Optimising Alumina Feeders in Aluminium Smelting Pots

University of Wollongong
Research Online
University of Wollongong Thesis Collection University of Wollongong Thesis Collections
1995
Optimising alumina feeders in aluminium smelting

pots
J. P. Kissane
University of Wollongong
Recommended Citation
Kissane, J. P., Optimising alumina feeders in aluminium smelting pots, Doctor of Philosophy thesis, Department of Mechanical
Engineering, University of Wollongong, 1995. http://ro.uow.edu.au/theses/1589
Research Online is the open access institutional repository for the

University of Wollongong. For further information contact the UOW
Library: research-pubs@uow.edu.au
OPTIMISING ALUMINA FEEDERS IN
ALUMINIUM SMELTING POTS
A THESIS SUBMITTED IN PARTIAL FULFILMENT OF TH

REQUIREMENTS FOR THE AWARD OF THE DEGREE OF
DOCTOR OF PHILOSOPHY
FROM
UNIVERSITY OF WOLLONGONG
BY
J.P.KISSANE, B.E., MTEAUST, CPENG
DEPARTMENT OF MECHANICAL ENGINEERING
1995
This is to certify that this w o r k has not been submitted
for a degree to any other university or institution.
J.P.Kissane
ABSTRACT
Crustbreak/feeding systems in prebake aluminium smelting pots are critical to efficient

pot control as they contribute in a large w a y to the alumina balance in the pot.
Inefficient feeder and pot operation m a y cause emission of global warming gases,
severe cost penalties and m a y affect the health and safety of operating personnel This
study investigated the causes of failure and the performance of existing designs. Most
of the research was conducted at Portland Aluminium (Victoria) and contact was made
with about 30 smelters world-wide.
Methods have been developed to improve shot size accuracy, reduce costs, increase
crustbreak/feeder life and improved the safety of operators and tradespersons. N e w
designs were developed and two were subsequently patented in Australia, N e w Zealand
South Africa and the U.S.A.
Over 30,000 measurements were taken of dose mass on 7 different feeder designs.
Changing the shape of the dosing unit to accommodate better alumina flow rate reduced
the standard deviation of the shot size to 0.6% of shot mass for one of the patented
designs and 1.9% for the most commonly used pot feeder (compared to an original 7-
14%). These figures compare with 2.5% for the Aluminium Pechiney independent
design.
In the investigation into the mechanism of plunger wear and optimising materials, o
10,000 measurements of plunger diameter were taken during feeder maintenance and
over 2,500 were taken in 3 monthly measurements in operating pots. S o m e 19
materials were tested. Plunger life at Portland increased from 1 year to 8 years (for
plant scale operation) and trials in operating pots achieved life up to 15 years for one
design.
Feeder life at Portland improved by over 80% and repair costs reduced by $1.4m per
annum. Feeder life improved from 9 months to 55 months. M a n y recommendations
have been shared at many smelters which has led to significant benefits in improving
the existing crustbreaker/feeder units used across the world.
ACKNOWLEGDEMENTS 1
ACKNOWLEDGEMENTS
When people consider improving pot feeders, they often think the objectives are to
improve life and reduce maintenance costs. These are desirable, but the main benefit is
to improve the operation of the pot. A repeatable alumina dose is critical to control of
pots, so a good feeder helps the people w h o control the pot to optimise its operation and
reduce generation of potentially global warming gases. Because the feeder is only one
part of the equation in pot operation, it is not clear as to h o w good a feeder needs to be
to make a significant effect on h o w well a pot runs; but poor feeders make pots run
poorly.
A long life feeder has a significant effect on the exposure of people to falls, heat,
and fumes, so some of the main customers in this investigation were the operators and
tradespersons. These latter two groups never lost sight of what was the targets were and
that ultimately this would help them to work safer in a better environment. These were
major incentives to continue w h e n the going got tough.
To complete a study such as this, a team of people need to work together. The bulk of
the "hands on" work was done by operators in Portland potrooms and tradespersons
from K e m p e Project Engineers, Parker Hannifin and Portland Air and Hydraulics.
Their assistance and dedication to the strange and repetitive jobs they were asked to do
were invaluable in unmasking the vagaries of pot feeders.
Although many people assisted in some way, it would be inappropriate for me not to
thank Mike Wakefield for contributing a large part to the data collection on plunger
wear. His patience is exemplary. Thanks also to all the "hands on" people: Phil
Harvey, Geoff Dover, D o u g Lucas, and Kevin (Bronco) Bevis for co-ordinating and
recording many tests; Brett Bampfield and Robert Logan for their often hot work on
operating pots; Jack Cholewski for co-ordinating the shot size testing and design
ACKNOWLEGDEMENTS II
changes; Neville Stutchberry for conducting the tests for the pneumatics investigation;
Greg Mansfield for the tedious hot job of monitoring the cylinder trials.
The hardest job probably was the tedious brain-deadening operation of the shot size
testing...about 30,000 individual mass determinations. Tony Cannon, S a m Kelly and
D o u g Lucas mainly had this unenviable task. M a y b e the most frustrating job was for
Val Ellis and Cathy Enscoe for trying to read m y writing and translating it into m y
earlier reports.
There were also a number of people who tried valiantly to improve my engineering
knowledge: Alf Jones, Gary Nauer, and Greg Noonan for trying to teach a chemical
engineer something about mechanical engineering and pneumatics; Matthew Langmaid
and Colm Fitzpatrick for their assistance in turning the A 3 feeder ideas into reality;
Hans K e m p e and H u g h Stark for their valuable advice and knowledge on metallurgy
and fatigue; Andrew Morphett for his assistance as a process and statistics advisor; and
Fethon N a h o u m for his willing help in data analysis and for optimizing feed control so
that the hardware improvements were not wasted.
Peter Arnold (my PhD supervisor) was a ready mentor in this development and gave me
the inspiration and direction to complete it. T o read this document cover to cover is a
feat, but to read it several times and to find something to write red comments about on
every page w a s indeed an accomplishment. T h e quality of the final version is a
testament to his patience.
I would also like to thank my opposite numbers from other plants for discussing on
phones and faxes the feeder improvement project. Their time and patience during long
telephone calls has been appreciated. Probably Telecom and A T & T profits have
dropped n o w the project is complete. The success at other plants is a credit to your
teamwork and dedication. Although I helped place a bit of focus on opportunities for
improvement, the hard work in the plant is what makes results.

ACKNOWLEGDEMENTS m
Portland and Alcoa have been most generous in allowing m e to use data and facilities to
conduct this research. In particular, I want to thank m y Potroom bosses...Trevor
Minchinton, Gerard Waller and Trevor Adams. They still supported m e even at times
when results were not coming as fast as they would have liked. The vision of the late
David Judd and Geoff Hayward to allow people such as myself to expand our horizons
has given m e opportunities to learn things and to influence the industry which seldom
occurs for engineers. The support of Alcoa U S A Primary Metals people Keith Isakson,
Bob Seymour and Dick Taylor gave m e the tenacity to go on and ensured that others
contributed to and benefited from feeder research.
Lastly, I would like to thank my wife Linda and my children Patrick, Anne, Katie a
Ben. Ben did a great job in cleaning up the scanned images which took a lot of time.
M y family listened patiently to all m y "feeder stories" and supported m e when I needed
it, and often lived without m e (physically or mentally) on many occasions. It has not
been an easy job putting up with m e on one of m y crusades.
That's a lot of people, but 1 had a lot of help. Thanks.
JimK.
NOTE TO THE READER IV
NOTE TO THE READER
Due to the confidential nature of some of the data presented in this document, this the
is restricted from public access except under written approval from the Technical
Manager of Portland Aluminium.
The whole of this work was initiated, controlled and co-ordinated by the author, but
considerable input has been given by others. Most of the test work was conducted at
Portland Aluminium by the author, operators, and tradespersons. Major contributions
that were the work of others has been specified in the text with the n a m e of the person
involved.
To simplify references to plants, any plant specified has been coded throughout the tex
as follows:
T integrated feeder plant

D independent feeder plant
Appendix 10 lists the plant names against the codes.
To assist with brevity and confidentiality, data or opinions of individuals from variou
plants or organisations have not been cited unless the comment is critical to the
argument. However, personal correspondence and drawings are available if required to
validate comments. Data from smelters or from organisations have been taken first
hand from a suitable representative of that smelter or organisation. If assumptions or
interpretations of the work of others are made, they have been noted accordingly.
J.P.K.
TABLE OF CONTENTS V
OVERVIEW
Page
Abstract
Acknowledgements I
Note to Reader IV
Overview V
Table of Contents VI
Index of Figures XV
Index of Tables XIX
Index of Appendices XXII
Glossary XXIU
Chapter 1 Introduction and Literature Survey 1

Chapter 2 Shot Size Analysis 61
Chapter 3 Dose Delivery 119
Chapter 4 Pneumatics 143
Chapter 5 Plunger Wear 169
Chapter 6 Plunger Buildup Analysis 230
Chapter 7 Pneumatic Cylinder Investigation 254
Chapter 8 Extending Feeder Unit Life 285
Chapter 9 Cost Reduction 328
Chapter 10 Feeder Development 354
Chapter 11 Impact of Research on Alcoa and Other Smelters 373
Chapter 12 Summary Of Results 391
References 395
Appendices
TABLE OF CONTENTS VI
TABLE OF CONTENTS
Page
CHAPTER 1 INTRODUCTION AND LITERATURE SURVEY
1.1 W H Y ALUMINA FEEDER TECHNOLOGY is IMPORTANT 1
1.2 LITERATURE SURVEY 7
1.2.1 Aluminium Smelting Technology. 7
1.2.2 Objectives ofAlumina Feed Control 13
1.2.3 Initial Prebake Feeding Systems 17
1.2.4 Integrated Feeders 19

1.2.5 Independent Feeders 24
1.2.6 Comparison ofIntegrated and Independent Feeder Designs 27

1.2.7 ConversionfromIntegrated to Independent Feeding 29
1.2.8 Continuous Feeders 33
1.2.9 Aeration Feeders 34

1.2.10 Computer Control. 36
1.2.11 Pot Alumina Supply Systems 38

1.2.12 Plunger Developments 39
1.2.13 Dissolution ofAlumina 42
1.2.14 Alumina Dose Design and Accuracy 44

1.2.15 Pneumatic Air Cylinders. 45
1.2.16 Pneumatic Systems 50
1.2.17 Other Feeder Developments 52
1.3 DEVELOPMENT OF THIS RESEARCH 54
1.3.1 Research Objectives 54
1.3.2 Optimizing Existing Feeder Designs 54
1.3.3 New Feeder Designs 56

1.3.4 Across Alcoa Optimisation 58
1.3.5 Document Layout 59
CHAPTER 2 SHOT SIZE ANALYSIS
2.1 SUMMARY 61
2.2 INTRODUCTION 62
TABLE OF CONTENTS VII
PAGE
2.3 PROCEDURE 65
2.3.1 Dwell Time 65

2.3.2 Dosing Unit Shape 67
2.3.3 Stroke Time 69
2.3.4 Validity of Off-Site Testing.
2.4 O F F - S I T E T E S T I N G O F S H O T SIZE 73
2.4.1 PortlandAEDD 73
2.4.2 T6 Integrated Feeder 78
2.4.3 Tl 1 Integrated Feeder 80
2.4.4 Portland A2 Feeder 80
2.4.5 A3 Feeder and Sequential Feed. 84
2.4.6 Decision on Use ofSpool Inserts 85
2.5 SENSITIVITY A N A L Y S I S O N S H O T S I Z E 87
2.5.1 Bulk Density of Alumina 87
2.5.2 Flow Through Inaccuracies in Dosing Systems 8
2.5.3 Effect of Springs 89
2.5.4 Effects of Cylinder Design on Shotsize 9

2.5.4:1 Description of Cylinder Operation 91
2.5.4:2 Effect of Cushioning on Shot Size 91
2.5.4:3 Effect of Seal Material/Quality 96
2.5.5 Effect ofDwell Time 99
2.5.6 Effect of Height ofAlumina in the Hopper 9

2.5.7 Effect ofInsert Stroke Length 9
2.5.8 Effect ofSpool Manufacture Quality. 100
2.6 O N - S I T E T E S T I N G O F S H O T SIZE 102
2.6.1 Til-ColdPots 102
2.6.2 Portland - Cold Pots. 102
2.6.3 Portland-Operating Pots 103
2.6.4 Portland - A3 Feeders 108
2.7 EFFECTS OF PNEUMATICS ON SHOT SIZE 109
2.8 DIRECT FEED INTO THE HOLE Ill

2.9 FACTORS AFFECTING SHOT SIZE ACCURACY IN POTS 115
2.10 MAJOR FINDINGS FROM SHOT SIZE ANALYSIS 117
2.11 RECOMMENDATIONS FROM SHOT SIZE ANALYSIS 118

TABLE OF CONTENTS VIII
Page
CHAPTER 3 DOSE DELIVERY
3.1 SUMMARY 119
3.2 INTRODUCTION 120
3.3 IMPACT OF ANODE COVER ON HOLE BLOCKAGE 121

3.4 CRUST BREAKAGE TECHNIQUES 124
3.4.1 Kinetic Energy 127

3.4.2 Pressure 127
3.4.3 Stroke Time to Break the Crust 128
3.5 BLOCKED FEEDER SURVEY 131
3.5.1 Examination of Blocked Feeder Survey Data 131
3.5.2 New Set Carbon 132

3.5.3 Repeaters 135
3.5.3:1 Dwell Time Tune-out 135
3.5.3:2 Lack of Supply Pressure 136
3.5.3:3 Back Pressure 136
3.5.4 Low bath 138
3.5.5 Pointed Plungers 138

3.5.6 Ore Leaks 140
3.5.7 Feeder Location 140

3.5.8 Minimum Dwell Time Determination 140
3.6 RESULTS OF IMPROVEMENTS TO BLOCKAGE PREVENTION 141
3.7 MAJOR FINDINGS FROM DOSE DELIVERY INVESTIGATION 142

3.8 RECOMMENDATIONS FROM DOSE DELIVERY INVESTIGATION 142
CHAPTER 4 PNEUMATICS
4.1 SUMMARY 143
4.3 GENERAL OBSERVATIONS ON PRESSURE TRACES 147
4.4 M E T H O D USED TO SLOW FEEDERS 153
4.5 RESULTS 157
4.5.1 Location 157
4.5.2 Feeder Age 157
4.5.3 Mufflers 158

TABLE OF CONTENTS LX
Page
4.5.4 SpeedChange 159

4.5.5 Dual Dwell Time 16
4.5.6 Pressure Required to Break a Hole
4.6 A I R SUPPLY OPTIONS 163
4.7 G R A D U A L INCREASE IN D W E L L T I M E 167
4.8 M A J O R FINDINGS F R O M P N E U M A T I C S INVESTIGATION 168
4.9 R E C O M M E N D A T I O N S F R O M P N E U M A T I C S INVESTIGATION 168
CHAPTER 5 PLUNGER WEAR
5.1 SUMMARY 169

5.2.1 Pattern ofPlunger Wear.
5.2.2 Wear Rate Measurement. 171

5.2.3 Theory ofPlunger Wear Mechanism 1
5.2.4 Procedure for Analysis ofPlunger Wear Mechanism
5.3 VARIABILITY O F P L U N G E R W E A R R A T E 178
5.3.1 Overall Observations 17

5.3.2 Effect ofSingle Batch Variability on Plunger Wear Rate
5.3.3 Effect ofBatch-to-Batch Variability on Plunger Wear Rate
5.3.4 Effect ofDiameter on Plunger Wear Rate

5.3.5 Effect ofBetween-Pot and Between-Plant Variability on Plunger Wear Rate
5.3.6 Effect of Within-Pot Variability on Plunger Wear Rate
5.3.7 Effect of Location Within a Pot on Plunger Wear Rate

5.3.8 Recommendations for Data Analysis of Plunger Wear Rates
5.4 TESTING C O R R O S I O N T H E O R Y 193
5.4.1 Wet Time 194

5.4.1:1 Wet Time per Crustbreak 194
5.4.1:2 Crustbreak Frequency 197
5.4.1:3 Cumulative Wet Time 201
5.4.2 Penetration Distance in Bath
5.4.3 Bath Temperature 206
5.4.4 Plunger Preheat 20
5.4.5 Anode Effect Control Strategy
5.4.6 Plunger Mass 21
5.4.7 Plunger Buildup 21

TABLE OF CONTENTS X
Page
5.5 TESTING EROSION THEORY 211
5.5.1 Plunger Speed 211

5.5.2 Crust Hardness 213
5.5.3 Plunger Hardness 213
5.5.4 Plunger Shape 215
5.6 PLUNGER MATERIAL 216

5.6.1 General Smelter Experience 216
5.6.2 Kaiser Plunger Wear Experiment 218
5.6.3 Portland Experience 220

5.6.3:1 Cast Iron 220
5.6.3:2 Inconel 222
5.6.3:3 Niresist and Incoloy 222
5.6.3:4 Stainless Steels 223
5.6.3:5 Silicon Carbide 225
5.6.3:6 Alloy Coated Cast Iron 226
5.7 CONCLUSIONS ON PLUNGER W E A R MECHANISM 226
5.8 MAJOR FINDINGS FROM PLUNGER W E A R ANALYSIS 229
5.9 RECOMMENDATIONS FROM PLUNGER W E A R ANALYSIS 229
CHAPTER 6 PLUNGER BUILDUP ANALYSIS
6.1 SUMMARY 230

6.3.1 Dag Frequency 235
6.3.2 Plunger Type 235
6.3.3 Dags Removed. 235
6.3.4 Anode Effects 236
6.3.5 Liquid Levels 236
6.3.6 Location in Pot 236
6.4 CONCLUSION ON CAUSES OF DAGS 239

6.5 FACTORS AFFECTING PLUNGER TEMPERATURES 240
6.5.1 Pot Temperature 240
6.5.2 Plunger Temperature Over Time 240
6.5.3 Hole Size and Flame 242
6.5.4 Number of Holes Open 242
6.5.5 Plunger Length and Penetration 243

TABLE OF CONTENTS XI
Page
6.5.6 Proximity ofPlunger to Hole 244
6.5.7 Wet Time 246
6.6 D A G PREVENTION A N D CONTROL 249
6.6.1 Plunger Coolers 249

6.6.2 Liquid Level Control 249
6.6.3 Plunger Penetration Depth 251
6.6.4 Pot Gas Venting 251
6.6.5 Reduced Anode Cover Depth 252
6.7 M A J O R FINDINGS F R O M PLUNGER BUILDUP ANALYSIS 253
6.8 RECOMMENDATIONS FROM PLUNGER BUILDUP ANALYSIS 253
CHAPTER 7 PNEUMATIC CYLINDER INVESTIGATION
7.1 S U M M A R Y 254
7.3 CYLINDER TRIALS 254
7.3.1 Introduction 254

7.3.2 Procedure 254
7.3.3 Results. 259
7.3.3:1 Sensitivity Analysis 259
7.3.3:2 Teflon Versus Viton Seals 263
7.3.3:3 Temperature 263
7.3.3:4 Speed 264
7.3.3:5 Eccentricity 264
7.3.3:6 Grease 265
7.3.3:7 Piston rod/shaft connection 265
7.3.3:8 Top cushion length 265
7.3.3:9 Spring design 265
7.3.3:10 Deceleration rate 266
7.3.3:11 Air leak rates 266
7.3.3:12 Insulation 268
7.3.4 Cylinder Trial Summary 268
7.5 LUBRICATION INVESTIGATION 254

7.5.2 Procedure 257
7.5.3 Kepner Tregoe Decision Analysis Procedure 258

TABLE OF CONTENTS XII
Page
7.5.4 Discussion 260
7.5.4:1 Benchmarks 260
7.5.4:2 N o Lubrication 261
7.5.4:3 Lubricant Options 263
7.5.5 Conclusions and Plant Scale Results 264
7.11 MAJOR FINDINGS OF PNEUMATIC CYLINDER INVESTIGATION 265

7.12 RECOMMENDATIONS FROM PNEUMATIC CYLINDER INVESTIGATION 265
CHAPTER 8 EXTENDING FEEDER UNIT LIFE
8.1 SUMMARY 285

8.3 CAUSES OF FEEDER FAILURES 297
8.3.1 Ore Leaks 288
8.3.2 Pointed and Short Plungers 290

8.3.3 Arcing 290
8.3.4 Fatigue ofPiston Rods and Spools 297
8.3.5 Cylinder Bypass 291
8.3.6 Pot OffLine 292

8.4 R O D SEALS 292
8.4.1 Causes ofRod Seal Failure 292

8.4.2 High Temperature Exposure Control 294
8.4.2:1 Pot and Feeder Design 294
8.4.2:2 Normal Operation 295
8.4.2:3 Hot Pots 296
8.4.2:4 N e w Pots 297
8.4.2:5 Rod Seal/Bush Orientation 297
8.4.3 Speed. 298
8.4.4 Eccentricity - 298

8.4.5 Rod Seal Material 301
8.4.6 Rod Seal Design 303
8.4.7 Rod Wiper Design 306
8.4.8 Venting Rod Seal Air Leaks 307
8.5 PISTON SEALS 310
8.6 BROKEN PISTON RODS 310
8.6.2 Fatigue Analysis 311

TABLE OF CONTENTS Xin
Page
8.6.3 Results and Conclusions ofFatigue Tests
8.7 A R C I N G 315
8.7.1 Insulation of the Assembly
8.7.2 Cylinder Insulation J/
8.7.3 Other Methods to Control Arcing
8.8 D O S I N G U N I T FAILURES 319
8.8.1 Broken and Worn Spools 319

8.8.1:1 Integrated Feeders 319
8.8.1:2 Independent Feeders 320
8.8.2 Spool Jamming 321
8.8.3 Spring Failure 323
8.9 MAJOR FINDINGS OF EXTENDING FEEDER UNIT LIFE 326
8.10 RECOMMENDATIONS OF EXTENDING FEEDER UNIT LIFE 326
CHAPTER 9 COST REDUCTION
9.1 SUMMARY 327

9.3 PLUNGER SAVINGS 330

9.4 AIR SAVINGS 337
9.4.1 Ore leaks and Air Leaks 33

9.4.2 Air Leak Test Method 33
9.4.3 Boyne Feeder Maintenance Strategy 3
9.4.4 Non-routine Cylinder Overhaul Strategy
9.5 SAVINGS F R O M R E U S E D F E E D E R S 345
9.6 F E E D E R T R A C K I N G S Y S T E M S 346
9.6.1 The Purpose of a Feeder Tracking System
9.6.2 Portland Feeder Reporting System
9.6.3 Tracking Systems and Quality Control.
9.7 MAJOR FINDINGS FROM COST REDUCTION 353
9.8 RECOMMENDATIONS FROM COST REDUCTION 353
CHAPTER 10 FEEDER DEVELOPMENT
10.1 SUMMARY 354
10.2 PROJECT HISTORY 355

TABLE OF CONTENTS XTV
Page
10.3 A2 FEEDERS, DRIBBLE FEED A N D PULSE C H U T E 358
10.4 MULTIPLE D C V S PER P O T OPTIONS 359
10.4.1 Split Header 36

10.4.2 Sequential Feed. 3
10.5 A3 FEEDER 362
10.5.1 A3 Pot Performance 362

10.5.2 Possible Use in Other Smelters
10.5.3 A3 Design Features 365
10.6 E N D STROKE SENSING 367
10.7 V A L V E A N D CYLINDER DESIGN 369
10.8 INTEGRATED VERSUS INDEPENDENT FEEDERS 370
CHAPTER 11 IMPACT OF RESEARCH ON ALCOA AND OTHER SMELTERS
11.1 SUMMARY 373
11.3 PORTLAND RESULTS 374
11.3.1 Plant Performance 374

11.3.2 Improvements for Health 3
11.3.3 Worker Involvement. 383
11.2 ACTIONS AT A L C O A SMELTERS 384
11.3 ACTIONS AT N O N - A L C O A SMELTERS 390
CHAPTER 12 SUMMARY OF RESULTS
12.1 RESEARCH OBJECTIVES 391
12.1.1 Feeder and Pot Performance 39

12.1.2 Feeder Development. 392
12.1.3 Conclusions on Achievements Against Objectives
12.2 FUTURE W O R K 393
12.3 CONCLUDING R E M A R K S 394

INDEX OF FIGURES XV
INDEX OF FIGURES
Page
CHAPTER 1
1-1 Portland Aluminium Smelter

1-2 Portland Aluminium Pot 3
1-3 Portland Pot Schematic 9
1-4 Potroom Layout 10
1-5 Aluminium Smelting Pots 12
1-6 Relationship between Pot Resistance and Bath Alumina Concentration 13
1-7 Early Pot Feeding System 18
1-8 Integrated and Independent Feeders 20
1-9 Integrated Feeder Operation ( A E D D ) 21
1-10 Crater below Integrated Feeder 22
1-11 A E D D Feeder Assembly 23
1-12 Pechiney AP18 Feeder 25
1-13 Pechiney AP18 Pot Showing Independent Feeder 26
1-14 A 2 Feeder Operation 30
1-15 A 3 Feeder Operation 31
1-16 Sequential Feed A E D D 32
1-17 Aeration Feeders 35
1-18 Pulse Chute Operation 37
1-19 Plunger End Stroke Sensing 41
1-20 Pneumatic Cylinder Parts 46
1-21 Portland Pot Feeder System 51
1-22 Portland Feeder Designs 57
CHAPTER 2
2-1 Off-Site Shot Size Test Rig 63

2-2 Portland A E D D Shot Size Graphs 66
2-3 Spool Insert Assembly 68
2-4 Parker Standard Checks 71
INDEX O F FIGURES XVI
Page
2-5 Effect of Cylinder Design on Portland A E D D Shot Size 74
2-6 Portland A E D D Repeatability Ex Plant. 76
2-7 Effect of Insert Stroke on Portland A E D D Shot Size 77
2-8 T6 Feeder Shot Size 79
2-9 Til Feeder Shot Size 81
2-10 Til #2 Feeder With and Without Insert 82
2-11 A 2 Feeder Shot Size 83
2-12 Effect of Temperature on Springs 90
2-13 Cylinder Cushioning Operation 92
2-14 Effect of Cushioning on Portland A E D D Shot Size 95
2-15 Effect of Spool Design and Seal Age 97
2-16 Spool Wear Points 101
2-17 A 2 Feeder with Double Outlet Chute 114
CHAPTER 3
3-1 Plant Crustbreaker Kinetic Energy and Pressure Comparison 126

3-2 Portland Feeder Mufflers 138
CHAPTER 4
4-1 Pot Pressure Sensing Points 145

4-2 Stroke Testing Unit 146
4-3 Pressure Traces - Old versus N e w Feeder 148
4-4 Pressure Traces - Speed Changes 156
4-5 Pressure Traces - On-Site and Off-Site Testing 151
4-6 Piston Rod Failure History 160
4-7 Pressure Traces-Number of Feeders per D C V On-Site 164
CHAPTER 5
5-1 Typical Worn Plungers 170

5-2 Trial Plungers 172
5-3 Plunger Diameter versus Life 178
5-4 Cast Iron Plunger Wear by Batch Number 183
5-5 Pot-to-Pot Plunger Wear Variation I88
19
5-6 Within-Pot Plunger Wear Variability °
INDEX O F FIGURES XVII
Page
5-7 Cast Iron Plunger Wear Rate versus Wet Time per Crustbreak 195
5-8 Cast Iron Plunger Wear Rate versus Crustbreak Frequency 197
5-9 Effect of Crustbreak Frequency Change on Plunger Wear Rate 198
5-10 Plunger Wear by Month Installed 200
5-11 Effect of Cumulative Wet Time on Cast Iron Plunger Wear Rate 201
5-12 Plunger Wear Ratios 204
5-13 Calculated Plunger Life 205
CHAPTER 6
6-1 Dagged Plunger 231

6-2 Plunger Temperatures for 945°C Pots by Location 238
6-3 Plunger Temperature Profiles 248
CHAPTER 7
7-1 Cylinder Trial Equipment 257

7-2 Cylinder Trial Rig 258
7-3 Performance of Test Cylinders in Cylinder Trial 263
7-4 Cylinder 22A Rod Seal Leak Rate 269
7-5 Effect of Amount of Grease on Rod Seal Life 281
CHAPTER 8
8-1 Cylinder Rod Seal and Bush Designs 293

8-2 Seal Material Friction Coefficient 302
8-3 Weibull Analysis for Parker Rod Seals 304
8-4 Boomerang Brackets and Flange Insulation 308
8-5 Mounting Bolt and Bracket 309
8-6 Sketch of Broken Piston Rods 312
8-7 Weibull Analysis by Quarter of Piston Rods 314
8-8 Frequency of Jammed Objects 316
8-9 Weibull Analysis of Springs 325
CHAPTER 9
9-1 Bimetal Plunger 331

9-2 Calculated Plunger Cost per Tip-Day at A$75/Repair 335
9-3 Calculated Plunger Cost per Tip-Day at A$600/Repair 336
INDEX O F FIGURES XVHI
Page
9-4 Ore Leak History 339
9-5 Cylinder Overhaul Strategy Costs 344
CHAPTER 10
10-1 Multiple DCV's per Pot Feed Options 361

10-2 Independent Feed Option 361
10-3 A3 and Sequential Feeders 364
CHAPTER 11
11-1 Portland Development Implementation 375

11-2 Portland Feeder Changeout History 378
11-3 Life of Portland Feeders by Month Installed 380
11-4 Portland Feeder Performance - Summary 381
11-5 Alcoa Pot Feeder Performance-By Plant 382
11-6 Alcoa Feeder Performance - Summary 387
INDEX OF TABLES XLX
INDEX OF TABLES
Page
CHAPTER 1
1-1 Smelters of the World 2

l-II NonAlcoa Point Feeder Systems 4
1-IJJ Duties of a Pot Feeder 5
1-IV 1991 Alcoa Feeder Costs (A$) 6
1-V Causes of Integrated Feeder Removal (1991) 16
1-VI Smelters Contacted in this Research 48
1-VTI Karmoy Cylinder Repair History 49
1-Vm Alcoa Feeder Recommendations 59
CHAPTER 2
2-1 Summary Results of Off.Site Shot Size Tests 73

2-II A 3 / A E D D Shot Size Comparison (Room Temperature) 84
2-UJ Anode Effects for Pots with/without Spool Inserts 86
2-IV Spring Force Comparison 98
2-V Shot Size Testing on N e w Pots 103
2-VI Summary of Shot Size Tests on an Operating Pot 105
2-VII Repeat Shot Size Runs using Same Feeder on an Operating Pot 106
2-VM Shot Size Comparison of Dwell Time and Speed on Feeders in an
Operating Pot 107
2-LX A 3 / A E D D Shot Size Comparison (Operating Temperature) 109
2-X Comparison of Off Site versus Operating Pot Shot Size Tests for
the Same Feeder 110
2-XI Things that Affect Shot Size 116
CHAPTER 3
3-1 Techniques to Break a Hole in the Pot Crust 120

3-U Plant Crustbreaker Kinetic Energy and Pressure Comparison 125
3-ID. TimetoBreak a Hole at Zero Velocity 129
3-IV Causes of Blocked Feeder Holes at Portland 131
INDEX OF TABLES XX
Page
CHAPTER 4
4-1 Effect of Muffler and Speed Change-Location 1 154

4-U Effect of Muffler and Speed Change-Location 5 155
4-111 Dual Dwell Wet Times 162
CHAPTER 5
5-1 Parameters Potentially Affecting Plunger Wear 176

5-II Plunger Harness Trial Data 180
5-ffl Plunger Hardness Trial Data 182
5-IV Cast Iron Wear Rate Matrix 186
5-V Effect of Location on Plunger Wear 191
5-VI Plant Penetration Comparison 202
5-VTI Plunger Temperature Survey 207
5-Vni Effect of Plunger Speed and Dwell Time on Plunger Wear Rate 212
5-rX Plunger Material Comparison 214
5-X Effect of Tip Shape on Plunger Wear Rate 215
5-XI Kaiser Plunger Wear Tests 218
5-XJJ Plunger Materials Tested at Portland 221
5-XIH Effects of Plunger Wear Parameters for Cast Iron 227
CHAPTER 6
6-1 Dagged Feeder Survey Summary 234

6-II Link between Dags and Anode Effects 237
6-m Plunger Temperature Over a Three W e e k Period 241
6-IV Survey of Feeder Holes Open on Operating Pots 243
6-V Plunger Temperatures Relative to Chute Temperature 245
CHAPTER 7
7-1 Portland/Parker Cylinder Trial Data 262

7-U Portland/Parker Cylinder Trial Data Summary 271
7-III Recommendations from Cylinder Trial 272
7-IV Lubricant Decision Analysis Summary 275
7-V Final Lubricant Ratings 278
INDEX OF TABLES XXI
Page
CHAPTER 8
8-1 Causes of Failure of Rod Seals at Portland 292

8-II M a x i m u m Temperature Limits for Seal Materials 296
8-ni Plunger Side Deflection 300
8-IV A E D D Spring Forces 324
CHAPTER 9
9-1 Portland Plunger Comparison 333
CHAPTER 10
10-1 Pot Feed Alternatives 356

10-H Feeder Trial Data 357
10-m Advantages and Disadvantages of Independent Feeders 371
CHAPTER 11
11-1 Portland versus other Feeders 376

11-11 Comments from Alcoa Plants on Feeder Development 386
11-111 Actions at Non-Alcoa Plants 390
INDEX OF APPENDICES XX11
INDEX OF APPENDICES
Appendix 1 Examples of Shot Size Control Charts
Appendix 2 Off.site Shot Size Test Data
Appendix 3 Alumina Sizing and Density
Appendix 4 Bob Seymour Memo on Feeder Recommendations
Appendix 5 Dick Talyor Memo on Results of Feeder Research
Appendix 6 A2 U.S.A. Patent
Appendix 7 Pulse Chute Patent Application
Appendix 8 A3 U.S.A. Patent
Appendix 9 Diary of Changes
Appendix 10 Plant Codes

GLOSSARY PAGEXXIII
GLOSSARY
A2 Integrated feeder which feeds when the plunger is up;

invented by the author.
A3 Independent feeder invented by the author.
AEDD Alcoa integrated feeder designed by the Alcoa Equipment
Development Division.
AEPD Average "anode effect per day" per pot.
Airbum Combustion of an anode due to exposure of the surface to
air when at elevated temperature.
Alumina Aluminium oxide used for aluminium production.
Anode Positive electrode of carbon.based material suspended
from the superstructure into the molten bath. Conducting
rods of copper or aluminium join the superstructure to the
carbon.
Anode Cover Material that covers the anodes to prevent air burning the
carbon and to maintain heat balance in the pot.
Anode Effect (AE) When a rise in voltage occurs on a pot due to the depletion
of alumina in the bath to a level which the current can not
be carried by the reduction of alumina and reduction of
the electrode commences.
Assembly The dosing unit of a pot feeder.
Auxiliary Power In a smelter, power used which is not related to
manufacturing of metal. Generally refers to power used in
compressors, lighting and computers.
Bath Liquid electrolyte used in a pot. Comprised mainly of
cryolite but with low concentrations of alumina and
calcium.
Beehive Another name for a dag.
GLOSSARY PAGE XXIV
BHN Brinell Hardness Number. Unit of hardness used for

plunger wear assessment.
Blocked Feeders Blocked feeder hole in a pot causing buildup of alumina
from the feeder. Also called mounding.
Bridge Horizontal aluminium bar on the superstructure that
suspends the rods of the anodes.
Carbon Another n a m e for anodes.
Cathode The negative electrode forming the bottom of the pot.
M a d e of carbon-based material.
Chisel Another n a m e for a plunger.
(In) Control In statistical control. All values are within predicted
probability limits...the process is predictable.
Control chart A graph of values and a graph of variability between
values with horizontal lines showing acceptance limits for
statistical control.
Crucible Cylindrical vessel used to retrieve aluminium from the
pot. M a y contain over 20 tonne of molten metal which is
withdrawn from pots (tapped) by vacuum.
Crust Concrete-like hard layer above and between anodes in a
pot. M a d e of alumina and/or crushed bath.
Crustbreakers The pneumatic cylinder, shaft and plunger unit that creates
a hole in the crust for alumina to enter the bath.
C E (Current Efficiency) Percentage of efficiency of metal production from a given
amount of electricity relative to the theoretical conversion
rate.
Cylinder bypass Piston seal leaks causing air to bypass through the
cylinder.
Dag Buildup of bath material on the end of a plunger.
DCV Directional Control Valve which supplies air from an air
supply generally to an air cylinder. Activated by an
electrical signal, an internal spool or poppet has its

GLOSSARY PAGE XXV
direction changed by a solenoid when air needs to be
applied to the inlet ports and the outlet ports need to be

vented to the exhaust.
Design of Experiment Statistical technique to trial multiple parameters in a
disciplined manner to identify causes of variability.
Dwell Time Time from D C V movement when feeder (or crust breaker)
starts to m o v e down, to when the D C V moves to raise the
feeder (or crust breaker).
Dense Phase Pneumatic conveying system supplying alumina to pot
superstructure.
Dosing Unit The component of a pot feeder that measures and delivers
the shot of alumina. Also called the spool in the feeding
assembly.
Dough Ball Another name for a dag.
Downstroke Movement of the piston vertically down a cylinder
installed in a pot.
Duct F u m e duct.
Endothermic Reaction which requires energy to occur.
Energisers Spring or " O " ring type material that forces the seal
against the moving part. Generally positioned between
the seal and the fixed housing.
FEC Front end cartridge. R o d seal unit on a pneumatic
cylinder. Includes rod seal, rod bush, and wiper.
Feeder Alumina feeding unit. This document refers to a feeder as
the complete feeder/crustbreaking unit, but some plants
(especially those which use independent feeders) m a y

refer to a feeder as only the small pneumatic cylinder,
shaft and dosing unit that doses the alumina.
Foot Another n a m e for a plunger.

GLOSSARY PAGE XXVI
Fume Duct Cavity in superstructure for pot gases to vent. May also
refer to pipe from superstructure which transfers the gases
to the scrubbing systems.
Greens Newly manufactured anodes.
Hornet's Nest Another name for a dag.
H R Cast Iron Heat Resistant cast iron developed by Alcoa.
ICC Team In Control and Capable Team - co-ordinators of research
and implementation of improvements in Alcoa plants.
ID. Internal diameter.
Incoloy Trade name of a high chrome/nickel stainless steel alloy
used for plungers
Inconel Trade name of a high chrome/nickel stainless steel alloy

used for plungers
Independent Feeder Crustbreak and alumina dosing are operated by separate
pneumatic cylinders.
Integrated Feeder Crustbreak and alumina dosing are mutually operated by
one pneumatic cylinder.
Kidney Plate Valve on superstructure that isolates alumina entry into

the feeder.
LCL Lower Control Limit - usually 2.5% of values fall below

this limit.
Lights Colloquial expression for an anode effect. Refers to

illumination of a light bulb mounted on a pot when the
voltage rises above 6 volts during anode effect.
Location Position of feeders in the pot superstructure - #1 closest to
narrow (control panel) aisle and #5 closest to wide (tap)
aisle.
Mack Valve Colloquial expression used for a directional control valve
(DCV). Is a trade name for a DCV.
Metal Pad Layer of molten aluminium at the base of a pot.
Mounding Blocked feeder hole.

GLOSSARY PAGEXXVII
Muffler Exhaust silencer on a D C V .

Nicrofer Trade n a m e of a high chrome/nickel stainless steel alloy
used for plungers
O.D Outer diameter.
Oiler/Filter Unit mounted upstream of a D C V .
Ore Aluminium industry jargon for alumina.
Ore leak Unmetered alumina leak into the pot from the feeder or
kidney plate.
"O" ring Ring shaped seal used often as an energiser behind a rod
or piston seal.
Outliers Extreme values.
P A Spec Cast iron specification plunger developed in this study for

Portland Aluminium.
Pareto Chart Bar chart with values decreasing from left to right
Penetration Distance that the plunger is wetted by molten bath.
Pick Another n a m e for a plunger.
Plugger Another n a m e for a plunger or feeder/crustbreaker unit.
Plunger Metal chisel used to break a hole in the crust above the
liquid components of a pot.
Plunger wear rate Reduction in O.D. of the bottom edge of the plunger over
time.
Pot Electrolytic cell used for manufacture of aluminium.
Pot dressing Covering anodes in operating pots using a rake type
device; usually not related to anode setting.
Potline A n array of pots electrically connected in series.
Potlining Carbon material that lines a pot to protect the insulation
from corrosion.
Potroom Half of a potline.
Prebake pots Pots that have anodes baked prior to position in the pots;
alternative to Soderberg pots where the anode is formed
from a paste in the pot itself.

GLOSSARY PAGEXXVIH
PTFE Polytetrafluoroethylene; often called by the trade name

Teflon.
Reactors Systems to scrub fluorides from the pot vent gases.
Restrictor A 1 2 m m orifice plate in the air lines to and from a
Portland pot. Used to reduce stroke speed of the feeder
cylinder piston.
Rod The vertical bar attached to the carbon block of an anode.

This is suspended off the bridge in the pot.
Scheduled AE Anode effect which is programmed (scheduled) to occur
by manual intervention.
Set Installation of n e w anodes and removal of spent anodes.
Sequential Feed Each feeder in the pot operates at different times. For
example, order of feeding m a y be 3,1,5,2,4 in sequence
for a 5 feeder pot.
S.D Standard deviation.
Sigma O n e standard deviation of a sample.
Silencer Straight.through silencer on the D C V of the pot feeder air

supply system. Similar principle to ariflesilencer.
Spa Calcium fluoride used to control the hardness of the crust

above the molten electrolyte in the pot. Lower spa gives
harder crust.
Spool The measuring unit that measures and delivers the shot of
alumina. Also k n o w n as the dosing unit.
Statistical SignificanceW h e r e there is less than (usually) 9 5 % that something

occurred by chance and chance alone.
Stroke The movement of the piston in a pneumatic or hydraulic
cylinder from one end of the barrel to the other
Super Superstructure.
Superstructure Central storage vessel above the pot cavity. Holds

alumina, feeders and anodes. F u m e duct vents from
superstructure.
GLOSSARY PAGEXXLX
Tap Removal of metal from a pot via suction.

Tapout Pot leaks molten metal or bath and has to be cut out of the
process.
Teflon A trade name of PTFE; a high temperature slow wearing
plastic commonly used for seals on pneumatic cylinders.
Thermoset Hardenning of a material due to heat. Commonly found
for rubber "O" ring seals used as energisers on pneumatic
cylinders under extreme temperature excursions.
UCL Upper Control Limit - usually 2.5% of values fall above
this limit.
Upstroke Vertical movement of the piston from the low point.
Viton A trade name for a high temperature rubber, like
hydrocarbon material commonly used for seals on
pneumatic cylinders.
Wallbox Control cabinet on the narrow aisle side of each pot to
allow manual intervention of pot control, and to monitor
pot performance.
Weibull Analysis Logarithmic graph of failure data that is used to predict

failure rates and mechanism of failure. The Y axis shows
percent failed and the X axis shows life of the component
or part. A curve fit is conducted and the two critical
parameters that explain the curve are displayed viz. (3
(shape parameter) and TJ (age when 62.3% have failed).
When P is 1.0, failures follow an exponential relationship,

and when 3.5 the distribution is normal. In most of the
failures in this text, p is close to 1.0. (Refer Juran (96)
page 23-34 for more details).
Wet Time Time that the plunger is fully extended and potentially
immersed in molten bath.
Wiper Outer unit of the rod seal (front end cartridge) that wipes
material off the piston rod of pneumatic cylinders.

FIGURE 1 -1
PORTLAND ALUMINIUM SMELTER
Photo courtesy Portland Aluminium, Portland, Australia

CHAPTER 1 INTRODUCTION AND LITERATURE SURVEY PAGEI
CHAPTER 1
INTRODUCTION AND LITERATURE SURVEY
1.1 WHY ALUMINA FEEDER TECHNOLOGY IS IMPORTANT

World production of aluminium is about 20 million tonnes per annum (Table 1-1).
Alcoa is the largest manufacturer of aluminium in the western world with interests in
the U S A , Brazil, Europe and Australia. Their newest smelter is Portland Aluminium at
Portland in western Victoria, Australia which was commissioned in 1986 (Figure 1-1).
A key factor in maximising the efficiency of smelting alumina to aluminium is the

control of the concentration of the alumina in the molten electrolyte of the electrolytic
cell (commonly called a "pot") (Figure 1-2). The concentration needs to be kept within
a tight tolerance in the 2 % to 5 % range or the pot will rise in voltage (if too low) or
form sludge (if too high). The ideal range is 2-2.5%, though many pots operate closer
to 3-5%.
Low concentration causes starving of the pot. This is called an "anode effect". The
resulting high voltage causes a rapid rise in temperature which affects the heat balance.
Sludge (commonly called "muck") drops to the bottom of the pot, increases voltage
across the pot and the pot increases temperature rapidly. The closer the pot operates to
an anode effect, the more efficient and cost competitive the process but the more
sensitive it is to upsets. O n e can see that the pot operates on a knife edge of
concentration. Either side of this acceptable range, efficiency drops and profit margins
reduce.
The high voltages associated with low alumina concentrations are caused by the
generation of C F 4 and C 2 F 6 which are "green house" gases which m a y contribute to
global warning. Research work has shown that C F 4 and C 2 F 6 have about 5,000 and
CHAPTER 1 INTRODUCTION A N D LITERATURE S U R V E Y PAGE 2
10,000 times respectively more global warming capacity than carbon dioxide (1). For
these reasons, for virtually all smelters, a major target is to reduce anode effects. Large
smelters use 5 0 0 - 7 0 0 M W of power for production rates up to 500,000tpa. The
efficiency of the pot is the main contributor to this power demand as potlines and pots
consume 12-14kWhr/kg of metal produced which comprises up to 8 0 % of smelter
power usage.
T A B L E 1-1
SMELTERS OF THE W O R L D
No. of Smelters Capacity
100% Share (tpa x1000)
Russia (Government) 11 0 3008

Alcoa 8 6 1778
Alcan Smelter & Chemicals 13 7 1735
Reynolds Metals 4 5 1018
China (Government) 17 1 898
A m ax 1 4 773
Aluminium Pechiney (Government) 4 6 723
Venezuela (Government) 0 2 541
Tadzhikistan (Government) 1 0 520
Vereinigte Industrie Unternehmungen A 3 3 499
Maxxam Group Inc. 0 4 445
Alusuisse-Lonza Holding Ltd 2 3 349
Bahrain (Government) 0 1 344
Instituto Nacional De Industria (Gvmnt.) 0 3 336
Norway (Government) 0 5 334
Conzinc Riotinto Of Australia Ltd 0 3 275
Romania (Government) 1 0 270
Ohio River Associates 1 0 245
Dubai (Government) 1 0 244
Votorantim 1 0 240
Companhia Vale Doe Rio Doce 0 2 224
Hodgovens 2 1 220
EFIM-MCS 3 1 218
Noranda Mines Ltd 1 0 215
Austrian Industries A G (Government) 1 3 209
India (Government) 2 0 209
Kazakhstan (Government) 1 0 200
Others - ~ 4610
20340
Notes: (i) Data from Pawlek (94) updated in 1994.

(ii) "Capacity" is installed capacity but not necessarily at the level
that the smelter is currently operating.
(iii) "100%" refers to the smelters owned by the company and
"Shared" refers to the number of smelters that the company
has shares in.
CHAPTER 1 INTRODUCTION AND LITERATURE SURVEY PAGE 3
FIGURE 1 - 2
PORTLAND ALUMINIUM POT
Photo coutesy Portland Aluminium, Portland, Australia
To achieve optimum concentration in the cell, the control of alumina feed to the pot
critical. Almost all of this material is delivered by the alumina feeder.
Each pot has feeding equipment that varies in design depending on the pot design.
Modern plants use 1-5 point feeders that dose about l-2kg of alumina into the pot about
every 2-4 minutes. They are called "point" feeders as they deliver the dose into a
localised point, rather than on a large slot in the crust above the electrolyte. Alcoa
plants have 1-3 feeders per pot except for Portland which has 5 feeders per pot...the
largest number of point feeders in any pot in the world. Table l-II shows some feeder
systems for non-Alcoa plants.
TABLE 1-D
NON-ALCOA POINT FEEDER SYSTEMS
INSTALLED FEEDERS/ O.D. STROKE '

SMELTER POT CYLINDER PLUNGER VALVES
FEEDERS (mm) (mm)
D1 640 2 Atlas Copco 200 600 Stainless Atlas Copco
2 Mechman 200 600 Stainless Atlas Copco
D2 888 4 Atlas Copco 200 500 AISI 309 Atlas Copco
4 CPOAC 200 500 AISI309 Atlas Copco
D3 128 2 Atlas Copco 160 600 Alloy Casting Atlas Copco
D4 - 4 - 200 600 - -
D6 960 4 Atlas Copco 200 500 Z12CN Atlas Copco
D8 - 4 - 200 600 - -
D9 1920 4 Atlas Copco 200 500 Nicrofer Atlas Copco
4 Atlas Copco 200 500 Hastelloy X Atlas Copco
D10 1920 4 50%CPOAC 200 500 31 OSS Parker
4 50%Atlas Copco 200 500 31 OSS Parker
D13 684 3/2 Atlas Copco 160 - Cast Iron Atlas Copco
D14 720 4 - 200 600 NS24 -
D15 - 4 - 200 600 31 OSS -
D16 1152 4 Atlas Copco 160 500 31 OSS Atlas Copco
D17 - 4 Atlas Copco 200 - - -
D18 3600 4 Atlas Copco 160 - - Atlas Copco
D19 1056 4 50%Atlas Copco 200 - 309SS Atlas Copco
4 50% CPOAC 200 - 309SS -
D20 - 4 Atlas Copco 160 - - Atlas Copco
D21 - 4 Atlas Copco 160 - - Atlas Copco
T2 352 2 Atlas Copco 125 508 Cast Iron Ross
T3 328 2 Mechman 125 500 Alloycoating Mechman
T4 864 3 Lindberg 125 356 304SS Mac
T8 1920 4 Parker 125 470 Steel/304SS Parker
T5 768 3 Lindberg 100 356 304SS Mac
T14 768 2 Lindberg 100 356 304SS Mac
T10 240 2 Terry 125 356 Cast Iron Ross
T17 1080 3 5 0 % Scheffer 100 356 Cast Iron Mac
Notes: (i) "T" refers to integrated and "D" refers to independent feeder plants.
(ii)"-" refers to data not being available.
The capital and operating costs for feeders are a significant cost to smelters as the
up to 3,000 feeders in large modern plants at capital costs of up to $30,000/pot for
several hundred pots. The duty is arduous due to the high temperatures (up to 1,100 C ) ,
corrosive liquids and gases (cryolite, H F , H 2 S , S O 2 ) abrasive crust above the
electrolyte, and the need to provide safe and hygienic conditions for personnel in
operating and maintaining the equipment.
Pot feeder designs used in the modern smelters have basic requirements to:
(i) measure an accurate and repeatable dose.

(ii) deliver the dose every time it is required.
(iii) ensure the dose enters and mixes efficiently in the molten electrolyte (bath).
If these actions are carried out effectively, low frequency of anode effects should re
Table I-HI lists the main duties of a pot feeder. Note that capital/maintenance costs are
only one of the major aims of a good feeder. Table 1-IV shows typical maintenance
costs. These costs are not trivial, however, it is generally considered that these costs are
small in comparison to the greater needs of control of anode effects and pot efficiency.
If market prices of aluminium are low, these costs are important factors in the cost
competitiveness of a smelter.
TABLE l-III
DUTIES OF A POT FEEDER
ACCURATE MEASURE AN ACCURATE AND REPEATABLE DOSE (SHOT SIZE)
DOSE IS DELIVERED ENSURE ALL THE DOSE ENTERS AND MIXES EFFICIENTLY
IN THE MOLTEN ELECTROLYTE (BATH)
" RELIABLE DELIVER THE COMPLETE DOSE EVERY TIME REQUESTED
SAFE SAFE AND HYGIENIC OPERATION
COST EFFECTIVE MINIMUM OVERALL COST...CAPITAL AND MAINTENANCE
The effect of the design on the people w h o operate and maintain the pot feeders needs
special attention. Changing feeders in an operating pot results in greater exposure of
people to falls (up to 3 metres), heat, dust, and fumes as they operate the cell. Even
maintaining the feeders creates hazards due to manual handling of heavy equipment and
potential exposure to fluorinated alumina. This is an area of responsibility that
C H A P T E R 1 INTRODUCTION A N D LITERATURE S U R V E Y PAGE 6
engineering must address in the development of any piece of equipment. This facet of
engineering design is often significantly lacking.
TABLE 1-IV
1991 ALCOA FEEDER COSTS (A$)

PLANT # FEEDERS # FEEDERS AVE LIFE TOTAL $PER
INSTALLED CHANGED (months) COSTS CHANGE
(A) (B) (A/Bx12) ($ pax 1000)
PORTLAND 2040 2400 10 1523 635
T6 734 700 12 446 635
T7 1824 4794 5 2492 N/A
T9 1548 940 20 908 966
T11 564 200 34 148 738
T13 496 400 15 406 1015
T15 1936 940 24 908 966
T16 1700 880 23 457 520
D12 984 694 17 748 1077
ALL 11826 11948 12 8035 718
Notes: (i) '*' indicates this is an estimate.

(ii) "average $/change" does not include D12 as this is the only plant
which is not an integrated feeder plant.
(iii) Assumed A $ = US$0.65
(iv)" T" refers to integrated and "D" to independent feeders.
It is desirable, therefore, to optimise the efficiency of aluminium smelting cells for
several major reasons:
(i) potential global warming effect
(ii) cost control
(iii)occupational hygiene/safety.
1.2 LITERATURE SURVEY
1.2.1 Aluminium Smelting Technology
Aluminium is the third most plentiful metallic element in the earth's crust (7.3%) (
is found in 250 oxides exceeded only by oxygen and silicon. O n e of the oxide hydrates
(bauxite) is the dominating raw material for the production of aluminium. Extraction of
the metal has three main stages:
(i) mining of the ore (bauxite);

(ii) refining the ore to recover alumina (aluminium oxide); and
(iii) smelting the alumina to produce aluminium.
The refinery Bayer process is the prevalent method of refining bauxite to make
alumina. In this process, a caustic soda solution is used to dissolve the ore and alumina
crystallises as a 100 micron solid with generally a small particle size range.
(Descriptions of the mining and refinery processes are explained by Prider and
Grojotheim et al. in references 5, 6 and 7.)
The smelting process comprises the electrolytic reduction of alumina (A1203) in a bat
of molten cryolite (sodium aluminium fluoride, Na 3 AlF 6 ) using carbon anodes:
2A1203 + 3C + electricity = 4A1 + 3C02
As the current goes through the pot, electrons exchange between the disassociated
alumina in solution to produce aluminium:
Al+++ + 3e' = Al
Electrical current flows through the pot via anodes, through molten liquids of bath
metal, and out the carbon cathode base of the pot via a matrix of aluminium bus bars to
the next pot where the process continues in a similar manner.
Figure 1-3 shows the main features of the Portland pot which is typical of normal
prebake pots (though considerably larger; 1.5-3 times the size of most prebake pots). A
pot is basically a large tank which has a 100-3 0 0 m m layer of molten aluminium below
a 100-300mm layer of molten electrolyte (bath). A rectangular silo of alumina (the
superstructure) is centrally mounted for alumina feeders to dose alumina into the bath at
regular intervals. Carbon anodes are suspended on a large horizontal busbar (the
bridge) off the superstructure. These carbon blocks (of up to one tonne in size) have
either aluminium or copper vertical bars (rods) that are fixed to their top surface usually
with cast iron or pressed carbon material. The vertical "rods" are attached to the bridge
which moves up and d o w n by electric motors in response to the voltage across the pot
using automatic computer programs.
Up to about 260 pots can be in a potline using the same current path (Figure 1-4). Th
potlines are generally divided into two potrooms of equal length, though sometimes (as
in Figure l-4(a)) this m a y be under the same roof. Despite being of only half the length
of a potline, potrooms can be over a kilometre in length. Another feature of potlines is
the use of overhead cranes (one of which can be seen in Figure l-4(b). These are used
to service the pots and are used to change anodes (setting) and extracting metal
(tapping). In m a n y older potlines, cranes also are used to supply alumina to the pots by
hoppers contained in (or hung off) the cranes. In more recent potlines pneumatic
conveying or air fed systems are used instead of cranes for alumina supply of pots to
reduce the amount of spilled alumina and to reduce the usage of cranes. Retrofitting
dense phase systems is becoming more c o m m o n since the mid 1980's.
Aluminium smelting is a continuous process with the alumina being dissolved in the
bath. The aluminium (being denser than the 950-960°C bath) forms in the bottom of
the pots. Alumina concentration is controlled at 2-5%. Molten metal is tapped on a 1-2
day routine by use of vacuum tanks (called crucibles). Gases (mainly C 0 2 , S 0 2 and
H F ) are released from the solution.

n
>
§
P
§
FIGURE 1 - 4
POTROOM LAYOUT
(a) End to end prebake (left) and Soderberg(right) pots except for
3 pots on lower left which are prebake pots).
(b) Side by side prebake pots

Photos courtesy GA Metall AB, Granges, Sweden
The pots are arranged in a series of cells forming the potline, with each cell operating at
4-5 volts D C at a current of 60-3 OOkA. Direct current passes from carbon anodes
through the bath to the cathode of the cell and then to the anodes of the next pot (and so
on). Steel bars embedded in the cathode carry the current out of the pot while the pots
themselves are connected via an aluminium bus-bar system.
Generally there is a similar current per unit area (current density) of pots, so it is
to refer to the relative size of a pot by referring to the current applied to a potline. Most
potlines are 120kA to 200kA. T h e Aluminium Pechiney A P 3 0 and the Alcoa model
817 pot (Portland) are the only potlines operating close to 3 OOkA. These are the largest
pots operating at present and are about 1 0 m long by 5 m wide using anodes of about a
tonne mass. This compares with the earlier pots of only a few k A and anodes so small
that they could be carried by hand. There has been rapid growth of the industry since
Charles Martin Hall and Paul Heroult invented the process in 1886.
These cells are of generally two different design types (Figure 1-5).
(a) Soderberg pots
Currently, 30% of aluminium production is from Soderberg pots (mostly from CIS
countries) (8). These pots use a carbon based anode paste (which is a mixture of coke
and pitch) that is dropped into a steel casing hanging above the pot. The carbon paste is
baked in the pot itself by virtue of the heat coming out of the molten bath while vertical
or angled steel pins (or studs) hold the anode and bring the current into it. This type of
pot has poor fume control and generally poor efficiency. There are two types: vertical
(as shown in Figure l-5a) and horizontal stub (which is the steel conducter in the anode
paste).
(b) Prebake pots
These pots have the carbon paste baked prior to being placed in the pot. They have t
carbon based anode compressed or vibrated into a block which is then baked at about
1,100°C to precondition the 'green' anodes prior to placement into the pot. This results
in better pot fume control and the pots run at superior efficiency. These anodes are
changed at about 20-30 days as the carbon burns away with the oxygen generated from
the dissolution of the alumina. Carbon dioxide and other gases are driven off into
collection fume systems which are then routed to scrubbers to collect any fluoride in the
gas.
Soderberg pots were the main production units initially, but better designs of preba
pots have developed such that almost all n e w pots commissioned since the 1960's have
been prebake. A large number of potlines are n o w retrofitting prebake pots into
Soderberg lines or converting the Soderberg pots to point feed.
FIGURE 1 - 5
A L U M I N I U M SMELTING POTS
LIQUID PASTE
ALUMINA
BAKED A N O D E FROZEN LEDGE
LIQUID
ALUMINIUM
yj £\JM1I
(a) Soderberg
ALUMINIUM
RISER
FROZEN LEDGE
(b) Prebake
This document concentrates on crustbreaker/feeder units (generally called feeders) in

prebake pots.
1.2.2 Objectives of Alumina Feed Control
The control of the concentration of alumina in the bath is a critical factor in the
efficiency of a pot. A s the concentration drops, the voltage in the pot falls to a
m i n i m u m at about 2.5% w / w alumina in solution. Once this point is reached, there is a
very rapidrisein voltage (Figure 1-6). This rise in voltage is called an "anode effect".
FIGURE 1 - 6
RELATIONSHIP BETWEEN POT RESISTANCE

AND BATH ALUMINA CONCENTRATION
1.82 r
1.80 • -
o 1.78 -
X
m
E 1.76
x:
O
UJ
1.74 -
o
CO
CO
LU
1.72 -
1.70
1 3 4
ALUMINA (%)
It is desirable to run at the least number of anode effects as is possible in a day (or
days), as generation of anode effects creates rapid short term energy input into the pot.
This upsets the heat balance, current efficiency and subsequent power costs. However,
to run at too few anode effects, there is a risk that the pot has undissolved alumina
(muck) in the bottom of the pot which causes uneven current flow, unstable metal level
and loss of current efficiency. A mucky pot can become very hot and dangerous to
operating personnel and is also very inefficient. M u c k is a far worse evil than too many
anode effects, so smelters have traditionally programmed (scheduled) anode effects to
ensure that the alumina concentration does not get too high and create muck.
Recent concern that the gases given off during anode effects are potentially harmful
the atmosphere, and the findings of a close relationship of low anode effects to better
current efficiency has directed smelters to reduce anode effects yet still prevent m u c k
from forming. This is a difficult task as pots are operating m u c h closer to instability.
T o achieve these targets, better feeder efficiency and feeder life plus smart computer
feed control are required.
Hence, it is important to control the alumina concentration within a narrow range.

Feeding systems generally comprise 8 0 % of alumina input to a pot (with the balance
coming from the material that covers the anodes), so control of the feeding system is
very important to controlling pot performance.
To dose alumina into the liquid bath, a path has to be broken into the hard crust for
over the top of the anodes. This rock-like crust is generally a mixture of alumina and
solidified bath (called anode cover). The hard crust, even when broken, often reforms
quickly, so frequent breakage is required. A m a x i m u m time between feeds of less than
10 minutes is used in practice and feeding about every 3 minutes being typical.
Another complexity is that the bath is highly corrosive and dissolves most metals
rapidly. The abrasive nature of the anode cover and crust also tends to wear the plunger
away. Hence, the crustbreaking device must be capable of resisting this attack yet not
be uneconomic by being too expensive. Metals such as Inconel or high nickel based
materials are generally efficient materials, but costs are very high and considered by
many to be uneconomic. T o m a k e matters worse, the crustbreakers must withstand
high temperature cycles (100°C to 1,000°C every 3 minutes) and hydrogen fluoride gas.
The feeding device must have a long life to avoid exposure of people to heat, dust and
fumes during change-out. Corrosion of plungers contributes to contamination of
aluminium with iron or other elements in the material (especially if their life is
short... 1-2 years). For potlines of several hundred pots with 2 to 5 feeders per pot,
capital and maintenance costs quickly become significant if feeder life is low.
Table 1-V shows data on the causes of failure of feeders at 11 integrated feeder plant
This shows the variety of both the causes and the frequency of each type of failure
despite all plants having basically the same feeder design. Chapter 9 discusses causes
of failure and h o w to prevent them.
To get the most out of a pot horizontal area, the carbon area in the pot should be
maximised so m a x i m u m current can pass through the pot, thus, reducing the number of
pots and capital cost per tonne of aluminium. A s this happens, less area is available for
pot feeding. Also, there is a need to conserve energy as one of the highest costs of
operating a pot is the cost of power; breaking the crust and leaving open holes loses
energy, so is not desirable. Clearly, a balance of priorities is needed which results in
the duties of a pot feeder being difficult to achieve (Table I-in).
Although many papers have been published on dissolution of alumina, almost all
published literature on pot feeders or associated systems is contained in patents. A lot
of work is done on a plant-by-plant basis but not published. Hence, other smelters often
are not aware of developments in this competitive industry. N o papers have been traced
on the efficiency of operating feeders other than references in some papers by Pechiney
which provide a summary of patents (17) or general papers on the benefits of potlines
(19, 20).
TABLE l-V
CAUSES OF INTEGRATED FEEDER REMOVAL (1991)
AVERAGE BROKEN
OPERATING ANNUAL ORE POINTED POT OFF SHORT BROKEN NOT
SMELTERS LIFE ARCED PISTON
FEEDERS OVERHAULS LEAK PLUNGER LINE PLUNGER SPOOL FEEDING
(months) ROD
PORTLAND 2040 2400 10 28 17 40 1 11 3
T2 352 168 25 95 5
T6 734 700 13 15 60 10 10 15
T7 1824 4794 5 60 10 20 10
T9 1548 940 20 10 85 5
T10 240 310 9 80 10 5 5
T11 564 200 34 20 40 35 5
T13 496 400 15 15 75 10
T15 1936 940 25 80 20
T16 1700 880 23 10 60 20 5 5
T17 1080 364 36 75 25
ALL 12514 12096 12 44 12 13 13 8 5 3 2
SUMMARY OF ALL 11 SMELTERS SURVEYED
BROKEN
PISTON ROD
S% NOT
FEEDING
2%
BROKEN SPOOL
SHORT PLUNGER
8%
POT OFF LINE

13%
POINTED PLUNGER
12%
1.2.3 Initial Prebake Feeding Systems
The early feed systems were based on the feeding systems of the Soderberg potlines
which often drove a large toothed wheel along the side of the pot and poured alumina
onto the top of the holes (9 and 10). Figure l-7(a) illustrates one such design. This
method w a s difficult to control the unmetered alumina supply and m u c k often
developed which caused pot instability.
Because of the large number of existing Soderbergs (especially in the CIS), better
feeders are being developed that have similar features to prebake point feeders (14), use
of fluidised feeders (13), or fluidised airslides (14).
The next generation of feeding systems was a bar break (16,17). Figure l-7(b)
illustrates one such design. Along the centre of the pot a large guillotine, several metres
long, is used to break the crust. This device is operated by large air cylinders that, by
necessity, have to be located close to the dust, heat and fumes. The feeding system is
generally a dosing unit above the bar that discharges a dose of 30-70kg on top of the
slot in the crust.
The main problem with the barbeak design is the haphazard nature of alumina entry
into the bath. There is no guarantee that all the alumina reaches the bath and at what
rate. Also, the large area of the bar posed problems on economic materials of
construction for long life. Although better that the 'toothed wheel' of the Soderbergs,
bar break dosed too large a slug of alumina into the pot and pot efficiency suffered.
Reverdy (16) gives a good description of the patents of the Soderberg and barbreak
feeders in early prebake pot designs. M a n y operating plants still use barbreak feeders,
but some have converted or are converting to point feeders (e.g. Alusuisse (18) and
Comalco respectively).
In 1970, Pechiney patented an improved device for pot feed based on crustbreakers and
dosing into multiple drilled holes (15). The principle was in line with later point feeder
development, but the high cost of the large device and the need to use overhead cranes
resulted in the design not being pursued.
CHAPTER l INTRODUCTION AND LITERATURE SURVEY PAGE 18
FIGURE 1 - 7
EARLY POT FEEDING SYSTEMS
Adapted from Patent SU1301875 (Ref 9)
(a) Sidebreak
Adapted from Patent NO-143506 (Ref 17)
(b) Barbreak
The superior technology for alumina feeding is "point feeding" where an accurately
measured small dose (under 5kg) of alumina is dosed into a small hole (3 0-120mm
diameter) in the crust. The alumina is delivered into the pot via 1-5 dosing systems that
incorporate a crustbreaker and a dosing unit.
There are two basic types of point feeders (Figure 1-8):
Integrated feeders...one air cylinder used for dosing and crustbreaking in a

combined design; and
Independent feeders.separate cylinders for dosing and crustbreaking such that

the two functions operate independently.
Both types of feeder have a crustbreaker tip - called a plunger - which breaks a hole
the hard crust above the molten bath to provide a path for the alumina to get into the
bath. Plungers are called tips, pluggers, chisels or picks in various plants.
1.2.4 Integrated Feeders
A couple of designs were developed in the 1960's by Alcoa (Figure 1-9) and Reynolds
(21). In the integrated CÔA^ the downward movement of the crustbreaker releases a
premeasured volumetric dose onto the top of the plunger. The plunger is still in the hole
w h e n the alumina is released. W h e n the plunger moves upwards, most of the dose falls
into the hole. The alumina tends to build up a volcano shape around the broken holes
for integrated feeders (Figure 1-10). The alumina dose has to slip d o w n into the hole
after the plunger lifts out of the way. Thus, the feed is only indirectly reaching the bath
and is affected by the angle of the crater and blowback of gases out of the hole.
Integrated feeders generally use a 100mm to 150mm O.D. pneumatic cylinder which
activates the crustbreaker and dosing system as it strokes. They generally incorporate
high speeds to feed quickly the alumina out of the dosing unit and to develop high
kinetic energy to break the crust. This design is relatively cheap, simple and easier to
maintain than the barbreak, as the feeder can be lifted out of the pot by crane to repair
FIGURE 1 - 8
INTEGRATED AND INDEPENDENT FEEDERS
INTEGRATED INDEPENDENT
FIGURE 1 - 9
INTEGRATED FEEDER OPERATION (AEDD)
ALUMINA FROM
SUPERSTRUCTURE
ALUMINA DOSE
ASSEMBLY SEALING
FACE
DELIVERY
OF DOSE
CUP SEALING
FACE
PLUNGER UP
% %
PLUNGER DOWN
any failures. This type of design became the norm in new smelting lines until the
1980's when independent feeders were developed.
FIGURE 1 -10
CRATER BELOW INTEGRATED FEEDER
Note: This crater could have been the result of a blocked feeder hole.
The shape is similar to that of a normal feeder hole crater.
U S S R Patents (24 and 25) appear to be similar to the Reynolds/Alcoa integrated feeders
but incorporate baffles in the tank to push alumina into the dosing system. Baffles are
an attraction if the alumina is badly segregated, however, the more correct action should
be to minimise the segregation in the first place and avoid the extra capital cost of
equipment for each feeder.
The most popular integrated feeder is the Alcoa AEDD developed by the Alcoa
Equipment Development Division in the 1950s and 1960s (Figure 1-11) with the
Comalco feeder at Boyne Smelters being of similar concept but with a different feeding
mechanism.
FIGURE 1-11
AEDD FEEDER ASSEMBLY
SUPERSTRUCTURE
"TORPEDO TUBE"
PLUNGER SHAPT
-ASSEMBLY BODY
SPRING
VEN HOLES
KIDNEY PLATE*
BASE OP
SUPERSTRUCTURE
ALUMINA DOSE^
INSERT
ILATION BAND
GUIDE
HARDENED OR
BRONZE INSERT
SUPERSTRUCTURE
PLUNGER PIN-
PLUNGER
1.2.5 Independent Feeders
Independent feeders separate the function of crustbreaking and feeding though they
kept located in the same area of the pot. A large air cylinder (150-200mm O.D.) is used
as a crustbreaker and a small cylinder (40-50mm O.D.) is used to dose alumina.
In any pot, it is not necessary to break a hole every feed as the crust takes minut
hours to cover the hole. A s opposed to the integrated feeder that can only dose when
the crustbreak mechanism is extended, the independent feeder need only break a hole
when it is deemed prudent. Plants can break a hole every 2-3 feeds, thus reducing the
cycles of the cylinder and frequency of penetration of the plunger into the bath.
Maintenance costs and improved feeder life are evidenced. Reverdy (26) maintains that
another major advantage of independent feeders is direct feed into the hole and less air
usage. (Reverdy is a Director of Aluminium Pechiney.)
The most common independent feeder is that based on Pechiney's patents of

Gerphagnon and Wolter (22) and Bonney and Gerphagnon (23) (Figure 1-12 and 1-13).
The design is used by Pechiney in their A P I 8 (180kA) pots and with a revised model in
their A P 3 0 (3 OOkA) pots. Several other companies have adopted similar styles without
infringing patent rights e.g. Hydro Aluminium, V A W , Reynolds, Granges Metall,
Aluar. This type of feeder is rapidly catching up to Alcoa's integrated feeder in
popularity and will soon be the most popular feeder after commissioning of the
A L U S A F smelter in South Africa.
Alusuisse (Swiss Aluminium) recently have constructed high efficiency retrofit

packages for feeding and dense phase supply of alumina to the pot (18). S o m e of their
plants use an independent feeder that has a movable spout to direct the flow into the
hole in the crust in the one mechanism, while others use a similar design to Pechiney.
FIGURE 1 - 12
PECHINEY AP18 FEEDER
BOX FOR AIR

CONNECTIONS
(SEPARATE AIR
VALVES ON POT
CRUSTBREAKER FOR EACH CRUST-
CYLINDER BREAK & FEED
CYLINDER)
SPRING (TO ENSURE

DOSING SYSTEM MATES
WHEN DROPPED INTO
PLACE BY CRANE)
FEED CYLINDER
AIR BLEED FROM

DOSING CYLINDER
FIXED CHUTE IN
FIXED CHUTE IN SUPERSTRUCTURE
SUPERSTRUCTURE
PLUNGER SHAFT GUIDE FOR SHAFT

OF DOSING UNIT
AIR BLEED FROM

CRUSTBREAKER
CYLINDER
BUSH HOUSING
"DAG" SCRAPER
FLAT BOTTOMED DELIVERY CHUTE

PLUNGER
Adapted from Gerphagnon and Wolter Patent: U S A 4,437,964 (22)

FIGURE 1 -13
PECHINEY AP18 POT SHOWING
INDEPENDENT FEEDER
DOSING
NOZZLE
DAG
CLEANER
Adapted from thefrontcover of Light Metal Age, December 1992.
ALUMINA
OVERFLOW
GAP
Adapted from the front cover of Kongerud (74).

1.2.6 Comparison of Integrated and Independent Feeder Designs
The integrated feeder preheats alumina because the alumina dose rests on the hot crust
before it slips into the hole, but the independent feeder allows direct entry of alumina
into the hole. The latter m a y provide better mixing as the alumina has a vertical and
horizontal velocity as it hits the bath. The alumina dose enters as a slug rather than a
gradual flow; which reduces dust loss into the pot cavity (27). However, there is no pre-
heating of the alumina prior to entry into the hot bath which m a y cause the dose to drop
to the bottom of the pot as a lump until it is broken up by the turbulence of the metal
and/or it reaches dissolution temperature. This mechanism can result in m u c k forming
if this does not happen effectively to all the dose.
Examination of Figures 1-10 and 1-13 illustrate the smaller cavity for independent
feeders. Note that Figure 1-10 has a very large crater which is not typical of all
integrated feeders, but it does show the general trend of crater shape. Also of note is
the alumina overflow gap on Figure 1-13(b) which provides a path for alumina to still
get into the pot if the dose chute is blocked. Integrated feeders do not have a small
nozzle so do not have this type of problem and tend more to spread the alumina over the
anode cover. There is more chance that the independent feeder design will ensure that
alumina gets to the bath compared to the integrated feeder design.
Welch et al., (28,29) and Maeda et al. (30) deduced the effects on slug versus slow f
dissolution in laboratory cells. Roach (31) summarised the Welch/Kuschel results. All
concluded that slug dosing gives quicker dissolution than slow feeding. N o comparison
work has been published on the effect of alternative feed designs on operating pots to
verify this theory or to quantify any effects relating to anode effects, sludge formation,
or current efficiency.
The crustbreaker of independent feeders needs to be cycled only when a hole is

required. Hence, breaking the hole every 2-3 feeds is possible. This reduces plunger
wear and reduces air usage. This is not possible for integrated feeders where the
crustbreak cylinder operates the dosing unit. For integrated feeders, changes in plunger
speed affect the dosing unit (refer Chapter 2).
A n attraction of the independent feeder is the ability to reduce the size of the air
cylinder for dosing, but still use a large cylinder for crustbreaking. Thus, the cylinder
for feeding is a 50-80mm bore with 50-80mm stroke. Compared to the 125-200mm
bore x 400-600mm stroke crustbreaker cylinder; this reduces the air required by a factor
of 10-200. Air consumption on a potline varies, but usage of 10,000-20,000 m V l is
not unusual, so the efficient operation of pneumatics is critical to an efficient feeding
system and m i n i m u m use of power.
The choice of feeder design may be selected by economics. The independent feeder
costs about three times that of the integrated feeder, but costs less to maintain. It is up
to the company concerned to choose between capital versus maintenance costs. N o
published comparison tests have been conducted to determine which design is more
accurate in dosing alumina or if feeder design is significant enough to affect pot
performance in a measurable manner. Capital costs for a 3 O O k A pot is about $30,000
for independent feeders versus about $10,000 for integrated feeders, but feeder life is
about 5-6 years versus 1-4 years for integrated feeders. Thus it is a comparison of high
initial cost versus long term savings in maintenance.
The cost of maintenance is minor compared to the fact that feeder failure results in
inefficiency, and operator exposure to falls, heat, dust and fumes. Hence, any feeder
design that performs its duty with a long life is highly desirable. It would be
advantageous never to change a feeder on an operating pot. Because pot life is about 5
years, people normally seek a 5 year feeder as a target. However, with the longer life
being achieved by Pechiney and some other pot designs, there is a directional change to
a target of a 7 year feeder. S o m e pneumatics suppliers are already testing cylinders of
this standard.
As most plants achieving world class current efficiency and very low anode effect rat
also have independent feeders, it could be assumed that the independent feeder is the
best design. This conclusion m a y not be substantiated as there could be m a n y other
reasons that m a y be more significant than feeder design e.g. bath/metal height,
magnetics, alumina segregation control, shot size, operating practices, anode cover
quality/variability, alumina quality, maintenance strategy, quality control.
This research has developed a better understanding of the advantages and disadvantag
of the two designs and introduced trials to quantify their importance on pot performance
(Chapter 10). Chapter 11 concludes that the independent feeder is the preferred design
and the basis for that decision.
1.2.7 Conversion from Integrated to Independent Feeding
Considerable work on feeder design has been carried out by the author as part of thi
research. T w o designs have the attractions of the independent feeders, are cheaper to
make than independent feeders and can be retrofitted in Alcoa, Alcan and Alumax
plants that already have integrated feeders (2,4).
The A2 (Appendix 6) has only one cylinder, but the feeder doses when the plunger is
out of the hole, so the dose directly feeds into the hole (Figure 1-14). Improved dosing
unit profile ensures more accurate dosing for variable alumina particle size. T o date
patents have been approved in the U S A , Australia and N e w Zealand. O n e disadvantage
of this feeder found in plant trials at Portland is that the feeder can jam if there is a large
build-up on the plunger. This results in prevention of the dose being released and
anode effects can result. There has also been a history of build-up of material on the
plungers if liquid level in the pot is not kept under control.
The A3 (Appendix 8) has a small cylinder (with a hole in the centre) mounted under t
crustbreaking cylinder (Figure 1-15). This hollow cylinder independently doses the
pot. The retrofit costs are about half that of alternative independent feeders, yet the air
savings and more accurate dosing of the latter design are achieved. T o date patents
have been approved in the U S A , South Africa, Australia and N e w Zealand with
approval expected in several other countries due to a very good International
Examination Report by the International Co-operation Treaty examiners.
FIGURE 1 -14
A2 FEEDER OPERATION
ALUMINA FLOW FROM

SUPERSTRUCTURE
ALUMINA DOSE
DELIVERY
OF DOSE
FEEDING
NOT FEEDING
FIGURE 1 -15
A3 FEEDER OPERATION
ALUMINA FLOW FROM

SUPERSTRUCTURE
ALUMINA DOSE
DELIVERY
OF DOSE
FEEDING
FIGURE 1 -16
SEQUENTIAL FEED AEDD

Clearly the A 3 is considerably more complicated and consequently more expensive than
the A E D D . A compromise was sought that had the accuracy and the flexibility of the
A 3 yet be cheaper to install. The A E D D with "Sequential Feed" (Figure 1-16) had the
basic feeder shape of the A E D D but had the directional control valve ( D C V ) mounted
on the top flange to give better control. This does not have the ability to break a hole
every few feeds or the ability to deliver the dose directly into the hole, but does have
the air usage savings and similar shot size accuracy to the A 3 and is easy to retrofit.
1.2.8 Continuous Feeders
Another concept that has been in the development stage for a number of years is to
smaller and smaller doses of alumina. Continuous (rather than batch) feeding in a
continuous operating process appears attractive. T o achieve this efficiency, one must
develop a crustbreaking system that feeds often, yet does not use a lot of air. The
dosing system has be unaffected by the upwards velocity of gas from the hole, and the
dose must mix with the liquid bath.
Alcan, Comalco, and Pechiney have experimented with the design. Most have not
proceeded with the concept, but Comalco is still working in this area and has a
prototype pot in operation currently in Bell Bay, Tasmania (based on research
documented in 28,29,31 by Welch et al.).
In 1990-91, the author designed a "dribble feed" chute that replaced the existing c
under the A E D D feeder to feed the pot continuously. It was possible to operate in a
continuous manner for many occasions, but the design gave inconsistent results. The
problems experienced were similar to those of other continuous feeders. The main
problems associated with continuous feeders are:
(i) blockages at the small orifices needed for slow flow;

(ii) variable flow rates if the alumina size changes;
(iii) back draught from the hole in the crust blows the alumina away; and
(iv) low addition rates tend to allow the alumina to float on the bath rather than
mix efficiently below the surface.
Though theoretically a good idea, the Portland design (and all others to date) have not
proceeded past prototype testing due to these production problems.
1.2.9 Aeration Feeders
Another feeder design that was a hybrid between a continuous feeder and the
independent feeder was the design used initially by Alcoa in their Tennessee smelter
(32). Here an airslide was used to supply the pot directly via an air-fed manifold. This
required the measured dose to flow evenly to each feeder in the pot. This design has
not proved effective due to uneven flow to the individual feeders. Another aerated
dosing system has n o w been developed for this plant via in-house development. This
has proved both reliable and accurate. It is a relatively low cost independent feeder,
using conventional crustbreakers to m a k e the hole, however it uses significant amounts
of compressed air.
Alusuisse developed several designs incorporating a fluidised alumina dosing unit off
from a crustbreaker unit (33,34,35). Aeration feeder designs were developed by
Aluminium de Greece (36), Nippon Light Metals (37), Pechiney (35), Norsk Hydro
(39), Alusuisse (40) and Alcan (41). The design that appears to be the simplest and
most practical of this type of feeder is the Alcan design shown in Figure 1-17(a).
Kaiser (42) used a combination of airslides and a rotary feeder to dose the alumina into
the pot. These aeration feeders have not progressed past the development stage to date
except for Aluminium de Greece w h o have retrofitted their pots with the patented
design with claimed improvements of 2 % current efficiency and one anode effect per
day (16).
Aeration feeders suffer from the variability of alumina flow that occurs with changin
particle size. Fine alumina flows easily w h e n aerated but, w h e n it stops, it is extremely
hard to reaerate and have it flow consistently. This can be very significant if there is
segregation in the supply system (which is often the case). The Moeller and Pust (43)
design shows just h o w complicated it can be to meter alumina accurately from a
fluidised bed (Figure 1-17(b)). Arnold (44) identified this as a concern with the design
C H A P T E R 1 INTRODUCTION A N D LITERATURE S U R V E Y P A G E 35
of these types of feeders as part of this research project. Other concerns with
continuous feeders are the build-up of scale over time, the cost of air and the dif
of designing a reliable cost-effective air supply system to the hot parts of the po
FIGURE 1 - 17
AERATION FEEDERS
ALUMINA
EZZZ
AIRjh
Adapted from Patent US4,919,303 (Ref 41)
(a) Alcan
LEVEL
CONTROL
AIR
Adapted from Patent UK2.101,980 (Ref 43)
(b) Moeller
The author developed a feeding system based on a conventional dosing system but
utilising an aeration pad to pulse the feed into the pot; the "pulse chute" (Kissane (4)
and Figure 1-18). In this way, the accurate dose is achieved (due to the conventional
dosing system), yet it can be metered into the pot over a longer period of time to
minimise the amplitude of the concentration swings. Hence, it is a compromise
between volumetric feeding and aeration feeding.
After building the prototype, it was realised that the A3 was potentially a better f
design. Resources concentrated on the A 3 and the patent has not proceeded. The main
disadvantages of the pulse chute is the risk of alumina permeating through the
perforated plate into the air chamber and the need to use more air to feed the pot when
most smelters are trying to reduce air consumption. Appendix 7 shows the international
patent application for this design.
1.2.10 Computer Control
Even with the best pot feeder available, support systems are needed to ensure the pot
gets the alumina w h e n it is required i.e. computer controlled feed logic and alumina
supply systems. This has become more critical with the desire to maximise the anode
area in the pot to get the greatest current possible through a pot without increasing its
size. A s the size of the anode increases, the bath volume reduces and the ability to
achieve a constant alumina concentration becomes more difficult unless good logic is
used to feed the dose at therighttime.
Pechiney patented an over-feed/under-feed logic in the early 80's (23). This type of
logic is n o w widely used to achieve the correct feed at the m i n i m u m voltage possible
yet avoid the high voltage of the anode effect (which causes an upset to pot operation).
Hydro Aluminium (45,(46)) and Reynolds (47) also have patents using similar logic. It
is curious that the patents were awarded on almost identical concepts.
FIGURE 1-18
PULSE CHUTE OPERATION
AIR IN
NOT FEEDING
'EEDING
1.2.11 Pot Alumina Supply Systems
There are basically four ways of transporting alumina to the pots.
(a) Overhead Cranes

The supply of alumina to the pot has traditionally been by overhead cranes which have
a hopper suspended or a hopper as an integral part of the crane. Figure l-4(a) shows
white alumina dust piled up on the top of the superstructure. This is a hygiene problem
to operators and tradespersons as the alumina has already been through the fume
scrubbers and the dust has fluoride absorbed on the surface of the particles. This may
be a contributor to asthma. Overhead crane systems have been used in many plants but
some have been converted to dense phase. Plants that still have overhead crane systems
include T4, T5, T8, T9, TIO, T12, T13, T14, T15, T16, T17, D 2 , D 6 , D 9 , D10, Dll,
D19,D20,D21,andD24.
(b) Airslides
Air gravity conveyors (or "airslides" as they are commonly known) have been used on
several Alcoa potlines as a low maintenance design that is simple to operate. Their
principle of operation is to have the alumina above a permeable fabric with air passing
through to aerate the alumina. Alumina flows if the angle of the fabric is above about
7 to the horizontal. These airslides transport the alumina right up to the pot in some
locations e.g. T7, T11,D12.
(c) Dense Phase

The development of a dense phase system has been difficult due to build up of hard
scale in the lines which reduces the conveying rate and requires the pipes to be cleared
at intervals. Alesa (the materials handling division of Alusuisse) has developed a dense
phase design that is used in many modern smelters as either an original installation (e.g.
Portland, D 3 , D 1 8 , D 2 ) or retrofitted later (e.g. T3, T6, D l , D 7 , D13, D22, D23).
Buhler Miag installed a system at T13 and Johannes Moeller a system at T 7 but both
had to be removed because of poor equipment life and build-up of scale respectively.
The Alesa systems are still affected by the scale problem of an extent that varies from
plant to plant. The Alesa design (48) is marketed as being cheaper than an airslide
system, easy to operate via a P L C or computer and low in maintenance cost (except for
the scale...which they do not advertise).
(d) Hyper Dense Phase

Late in the 1980's Pechiney developed a "hyper dense phase" for their A P 3 0 pot (49).
The system incorporates a choked airslide that is self feeding as alumina is displaced
and a local low pressure region is created. It is only marketed as part of the A P 3 0
potline package. It is similar to an airslide but works on the principle of aeration due to
the reduction of pressure at an offtake. This lower pressure causes flow of alumina
from head tanks. The result is a low air usage system with little or no alumina
degradation.
However, the low rates do not allow removal of scale that develops in any system usi
reacted alumina (from the scrubbing system). Hence, they have developed a cleaning
mechanism to protect the system from scale (49). This also addresses the problem of
foreign objects fouling the dosing system. It allows Pechiney to explore further the
continuous feeder idea, as foreign objects are a major problem with continuous feeders.
The slower the rate of feed, the smaller the aperture required for flow and the more
prone is the system to blockage or partial blockage which in turn stops or changes the
feed rate. O n e can see h o w the supply and the feeding systems interrelate.
The hyper dense phase is very similar to a USSR patent (14) that describes fluidised
supply of alumina to the sides of a pot. (Note that the U S S R patent was granted several
years before the Pechiney patent.) The hyper dense phase system has been installed at
D 4 , D 5 , D 8 , D14, D 1 6 and D25.
1.2.12 Plunger Developments
To control plunger wear, the c o m m o n methods are to reduce the time the plunger is in
the molten bath and/or to use stainless steel material. Generally, plants use timers in
their computer logic or on local control panels near the pot to control the time the
feeder is fully extended towards the bath. There is thus a fixed "dwell time" between
when the crustbreaker descends to w h e n it is retracted for most (if not all) pots. The
time that the plunger is immersed in the bath is therefore dependent on the response
time of the valves and pneumatics and on the speed of the piston in the crustbreak
cylinder.
To minimise the time in the bath, several designs have been developed. Alusuisse has
sensing unit that detects a voltage change across the feeder w h e n the plunger strikes the
bath (52) so the plunger is retracted quickly (Figure 1-19(a)). Pechiney uses a similar
concept on a small scale crustbreaker to detect liquid level for bath level control in their
A P 3 0 pot (19,53) and illustrated in Figure l-19(b).
The problem with the Alusuisse and Pechiney designs is that they require the plunger
get wet to work. It is not necessary to do this. A good operating practice is to control
the liquid level below the plunger w h e n the plunger is fully extended. In this way, the
plunger does not get wet and the plunger life is extended significantly. In Chapter 5
this philosophy is discussed in more detail.
A better device to minimise bath corrosion of plungers is end-sensing of the

crustbreaker cylinder. This design is marketed by some cylinder manufacturers e.g.
Atlas Copco and M e c h m a n . Pneumatic designs are used due to the effect of the strong
magneticfieldassociated around operating pots. Pot magnetic fields are not only strong
but also variable. This can seriously affect the performance of magnetic sensors.
However, pneumatic sensors are seldom used due to their high cost and the large
number of feeders per potline. A pneumatic design was investigated as part of this
study (Chapter 10).
In the absence of an end-stroke sensing, the next option is to have a plunger that la
for long periods. The only published work is that of Kaiser in 1984 (54). Strips of
different analysis cast materials were welded to a cast iron plunger in a test pot to gauge
relative wear rates. The conclusions m a d e were that 2 0 % chrome was required to give
reasonable wear rates and that the cost was economic on a cost per tip-day basis. (In
Chapter 5 this paper was examined and a m u c h better relationship was extracted from
the data.) The conclusions of the Kaiser paper have been adopted by the industry and
most plants have changed from cast iron to high chrome plungers. However, plants
continue to experiment with different plunger materials.
FIGURE 1 -19
PLUNGER END STROKE SENSING
Adapted from Patent AUB-24373/84 (Ref 52)
(a) Alusuisse
Adapted from Patent AU14784/88 (Ref 53)
(b) Pechiney
The mechanism of plunger wear has been under discussion since pots were invented.
S o m e believe the cause is erosion and others believe it is corrosion. This study
systematically identified (by theory and plant trials) the causes of plunger wear and
examined 19 materials and different shaped plungers (Chapter 5). The economics of
different plunger materials and designs are discussed in Chapter 9.
1.2.13 Dissolution of Alumina
A key factor in the development of a better pot feed system is to understand the
properties of the alumina...both its physical flow properties, and the ability to dissolve
efficiently. Alcoa, Alusuisse, Mitsubishi, Comalco and Hydro Aluminium have either
researched in-house or engaged expertise from universities or research establishments to
investigate these areas. Pechiney has not published their "in-house" research.
Most published literature on alumina dissolution has been generated by Welch

(University of Auckland) (28,29,55,56,57,60,61,62,63), Bagshaw (formerly Alcoa)
(28,29), Kuschel (Pasminco Research Centre, Newcastle) (56,57) and Taylor (Comalco)
(60,62). Several of these papers refer to similar results and conclusions. For four
years, Alcoa sponsored this research (summarised by Roach (31)) then Comalco
engaged Welch to continue research into dissolution and to develop a continuous feeder
that w a s operating in Tiwai Point ( N e w Zealand) before changing to Bell Bay
(Tasmania) in 1992.
Most alumina used in pots is that which has passed through scrubbers that treat the ex
gases from the pots...called "reacted" alumina. This is ingrained with cryolite, S O 2 and
fluoride. It is sticky and does not flow like fresh alumina. M a n y papers published on
the reactivity of alumina by Welsh et al. (28,29,55,56,57,60,61,62,63) have used fresh
alumina. This m a y give false impressions of what actually happens in operating cells.
This is a major fault with their published studies, and casts doubts on their results.
Jain et al. (55), Welch and Kuschel (56) and Barrillon (58) discussed dissolution stu
and the equipment used for the experiments. Taylor et al. (60) and Thomstad et al.
(61) explained the effect of sludge and showed h o w difficult it is to dissolve once
formed. Modelling has been used to simulate the effect of dissolution in cells and cell
geometry by Liu et al. (62) and Haverkamp et al. (63).
The dissolution work of the University of Auckland tends to support results of Maeda
al. (30) that fast addition of alumina gives fast dissolution. Also, the predominance of
fines does not enhance dissolution (although one would expect it should give a faster
dissolution). This is due to non-wetting of the dust. T h e resulting 'raft' of alumina
floats on the bath surface and does not break up into individual particles easily.
Agitation of the liquid is necessary to physically break up the rafts.
Roach (31) identified the order of priority of factors that affect alumina dissoluti
(most important first):
1. addition systems;
2. bath agitation;
3. superheat;
4. alumina concentration; and
5. alumina properties.
Alcoa's Western Australian Refineries considered it w a s up to the smelters to get

control of the feeding system before more work w a s needed on identifying alumina
chemical factors. This led Alcoa to cease engaging outside consultants into dissolution,
although Alcoa's very substantial Research and Development Department at Kwinana
(Western Australia) continues to develop alumina quality. Hence, from all the above
work, it can be concluded that it is best to have good feed control and that the pot
feeder clearly has a large impact on sludge (muck) control.
The type of feeder dose will have an effect on dissolution rate. Purdie (103) showed in
experiments with operating pots that about half of the alumina feed in pots fed by bar
breakers comes from the bottom of the pot. Other pots fed by independent point
feeders only "backfed" 10-20%. W h a t is not quoted in literature are comparisons of
different feeders in the same pot. This is the only w a y to measure the significance of
integrated versus independent feeder designs.

1.2.14 Alumina Dose Design and Accuracy
The Pechiney API8 independent feeder is claimed to achieve a feed shot size "accuracy
within 5 % of the designfigure"(20). For a mass of 1kg, this relates to +/-50g. This is
the only published data source that specifies the accuracies of pot feeders. Discussions
with Holmsi and Leonard (101) verified this to be typical of Pechiney pots. It is
reasonable to assume that this is the accuracy of other independent feeders as they are
modelled on the Pechiney feeder. T h e Boyne integrated feeder accuracy is +/-100g
(102) for a 1.8kg shot and the A E D D accuracy is about +/-40-250g. Chapter 2 shows
results far superior to these for the sequential feeder and the A 3 feeder viz. +/-14g and
+/-12g respectively. The accuracies of the point feeders compares very well with the
Soderberg and bar break designs where 30-70kg shots were used.
To address the effect of segregation and alumina size variation, some dosing systems
are designed with approximately 45 angled surfaces at top and bottom. This is seen in
designs of a U S S R patent (24), the Pechiney A P I 8 feeder (22,23) and Figure 1-12.
Arnold (44) and Richards et al. (64) conducted tests on flow and angle of repose of
reacted alumina with relevance to feeder dosing units. Richards showed that the angle
of alumina at the top of the dosing unit varied depending on its size and fluidisation;
fine material had a stationary angle close to horizontal and coarse material was at a 45
angle. Arnold (44) recommended that the exit angle should be as steep as possible (at
60 or steeper) to ensure all alumina exits the dosing unit. Kissane (76) combined these
two observations and developed a cheap ($40) spool insert with 60 angles at top and
bottom that could be retrofitted into the A E D D integrated feeder (Figure 1-9 and 1-11),
and incorporated this design feature in the A 2 (Figure 1-14) and A 3 feeders (Figure 1-
15). Thus, alumina size and degree offluidisationwould have a minimum effect on
dosing accuracy. Data to support this is presented in Chapter 2.
The flow properties of alumina have been more tightly defined recently by Pechiney
both from a chemical and physical viewpoint (65). This has become necessary as
feeder development has highlighted h o w important the particle size of alumina is to pot
feeding.
Reverdy, a Director of Pechiney, states that direct entry into the hole is a benefit of
independent feeders (66). It is interesting that for some Pechiney plants, the
crustbreaker only operates every 2-3 feeds e.g. D 1 0 . This prevents direct entry to the
bath and suggests that direct feed m a y not be a major factor in pot control. T o identify
the effect of direct feed on dissolution, this research developed chutes under the feeder
that allow direct feed into the hole. S o m e of these can be seen in Figure 1-14 and 1-15.
Section 2.8 discusses the effect of chute design on the pot.
Patents from the USSR have explored different concepts in feeder design from
integrated to independent feeders. Chevonin et al. (67) have developed a dosing hopper
with an internal wire brush to clean out alumina. This is a costly solution to segregation
w h e n one considers there are a couple of hundred feeders in a modern smelter. Mirkin
et al. (68) have a patent for a double valve dosing unit controlled by sieves sequentially
releasing the dose into the pot. This design appears to reduce internal by-pass in the
feeder, but m a y still have poor accuracy due to leakage around the valves. Costs are
expected to be high.
1.2.15 Pneumatic Air Cylinders
Pot crustbreaker and feeding units only use pneumatic cylinders as there is a potentia
risk that hydraulic oil m a y createfiresw h e n exposed to the elevated temperatures of a
pot; up to 1,100°C is possible. Under the varying conditions of pot temperature, the
desire to last over 1 million cycles (about 5 years) historically has proven to be a
difficult assignment. The causes of failure of cylinder and assembly components are
similar in most plants, but the extent differs due to component design, to downstream
failure from other parts, to pot conditions and to repair/installation errors.
Traditionally, smelters have sought a cylinder with the longest life, rejecting all t
not achieve the desired results. For some plants, this has resulted in m a n y years of trials
of different cylinder designs. This study has shown that m a n y factors affect cylinder
life, not the least of which are the pot itself and the feeder or crustbreaker design
(Chapter 7).
Figure 1-20 illustrates the main components of a typical pneumatic cylinder. Each
cylinder brand has different configurations of seals (that prevent air leakage past the
piston rod or barrel), bushes (which prevent the shaft from moving transverse to the
centre line), cushioning (to slow the piston at the end of its travel) and seal/bush
orientation. The cylinder shown is an Atlas Copco cylinder that has one of the best
performance history in pot feeder operation.
F I G U R E 1 - 20
PNEUMATIC CYLINDER PARTS
Adapted from Atlas Copco, Series 40/41M brochure, 91270041-92, 2nd edition, Pp,3
N o papers or patents could be sourced on comparison of different cylinders nor any

other system components associated with the cylinders e.g. exhaust mufflers, D C V s ,
flow control valves. The performance of different components can only be found by
discussion with suppliers to plants or with representatives of the plants themselves.
Because of the long life of feeders, transferring of engineers and maintenance
personnel, and the lack of failure data, there are little valid data from most plants.
In order to establish optimum designs and suppliers, this study contacted many smelte
(Table 1-VI). Plants with integrated feeders generally had poorer performance to
independent feeder plants; 1-2 year life versus 5-6 years. Independent feeder
crustbreaker units generally operate at slower speeds to integrated feeders, so this
reduces frictional wear on seals and bushes. Independent feeders that break a hole
every 2-3 feeders have an added advantage over most plants by having less cycles per
year. Examination of the plant data shows that there are extreme inconsistencies in the
performance of the one type of cylinder in two different plants (Chapter 7).
For example, of the 16 integrated feeder plants contacted, T17 had the longest
integrated feeder cylinder life at just under 5 years. This plant uses Scheffer 1 0 0 m m
O.D. cylinders, but the key to the good performance was retrofitted Dover Teflon seals.
Identical cylinders (without Dover seals) at T 7 and T15 only achieved 1-2 year life.
Use of 20 similar seals at Portland showed life under 1 year. The longest life cylinder
of the 8 independent feeder plants contacted was from over four hundred Atlas Copco
cylinders at Karmoy (93) which theoretically had a life of 94 years! (Table 1-VTI). Yet
a 1 5 0 m m O.D. cylinder on an integrated A E D D design at T 2 only had a 2 year life.
Given that most plants are not immediately in a position to change their large number
cylinders (up to 2,000), there are still opportunities to improve cylinder life. Plants T 7
and T16 addressed the source of feeder failures and achieved a 6 5 % and 9 0 % drop in
failures respectively, and an 8 0 - 9 0 % drop in costs in two years, yet still use basically
the same cylinder and plunger designs as they did originally. A s will be seen Chapter
10, Portland feeder and cylinder life improved by 9 0 % as a result of this investigation.
TABLE 1 - VI
SMELTERS CONTACTED IN THIS RESEARCH

Capacity Integrated/
Location Smelters/Research Centres Visited AEDD
(t/year) Independent
Australasia Boyne Smelters Limited (Comalco) 230 integrated

Bell Bay (Comalco barbreak) — barbreak *
T o m a g o (Pechiney design) 380 independent
Kurri Kurri (Alcan) 153 integrated * *
N Z A S (Comalco) 269 barbreak *
Pt Henry (Alcoa) 180 integrated * *
Portland Aluminium (Alcoa) 320 integrated * *
South America Alumar, Brazil (Alcoa) 350 integrated * *

Aluar, Argentina 175 independent
U.S.A. Wenatchee (Alcoa) 220 integrated * *

Badin (Alcoa) 115 integrated * *
Massena (Alcoa) 127 integrated * *
Rockdale (Alcoa) 310 integrated * *
Warrick (Alcoa) 270 integrated * *
Tennessee (Alcoa) 200 independent *
Alumax, Mt Holly (Alcoa design) 181 integrated * *
Kaiser Research Centre, California — —
Mead (Reynolds) 200 integrated
Vanalco (Alcoa design) 115 integrated * *
Mosjoen, Norway (Alcoa design) 120 integrated * *

Europe, U K
Granges, Sweden (Reynolds) 100 independent *
Karmoy, Norway (Hydro) 220 independent *
Hydro Research Centre, (Norway) — —
Lynemonth, England (Alcan) 130 integrated * *
Anglesea, England (Kaiser) 117 barbreak
ISAL, Iceland (Alusuisse) 92 independent
Ardal, Norway (Alusuisse) 188 independent
— *
Canada Alcan Head Office, Montreal
Le Terriere (Alcan) 204 integrated *
Baie C o m e a u (Reynolds) 400 independent *
TOTAL 5,366 2,995
(26% of western world capacity)
Note: Installed capacity from Pawlek (94);updated 1994.
Suppliers Contacted
Parker-Hannifin, Sydney cylinders
Terry Fluid Controls, Melbourne cylinders
S M C , Sydney cylinders
Atlas Copco, Sweden cylinders
C P O A C , France cylinders
Norton, U.S.A. seals
Dover, U.S.A. seals
TABLE 1 - VTJ
KARMOY CYLINDER REPAIR HISTORY
Hydro Aluminium,
Karmoy Fabrikker,
Elektrolyse Prebake,
Karmoy.
21.10.92
To Jim Kissane
According to your questions yesterday about the Atlas Copco Monsun Tison cylinders type; C
40-200-50-500 F 00 installed in 1987, w e can give the following data;
Number of Atlas Copco breaker cylinders failures
in 1989 2 numbers
in 1990 11 "
in 1991 12 "
in 1992 8 "
For your other questions I am not allowed to answer because of the technology agreement.
Kind regards
Magne Stueland,
Editorial note:
Karmoy have 444 feeders. The average cylinder life for a turnover of 8.2 per year is calculated at 9 4 years
versus 2-5 yearsformost plants. Note that the quoted data only relates to cylinders and not to feeder or plunger
life. Evenforcylinder life, these results are outstanding. The "other questions" (mentioned in the last sentence
of the letter above) were relating to plunger and feeder information.
1.2.16 Pneumatic Systems
The invention of point feeders reduced the shot size from 30-70kg to l-2kg. As pot
size increased, the number of feeders per pot increased from one per pot to a m a x i m u m
of five per pot (Figure 1-21). T o distribute this alumina across the pot, different air
delivery systems (commonly called "pneumatics" in this text) have been used.
In Alcoa/Alcan/Alumax plants, the air supply to multiple feeders/pot have been by one
directional control valve ( D C V ) . T h e independent feeder systems require the
crustbreaker and dosing units to be separately controlled, so two D C V s are required per
feeder. In order to reduce the air usage, it was desirable to mount the D C V s as close to
the feeder as possible, so the valves were generally positioned on the pot superstructure.
(It will be shown in Chapter 2, 4 and 5 that this also improves shot size accuracy,
pressure at the crust and plunger wear.)
Air usage for pot feeders is quite significant e.g. about 30% of Portland's usage is
pot feeders. For a smelter, the cost of air compression m a y not just be the cost of power
and maintenance of the compressor equipment. There can be a very expensive "cost
penalty" as generally there is a m a x i m u m power limit on the plant. It is highly
desirable to maximise the availability of power to the pots. T h e more pots that can
operate, the more profit can be m a d e from metal sales.
Compressors and other uses of "ancillary power" have a cost penalty which is several
times the normal calculated air compression cost. T h e degree of "penalty" will vary
depending on whether one can sell the metal or not. If the metal cannot be sold due to a
oversupply (as has been the case for most plants in 1994/1995), then the cost of air is
the prices quoted above. If metal can be sold, the "penalty" can be several times these
values.
Not only is the cost of air generation a factor, but also the available quantity of
the net pressure that is available. Often, overuse of air leads to low supply pressure
with subsequent effect on crustbreaker plunger pressure. Dunstan (69) proposed
changing feeders based on air consumption.

FIGURE 1 - 21
PORTLAND POT FEEDER SYSTEM
WIDE O R
TAP AISLE
RESTRICTOR
(12mm)
NOTES:
(1) In closed position, ball valve NARROW
has a 12 m m hole. AISLE
(2) Line A supplies air to force
plunger up.
(3) Line B supplies air to force
plunger down.
RESTRICTORS
(12mm)
IR SERVICE UNIT
-BALL VALVE (BLEEDS
TO ATMOSPHERE)
-FILTER
-LUBRICATOR
GATE VALVE
SUPPLY
AIR
It is considered that this is not an effective manner of cost reduction and that the
particular problems at the plant concerned should be approached in a different manner
rather than accept that excessive air leaks are inevitable. A s no other study of
maintenance and cost reduction has been published, Chapter 8 examines the causes of
crustbreaker/feeder failures and systematically identifies opportunities for plants to
increase equipment life. Chapter 9 examines the economics of different maintenance
strategies and feeder designs.
To minimise the high air costs of feeding systems (especially the crustbreaker) some
smelters utilise pressure regulators or commercially available latching systems to hold
the air cylinders in the up position. Swiss Aluminium has patented a dual pressure
setting system for the cylinders to keep low pressure on the feeder w h e n in the up
position (70). B y reducing this pressure, there are cost savings due to less air usage.
Ross Operating Valve C o m p a n y market a similar device that is claimed to reduce air
usage by 5 0 % by a similar method, but there are few units operating in plants at
present. Terry Fluid Controls and Comalco jointly developed and have applied for a
patent for a pressure regulation system that also reduces the pressure holding the
cylinder retracted to 100-200kPa. The claimed benefits are a reduction of up to 5 0 % in
air leakage rate and faster downstroke time (98). The units will be on trial at several
smelters (including Portland) by end of 1995.
All four of these designs accept that air leaks are inevitable. However, these design
m a y be uneconomic if seals do not leak or do so to a minor extent, as each feeder needs
a unit worth several hundred dollars. It appears to be more efficient to address the
primary cause of leakage rather than work on the effect.
1.2.17 Other Feeder Developments
Alcoa has worked on a gravimetric feeder that avoids the bulk density change resultin
from a change of alumina size, but this has not led to plant scale installation due to high
costs (71). A s will be seen in Chapter 2, this research discovered that the variation of
bulk density had only a small effect on shot size accuracy provided size changes are not
extreme. Also, this research identified that accuracy from shot-to-shot for volumetric
feeders on operating pots can be as low as 0.5% of shot size, which is sufficiently
accurate not to consider costly alternatives and is far better than the 5 % accuracy quoted
by Pechiney (20).
The gases given off by anode effects are not desirable to people who work in the
potroom or operating areas (1). Greater awareness of the health and hygiene of
operators and tradespersons in potrooms has driven managers to address exposure of
people to heat, dust and fumes which are faced during change-over of feeders,
unblocking feeder holes, and anode effects (72,73). Asthma is of concern in most
smelters as this m a y be aggravated by feeder monitoring, changeout or servicing where
people are potentially exposed to dust or fumes (74).
In recent years, the reduction of anode effects has become a high priority for alumin
companies across the world as anode effects m a y lead to global warming. However,
recent research has questioned the relative magnitude of the effect of aluminium
smelting on global warming. Studies between the environmental protection agencies
and aluminium producers world-wide are assessing the impact of aluminium smelting
on global warming (95). Clearly this is an area that is the subject of m u c h research.
Irrespective of the relative significance of the effect of anode effects on global

warming, there are m a n y reasons w h y it is useful to optimise pot alumina feeders...cost
control, pot stability, health and safety of personnel. T h e financial benefits alone are
substantial.
1.3 DEVELOPMENT OF THIS RESEARCH
1.3.1 Research Objectives
The research objectives of this project were to:
(i) optimise existing feeder designs (especially the AEDD feeder),

(a) reduce failures
(b) reduce anode effects
(c) reduce costs
(ii) compare integrated and independent feeder/crustbreaker units (called

"feeders" in this text) and develop a state-of-the-art feeder which has the
following features:
(a) life of over 5 years,
(b) cost effective in both capital and maintenance costs, and

(c) can be retrofitted in integrated pots easily.
1.3.2 Optimizing Existing Feeder Designs
In 1990, the author commenced optimising the AEDD feeder at Portland, but with a
view that this research would have a wider application across Alcoa and other plants in
the industry. Contact with other Alcoa smelters showed there w a s a wide range of
crustbreaker/feeder life being achieved across all 9 plants, despite having almost
identical designs (Table 1-IV). T h e distribution of causes of failure varied significantly
from plant to plant (Table 1-V). Thus, plant to plant differences seem to be significant
even with the same basic design.
Early in the investigation, it became apparent that the conditions at Portland were muc
worse than m a n y (if not all) Alcoa and non-Alcoa plants e.g. hotter, multiple feeders
per air valve, high speeds, lubrication breakdown, inferior cylinders, eccentric
mountings, design faults, deep penetration in the bath. It was considered by several
pneumatic cylinder manufacturers (Atlas Copco, Terry Fluid Controls, Parker Hannifin)
that the conditions at Portland were the worst seen in their experience...especially the
temperature exposure. Thus, optimising Portland feeders was indeed a challenge.
An extensive literature search on feeder technology (mentioned previously in this

chapter) showed that feeder developments are often kept secret within m a n y
organisations. Often published literature contains information that is either already
widely k n o w n or is published as a marketing strategy e.g. Pechiney. For this reason,
except for dissolution studies conducted by a handful of university researchers, the
majority of published work was in patents. It was apparent that to get improvements in
feeder technology, one had to do the development oneself, using personal contact with
other plants wherever possible to flesh out potential problems.
It has been the custom in the aluminium industry that different companies do not assis
each other, very m u c h due to the competitive nature of the industry. From 1990 to
1994, the author attempted to break this d o w n by contacting by phone and faxcimile
m a n y non-Alcoa plants w h o also were frustrated that similar feeder problems were
experienced. Representatives from these plants saw that there was an opportunity for
all parties to benefit jointly from cross pollination of ideas. Often confidentiality
concerns restricted data or information being shared, but comments on cylinder design
and feeder performance were usually freely given.
The author visited 18 smelters to inspect and discuss feeders (9 integrated, 7

independent, 2 barbreak) The plants and suppliers visited are listed in Table 1-VIData
and input from u p to 2 7 smelters are presented in this report; 17 integrated, 24
independent, 1 Soderberg, and 2 guillotine. The plant n a m e is generally coded to
protect confidentiality unless the data is on public record or presented as a formal
correspondence. A "T" code is used for plants w h o use integrated feeders, and "D" is
used for those w h o use independent feeders (as mentioned in the Note to the Reader on
page IV).
The author approached optimising existing feeder designs by extensive use of statistics
and economic analyses. B y use of chemical engineering as well as mechanical
engineering principles, pot operation and feeder maintenance aspects were investigated.
Close co-operation with operators and tradespersons w a s used as a means of
implementing designs that were applicable to all plants, but not designs that only had
attractions on paper...they also worked in the field. Designs that assisted the safety of
operator and tradespersons were developed in association with improvements to costs
and feeder life.
There are common items in integrated and independent feeders that this research
identified....cylinder design, plungers, dosing units, air supply hardware. The
investigations undertaken resulted in an understanding of "crustbreaker/feeder
technology" in general, rather than just the technology peculiar to Alcoa A E D D feeders.
In this document, suggestions are m a d e on optimising the design of all existing feeders
as each design and plant location has the potential to improve performance.
1.3.3 New Feeder Designs
Contact with plants using independent feeders identified usually much better feeder li
and lower maintenance costs than plants using integrated feeders. However, the capital
cost is several times higher and the design cannot be retrofitted in pots currently using
integrated feeders.
Hence, this research developed designs that could be retrofitted into plants that
currently use integrated feeders. The designs were to be cost efficient alternatives that
had the attractions of the independent feeders, but not the same high capital cost. A s
noted previously, the A 2 design has been successfully patented in Australia, N e w
Zealand and the U S A (Reference 2 and Appendix 6) and the A 3 in Australia, N e w
Zealand, South Africa and the U S A (Reference 3 and Appendix 8). The A 3 has the
potential to achieve independent feeder performance (or better) and still be able to be
retrofitted into existing A E D D feeder pots with m i n i m u m installation costs. Figure 1-
22 illustrates the feeder designs invented as part of this research along with the A E D D
feeder.
FIGURE 1 - 22
PORTLAND FEEDER DESIGNS
AEDD A2 PULSE CHUTE SEQUENTIAL A3

FEED
1.3.4 Across Alcoa Optimisation
It is useful to explain the process on how Alcoa plants responded to, and assisted
research.
Each of the nine Alcoa smelters world-wide has a team of people investigating
improvements in key activities that are considered the major items to make Alcoa world
class in all aspects of aluminium production. These teams are called "In Control and
Capable" (ICC) Teams and the leaders from each make up the ICC Network.
The Feed Alumina Network addresses the control of anode effects and pot control.
Massena ( U S A ) is the lead plant for this Network and was responsible for
benchmarking the best in the world. In 1991, all I C C teams submitted their key
findings to senior international Alcoa management. O n e of the Massena
recommendations was that Alcoa should "support feeder technology development in
Portland". Later, Alcoa listed "Portland Feeder Technology Development" as one of
the 39 enablers for Alcoa to become world class; only two items were listed for
Australia.
In April, 1992 at Wenatchee (USA), the Alcoa Feed Alumina Network hosted a
presentation by the author on feeder technology to discuss opportunities from which
each plant could benefit. A list of ten general items was raised to be considered by all
plants; commonly called the "Kissane Feeder Recommendations" (Table l-VlII).
Appendix 4 (reference 99) is the instruction from Bob Seymour (Alcoa Quality
Manager, Primary Metals) to the Alcoa Smelting Managers to implement the feeder
recommendations in all plants and w h y they are important. These recommendations
became one of the seven Critical Activities for Alcoa smelters for the next two years
and implementation became a high priority item in all nine international Alcoa plants.
TABLE 1 - VIU
ALCOA FEEDER RECOMMENDATIONS

1 TRACKING SYSTEM
2 SUPPLIER QUALITY A S S U R A N C E
3 S U R V E Y D O W N S T R O K E TIME
4 S U R V E Y BLOCKED FEEDER HOLES
5 R E D U C E D W E L L TIME
6 DUAL D W E L L TIME
7 P E R F O R M A N C E CRITERIA
8 S P O O L INSERTS
9 PLUNGER COMPOSITION/SHAPE
10 PISTON LEAKAGE S U R V E Y
(PLUS INDIVIDUAL PLANT RECOMMENDATION LISTS)
After two years, the recommendations were removed from the critical activities
to the success of feeder research (as outlined in Appendix 5 (reference 100) by Dick
Taylor (Director of Technology, Alcoa Primary Metals)). Teams are currently
operating at most locations to optimise pot feeders and they are continuing to achieve
very worthwhile results.
1.3.5 Document Layout
To become knowledgeable on the state-of-the-art, a literature survey was conduc

well as contact with other smelters and equipment suppliers (Chapter 1). This research
consideredfirstof all the desire to measure an accurate dose (Chapter 2) and deliver it
into the bath (Chapters 3). A critical factor in the operation of feeders is the air supply
system. Chapter 4 describes tests conducted on the pressure profiles of cylinder inlet
and outlet ports to establish what affects the speed and pressure of crustbreaking units.
As plunger failures are one of the predominant causes of crustbreaker unit remo
plunger wear and build-up on the plunger were discussed in depth (Chapter 5 and 6).
N o matter h o w accurate the feeder is, a requirement of maintenance personnel is to

contain costs and operating personnel do not want to be troubled by frequent
changeouts. O n e of the most frequent causes of failure is the pneumatic cylinder, so
Chapter 7 discusses extensive trials on cylinder operating variables and cylinder
lubrication.
The causes of failure of feeders is discussed in Chapter 8 and recommended solutions

are raised to extend feeder life. Chapter 9 establishes that long life can be
complemented by cost efficient designs and maintenance strategies to minimise feeder
costs. Note that the two m a y not go hand in hand on a short term basis, but, on a long
term basis, long life generally results in reduced costs.
An integral part of this research was not just to make existing feeders better. Several
designs were developed and the good results indicate that these designs of low cost
feeders are capable of achieving the targets of long life and enhance the operation of the
pots.
Finally, Chapter 11 collates the results of the feeder research both in Australia and
overseas, and indicates where the technology can improve in the future.
CHAPTER 2 SHOT SIZE ANALYSIS PAGE 61
CHAPTER 2
SHOT SIZE ANALYSIS

2.1 SUMMARY
About 200 tests involving over 30,000 measurements were conducted on the shot size
of several pot feeder designs viz. the existing Portland A E D D , the A 2 and A 3 , plus two
feeders from plants T 6 and Tl 1. These tests sought to establish the factors that affected
shot size mass and variability. Parameters considered were dwell time, stroke time,
alumina size, feeder-to-feeder differences for the same design, cylinder brand, spring
type and cylinder cushioning quality.
Portland AEDD feeders were significantly affected by many parameters with the most
important being feeder to feeder changeout differences, alumina size, type of cylinder,
and dwell time. The overall result was a potential variation in shot size of up to 3 7 %
and a standard deviation of up to 1 4 % shot-to-shot. There was an improvement of 5 6 %
in shot-to-shot standard deviation and 3 6 % in feeder-to-feeder standard deviation when
a 60 spool insert was installed for a changeover cost of $40.
The standard T6 and Til feeders were better than the Portland AEDD but were affected
by the same parameters. The A 2 was more accurate with a lower standard deviation
than all the integrated feeders, and was less affected by the parameters considered. The
A 3 independent feeder had a standard deviation of about 0.6% of shot size for shot-to-
shot accuracy, making it by far the most accurate feeder tested and more accurate than
Pechiney independent feeders.
The improved shot size accuracy due to installation of spool inserts on Portland AEDD
feeders w a s a major enabler to a significant reduction in anode effects. It w a s
established that up to 32 parameters can affect shot size accuracy for integrated feeders
and 20 parameters can affect independent feeders.
CHAPTER 2 SHOTS SIZE ANALYSIS PAGE62
2.2 I N T R O D U C T I O N
A study of shot size mass from pot feeders was initiated in December 1990 to determine
what factors affected shot size mass and accuracy. Several feeders were tested:
- existing Portland AEDD

- Portland A 2
- Portland A 3
- T 6 integrated feeder
-Til integrated feeder
Also, individual and multi-feeder operation per DCV were tested to see the effect of
pneumatics on shot size accuracy.
The A3 is an independent feeder that was expected to give similar results in shot si
accuracy to other independent feeders, as the effective piston area and stroke distance
are similar to those of most independent feeders. This should give a good comparison
of integrated and independent feeder dosing accuracy.
An insert was invented by the author to install inside the dosing spool on the Portl
feeder following examination of results of alumina flow properties by Richards et al.
(64) and Arnold (44). This design could be used in virtually all Alcoa, Alumax and
Alcan integrated feeders and was tested in Portland, T 6 and Til feeders.
The Portland feeders were tested using 125mm O.D. x 508mm stroke Parker, Terry
Fluids, and Atlas Copco pneumatic cylinders. The T 6 and Til feeders used 1 0 0 m m
O.D. x 3 5 6 m m stroke cylinders of different manufacturers...Terry and Lindberg
respectively.
Different stroke times were chosen for each feeder to be consistent with field testi
operating pots.
CHAPTER 2 SHOTS SIZE ANALYSIS p A G E 53
A n off-site testrigwas constructed based on the Portland superstructure dimensions

(Figure 2-1). Tests were conducted using normal and fine sizes for both pure and
reacted alumina. Alumina used in most smelters is not fresh alumina direct from
refineries. The alumina is generally passed through reactors that scrub pot gases before
being transported to the pots. This affects its flow properties and dissolution rates.
FIGURE 2 -1
OFF-SITE SHOT SIZE TEST RIG
Most papers published on the dissolution of alumina in cells have used unreacted
alumina (refer Section 1.2.13). This m a y give a false impression of what actually
occurs in pots as unreacted alumina has better flowability than reacted alumina.
CHAPTER 2 SHOTS SIZE ANALYSIS PAGE 64
Reacted alumina has a sticky nature due to absorption of exhaust gases from the pots.
It is indeed possible that dosing unit performance could be quite different for unreacted
and reacted alumina. Tests of feeder shot size accuracy should use reacted alumina as
this is what is used in the pots. Hence, reacted alumina was used in this study except
for comparison tests at the start of the investigation. Alumina was collected from the
superstructure of off-line pots (with the exception of these early runs).
Each shot of alumina was caught in a plastic bag around the outlet chute with minimal
dust lost. Shot size masses were graphed on control charts as they were measured
(Appendix 1) and summary data (mean mass and standard deviation as a percentage of
shot size) were calculated (Appendix 2).
Appendix 1 illustrates raw data from one typical run. This illustrates that the shot
was in statistical control as the range and m e a n values were within the control limits.
The bottom curve is the height of alumina in the hopper around the feeder. This shows
no correlation between the shot size and height. Thus, there w a s always enough
alumina to supply the feeder for these tests.
During each run of 25 shots, samples of alumina were taken by catching several shots.
These were placed in sealed containers and analyzed by E M L Laboratories for sieve
sizing, Malvern micro sizing and "funnel flow time". S o m e of the early runs were also
analyzed for loose and packed bulk density (Appendix 3).
Alcoa has established a linear relationship between microfines content (-20 micron)
the time taken for afixedquantity of alumina to flow through a standard shaped funnel.
The finer the alumina, the longer the "funnel flow time" (FFT). This is a simple
repeatable test to compare fines content whereas expensive microsizing equipment m a y
be affected by machine-to-machine variation. W h e n one considers that a point feeder is

largely affected by flowability of alumina, the F F T is a very useful parameter to
compare alumina types.

CHAPTER 2 S H O T S SIZE A N A L Y S I S P A G E 65
The early tests identified several factors affecting shot size that had not previously been
understood as being important viz. cylinder speed, spring tension, cylinder design,
cushioning quality and dwell time. These factors were examined in later tests. Tests on
operating and off-line pots were conducted to verify the off-site tests and to examine
the effect of "multi-feeder per D C V " and "one feeder per D C V " operation on shot size
accuracy.
Some 30,000 shot sizes were measured in about 200 runs. A full detailed analysis of
the results of this investigation is contained in a report by the author (76).
2.3 PROCEDURE
2.3.1 Dwell Time
"Dwell time" is the time from w h e n the computer energises the D C V to w h e n it
deenergises the D C V . It includes the stroke time of the cylinder plus the time the
piston/piston rod is stationary in the extended position. For integrated feeders, the
dwell time is selected to achieve two objectives:
(i) all the feed shot is discharged from the dosing unit; and
(ii) sufficient pressure is achieved for the plunger to break the crust.
For independent feeders, a different dwell time is used for the feed and crustbreak
cylinders, however the objectives are the same as for integrated feeders but split into
function (a) for the feed cylinder and (b) for the crustbreak cylinder.
Dwell times from 0.5s to 5.0s were used in a random manner for each test to avoid any
systematic errors. Repeatability w a s established by measuring the mass of 25
consecutive shots for the same dwell time conditions. There was no correlation between
shot size and order of testing in any testing done off-site or on-site. Figure 2-2
illustrates the relationship between shot size mass, shot size standard deviation and
dwell time for the existing Portland A E D D feeder. This type of figure has been used
throughout this chapter to illustrate the performance of different designs for varying
dwell time.
FIGURE 2 - 2
PORTLAND AEDD SHOT SIZE GRAPHS

P A R K E R CYLINDER.lOs down,1.1s upstroke
)
S H O T SIZE (g)
2000
#1 AEDD FINE #1 AEDD NORMAL #1 AEDD UNREACTED

W R U V «»(•»*•» •• a a •
#2 A E D D FiNE #2 A E D D N O R M A L
200 -L
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
DWELL TIME (s)
STD DEV (%)
12
10 -
8
6
4 ""Htra-iio*
2
R U N S 1,7,13,19,22
Note: 1.7/1.9 refers to 1.7s downtroke and 1.9s upstroke time.

CHAPTER 2 S H O T S SIZE ANALYSIS P A G E 67
2.3.2 Dosing Unit Shape
It was found from graphs such as Figure 2-2 that the feed shot varied with dwell time.
This suggested that the shot size increases the longer the dosing unit is stationary. It
appears that the alumina does not flow out smoothly (by mass-flow) from the dosing
unit. This is a phenomenon that could affect both integrated and independent feeders,
but is less critical for independent feeders which m a y use long feeder unit dwell times
as the dwell time does not affect the time the plunger is potentially in the bath.
This variation in shot size was alarming in its magnitude but not particularly
unexpected based on test work conducted on flowability of reacted ore. Arnold (44)
suggested that at least 60-70° angles (not 45°) should be used on the dosing spool for
mass-flow. The original outlet angle of all Alcoa, Alcan and Alumax feeders and most
Pechiney and independent feeders is 45°. The Pechiney and other independent feeders
have a 45° angle at the top as well (as illustrated in Figure 1-11).
Richards (64) also suggested that there was variability due to the angle of repose of
various alumina sizes. Coarse alumina has a 55 angle to the horizontal and fine
alumina has almost a 0° angle to horizontal (especially if aerated). These findings were
also confirmed by Arnold (44).
Thus, both inlet and outlet angles are important to shot size variability. The angles
independent feeders should be increased from 45°, and Alcoa/Alcan/Alumax feeders
should have an angle at the top as well as at the bottom for better control over shot size
variability.
On the basis of this information, an insert with 60° angles at top and bottom of the
dosing unit was designed to improve variability of shot size. This insert can be installed
inside the existing spool of the Alcoa/Alcan/Alumax integrated feeders, as most have
identical dimensions. Thefinalinsert designed at Portland is seen in Figure 2-3.

FIGURE 2 - 3
SPOOL INSERT ASSEMBLY
(a) Disassembled
BASE
(b) Assembled (c) Details
Note that it is important to install the insert the right w a y up. If it is installed upside
d o w n it restricts the outlet port of the dosing unit and can cause very long discharge
times and a further 2 0 % drop in shot size (Figure 1-9 and 2-3).
2.3.3 Stroke T i m e
The stroke time is the time that the piston or plunger takes to travel from fully retracted
to fully extended (or vice versa). For simplicity, the downstroke and upstroke times are
reported together; such as 1.0/1.2s for downstroke/upstroke times.
In 1991, investigations commenced in reducing the stroke speed to reduce fatigue

failure of piston rods and spools. The original speed at Portland was 1.0/1.2s. Stroke
times at Portland were increased by November 1991 in order to stop fatigue failure of
components. The n e w stroke time at Portland was 1.6s (S.D. 0.3s) downstroke and 1.9s
(S.D. 0.5s) upstroke for tests on 170 feeders. Both sets of stroke times were tested off-
site. In August 1991, stroke times at T 6 and Til were measured at 0.6/0.9s and
0.8/0.8s respectively. These stroke times were used for appropriate runs of T 6 and Tl 1
feeder designs. Stroke times were set by flow control valves and a stop watch for off-
site tests on the test rig.
2.3.4 Validity of Off-Site Testing.
It is far easier to do shot size testing off-site on cold pots than on hot pots where there
are problems with magnetic effects, crane access, heat/dust/fumes and clashes with
normal operations requirements. For these reasons, the majority of Portland tests were
conducted on a test rig off-site. A s outlined below, the validity of off-site testing was
given careful consideration.
(a) Stroke Times

Dwell time is the total of the following:
(i) wait time from DCV movement to when the stroke of the piston starts,
(ii) stroke time, and
(iii) waiting time w h e n the piston is fully extended to when the D C V is
reversed.
As discussed in Chapter 4, the wait times and the time to reach maximum pressure are
much longer for on-site multi-feeder operation compared to one feeder operating with
its o w n D C V . Stroke movement time has the greatest effect on shot size accuracy with
integrated feeders. For the tests off-site, stroke time was fixed to be similar to on-site
operation and the dwell time was varied across a 0.5s to 5.0s range.
(b) Validity of Alumina Size Variations
It is possible that the repeated use of the same alumina for off-site testing could have
given false results due to dust loss. T o replace the alumina for each test would have
required about 40t of alumina which was not acceptable. Alumina size did in fact
increase as the runs progressed (Appendix 3).
To minimize the effect of dust loss, runs were conducted in blocks of about 4 runs. I
addition, one feeder (Parker #5001 cylinder) using normal sized alumina and standard
components was used repeatedly throughout the investigation to check if there was a
significant change to the testing method. Results showed shot size changed when the
alumina changed, but the feeder combination retained m a n y of its shot size
characteristics (Figure 2-4). Note that Runs 3 0 and 31 are not sequential due to a mix
up in numbering.
For testing of the A3 feeder off-site, a worst case analysis was conducted. It was
decided to thoroughly test the A 3 feeder in the toughest manner possible in order to
draw out any failures of what appeared to be the best dosing unit design. For this
testing, alumina was replaced after each run. The alumina was taken from reservoirs in
the superstructure which are usually the finest alumina used in pots. A s will be seen,
the A 3 still proved superior in shot size accuracy.
(c) Temperature
The temperature of alumina exiting the superstructure is about 60-120 C, as measured
at Portland and T16. Arnold (44) tested alumina at ambient temperature and 100 C.
H e found that there was no discernible effect on parameters pertaining to the storage or
flow characteristics of reacted alumina. Hence, tests at room temperature should
exhibit similar characteristics to those expected at the elevated temperatures in a pot.

FIGURE 2 - 4
PARKER STANDARD CHECKS

#2 AEDD, NORMAL ALUMINA, NO INSERT
)
SHOT SIZE (g)

2000
1800 91
y'y'"
***"••**'.' .-•*'
.-**•>.*
1600 — / /'
/ y
/ y' jr
1400
// /
m M
1200
/jf
1000 STROKE TIMES
RUN 22 1.0/1.1
7
800 RUN 31 1.2/1.7
—
RUN 30 1.7/1.9
600
—
400 RUN 22, JAN29 RUN31.FEB25 RUN 30, FE828
_i 1 J L 1 1 I L ... I 1—
200
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
DWELL TIME (s)
STD DEV {%)
«. ""•nai»'*"*"* , *"»»».
_L i
R U N S 22,30,31
Note-. 1.7/1.9 refers to 1 J s downtroke and 1.9s upstroke time.
This assumption proved valid in testing on cold pots and on operating pots (refer
Sections 2.6.2 and 2.6.3).
(d) Times and Masses
Dwell time was controlled by an electronic timer. Stroke times were checked using a
stop watch. A cross check was carried out with the stroke testing unit using a high
resolution chart recorder. Times were accurate to +/-0.05s and were only quoted in 0. Is
increments. Scales were checked daily with five test weights in case the instrument
drifted. The masses were within +/-0.5g. Though shotsize were only quoted in lg
increments.
(e) Continuity of Procedure
Standard procedures were carried out for all testing. Periodic checks revealed the
procedures were followed consistently and on-site testing verified results. Only three
people conducted shot size tests so there was minimal person-to-person variability. It
was concluded that operating practices were well controlled, especially when one
considers the boring and repetitive nature of the job.
(f) Unreacted versus Reacted Alumina

The unreacted alumina behaved similarly to "normal" reacted alumina for all feeders
(Figure 2-2). This was due to the fact that it was the same size as the normal alumina
(Appendix 3). This agrees with the findings of Arnold (44) that they have similar bulk
flow properties. T o avoid any possible undetected effect on flow properties, reacted
alumina was used for all but thefirstfew runs.
Note that the above tests showed that feeding unreacted and reacted alumina gives
similar mass delivery. However, this does not m e a n that the dissolution in the bath is
necessarily similar. This is a possible shortcoming of bench scale testing of aluminas
published to date (refer Section 1.3.13).

2.4 O F F - S I T E T E S T I N G O F S H O T S I Z E
Data from all tests are provided in Appendix 2, and a summary of results is given in
Table 2-1. Dwell time of 2.5s was used for comparison purposes. A more detailed
analysis of this data is in Kissane (76). The key results were as follows:
TABLE 2 -1
SUMMARY RESULTS OF OFF-SITE SHOT SIZE TESTS
STROKE WITHOUT WITH

FEEDER
TIME INSERT INSERT
(down/up)
(s) (S.D.%) (S.D.%)
PORTLAND AEDD 1.0/1.1 7-14 2

PORTLAND AEDD 1.6/1.9 2-10 1
T6 0.6/0.8 1-2 1
T11 0.8/0.8 1-3 <1
A2 1.6/1.9 <1 n/a
A3 ALL <1 n/a
Notes: (i) Dwell time of 2.5s w a s used for comparison.
(ii) "n/a" means not applicable
2.4.1 Portland A E D D
Consider Runs 1-27 and Figures 2-2 and 2-5.

1. The main influences on shot-to-shot mass changes were as follows:
Different feeders 13%
Alumina size 13%
Type of cylinder 9%
Dwell time 7%
Stroke time 2%
Overall effect is variation of up to 3 7 % for any combination. Parker
cylinders had the worst results.
CHAPTER 2 SHOTS SIZE ANALYSIS PAGE74
FIGURE 2 - 5
EFFECT OF CYLINDER DESIGN ON

PORTLAND AEDD SHOT SIZE
NORMAL ALUMINA, STANDARD SPOOL
SHOT SIZE (g)

2000
... ... ».» »•» •»• ... •
-MW. •*: Jf-> "*• — - -j-^rajpfc*!..*-
1800
1600 '/ / /
/•' / /
1400 '•: ff
Is "/
* //
1200
'/ //
'' -fx
1000
800 h "y
600
- PARKER, 1.7/1.9 PARKER, 1.2/1.7 TERRY, 1.7/1.9 TERRY, 1.2/1.7
400
I 1 1 1 1 1 1 _| 1 1—
200
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
DWELL TIME (s)
STD DEV (%)
6
5
4
3
••>•*,
2
1 V _ . I L — . - —.-..—-»~?HI^'-«2!
0 R U N S 30,31,32,33
Note: 1.7/1.9 refers to 1.7s downstroke and 1,9s upstroke time.
2. For fine alumina, the main factors were:

Dwell time 13%
Type of cylinder 7%
Different feeders 2%
Stroke time 2%
3. Slowing d o w n the feeder to 1.6s downstroke had the following main effects:
Cylinder type 7%
Dwell time 7 % for Parker
< 1 % for Terry
Alumina size 4%
4. Dwell times less than the plant stroke times caused smaller shot size and
significant increase in variability compared to dwell times over the plant
stroke times.
5. Six (6) feeders taken at random from off line pots were tested to see if the
feeders being tested were representative of normal feeders (Figure 2-6). This
showed similar results to the test feeders and confirmed that Parker cylinders
were more affected by dwell time than Terry cylinders.
6. Shot size varied with several parameters, but was generally about l,600g at
2.0s dwell time for a standard A E D D feeder.
7. Initial tests of spool insert shape used cast iron inserts with and without
enamel coating. Enamel coating had little effect on variability. A s steel was
cheaper to manufacture and cheaper to install, steel inserts were used for all
future testing after R u n 27 and later installed in the smelter in production
designs. In the light of experience, there has been no deterioration of the
surface of the inserts in production over a period of three years; if anything,
steel insert surfaces became smoother over time.
8. For feeders with inserts, the stroke length had an effect on the magnitude of
the shot size. There was drop of 1 7 % in shot size for a 8 1 m m stroke and 2 3 %
for a 5 3 m m stroke compared to the same feeder without an insert
(Figure 2-7). Standard deviations reduced from 7-14% to 2 % .
FIGURE 2 - 6
P O R T L A N D A E D D REPEATABILITY E X P L A N T
S T A N D A R D A E D D , N O R M A L ALUMINA, 1,6s /1.9s
SHOT SIZE (g)

2000
600
PARKER.1029 PARKER.1975 PARKER.1282 TERRY,580 TERRY.474 TERRY.650
a s & m a » »•»
400
200 U
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4,5 5.0
D W E L L TIME (s)
STD DEV (%)
6
R U N S 40,41,42,43,44,45
FIGURE 2 - 7
EFFECT OF INSERT STROKE ON

PARKER CYLINDER, FINE ALUMINA
SHOT SIZE fg)

2000
1800
STANDARD 63mm STROKE 81mm STROKE

«)•«•«*
400
200 -i 1 1 r- -i r-
.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0

D W E L L TIME (s)
STD DEV (%)
6
R U N S 102,107,112
9. For most runs, the 25 consecutive points were statistically "in control" if
dwell time was greater than the stroke time as shown in Appendix 1 where the
data falls between the U C L (upper control limit) and L C L (lower control
limit) for both shotsize and range.
2.4.2 T6 Integrated Feeder
Consider Runs 1-23 and Figure 2-8.
1. Changes in dwell time and feeders had little effect on shot size i.e. about 1%.
2. Changing from normal to fine alumina dropped shot size by 6 % in the first
runs conducted, where there was a large difference in size between fine and
normal alumina. Later runs showed little drop in shot size as the size
difference reduced (Appendix 2, Runs 50-53 and 72-74).
3. Fine and normal alumina developed similarly sloped curves.
4. The S.D. was less than 2 % of shot mass for a standard feeder and 1 % if the
feeder had an insert.

5. Shot size was about l,700g at 0.6s stroke time for the standard feeder and
l,300g if it had an insert; a drop of 2 4 % .

FIGURE 2 - 8
T6 FEEDER SHOT SIZE

53mm SPOOL STROKE
S H O T SIZE (g)
2000
1800 »:*W».3»8*»* »*'»i,t*',.».w».[[|
»»-—•»»„
1800
1400
.tf •*• «*•*-"•*• «»•*•.«•• -m. "I..™ Jd *J!
1200
1000
800
NORMAL.WITHOUT.0.6/0.9 RNE,WJTHOUT.0.6/0.9 NORMAL,WiTH50.6/0.9
600 h
FINE,WITH,0.6/0.9 NORMAL,W!THOUT,1.1/1.3 F I N E , W I T H O U T S 1 . 1 / 1 . 3
400
200 l-i
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
D W E L L TIME (s)
STD D E V (%)
6 V '; 1
5 "— \i, •. i 1
~ * • 1
4 - \. I
V1 ?
3- a1
S
VI " ," *
2 —N. "• /
1 *• *
* * - *" " >--a-*-^S-
-
0 _ l 1- !• 1 1 1 1 1 1 '—'
R U N S 50,51,52,53,54,55
2.4.3 Til Integrated Feeder
Consider runs 56 to 61 and Figure 2-9 and 2-10.
1. The two feeders tested operated quite differently. One operated like the
Portland A E D D with a Parker cylinder and the other like the Portland A E D D
with a Terry cylinder. Dwell time had little effect, but one feeder (#2) was
badly affected by dwell time. Most testing was done on #2 to simulate the
worst case feeder.
2. Changing from coarse to fine alumina dropped shot size by 4 % . This was
reduced to 1 % with the insert.
3. Shot size was similar for both feeders for fine alumina, but not for normal
alumina over a dwell time of 2s.
4. The insert installed in #2 feeder brought all shots to a S.D. within 1 % of shot
mass for dwell and alumina size changes versus 2 % without the insert.
5. Shot size was identical for normal andfinealumina with the insert installed,
but the size dropped from about l,600g to l,300g viz. 2 5 % .
2.4.4 Portland A2 Feeder
Consider runs 4 to 64 and Figure 2-11.
1. Shot size changed by less than 1% for changes in dwell time (at fast and slow
speeds) and alumina size. Hence, the feeder was not affected by the alumina
(presumably due to the 60 angles at the top and bottom of the dosing unit
which assist mass flow).
2. Stroke time affected shot size by over 4 % for a 0.5s stroke time change.
3. Cylinder type changed shot size by 2 % .
4. Shot size was 900 gram (virtually independent of dwell time).

5. S.D. was similar to the A E D D with insert viz. 2 % .
FIGURE 2 - 9
Til FEEDER S H O T SIZE
SHOT SIZE (g)

2000
-
1800 -
1600 ^ r*t»v**»"** ^ •"*
, »f >
a» •*"
A* ' "'
1400
1200 - /' ' ' - .-

/ /
-ft
1000 ~ 1
_ /
800 _ f
?
- f
600
#1, N O R M A L , 0.8/0.8 #1, FINE. 0.8/0.8 #2, N O R M A L , 0.8/0.8
400
#2, FINE, 0.8/0.8 #2, N O R M A L 1.3/1.3 #2, FINE, 1.3/1.3
200 I I I U I I I . I I I
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.
DWELL TIME (s)
STD DEV (%)
6 It
5
4
3 - *Vl
2 \» »•• "" H
a
1 \ s. _ _ _ _ _ „, ** »•• .*
0 I • 1 1 1 1 1 1 a 1 1 1
R U N S 56,57,58,59,60,61
FIGURE 2 -10
Til #2 FEEDER WITH AND WITHOUT INSERT
SHOT SIZE (g)

2000
1800 r
NORMAL FINE N O R M A L WITH INSERT FINE WITH INSERT

• »« •••
200 L-1 • •- H H H 1 1-
0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.I
DWELL TIME (s)
STD DEV (%)
6
5
4
3
2
1
0
RUNS 58,59,62B,62A
FIGURE 2-11
A2 F E E D E R S H O T SIZE
S H O T SIZE (g)
2000
1800-
1600
1400
1200
1000
... ...
^•<iil»iiui[|«iaJIBgi»]B»»»a»aa*V»wTi.j«iWiijia[iiialfct,wU»'»»' la11
800
600
ft PARKER.NORMAU1.0/1.1 PARKER,F1N£,1.Q/1.1 TERRY,NORM AL.1,7/1.9
••HIâB RSR»R«XV »•• mm* «««
400 TERRY.FINE.1.7/1.9 PARKER!NORWIAL:1.2/1.* ARKER.NORWIAL,1.7/1.5 >
4-
200 "-1 u
x 4- 4- 4- 4- 4- 4-
0.5 1.0 1.5 2,0 2.5 3,0 3.5 4.0 4.5 5.0
D W E L L TIME (s)
S T D D E V (%)
6
4
af I * * $••?,$ %7-.m- ««-.*
R U N S 10,4,65,67,63,64
C H A P T E R 2 S H O T S SIZE ANALYSIS P A G E 84
2.4.5 A 3 Feeder and Sequential Feed
As the A 3 and sequential feed designs were not affected by dwell time, graphs as shown
in Figures 2-4 to 2-11 are not relevant. Table 2-H shows a comparison of the A E D D ,
A 3 and integrated feeders. Shot size testing has been carried out on nine A 3 feeders.
To test sequential feed operation, a Portland A E D D (with insert) was tested with a D C V
mounted on the top flange. (Previous off-site testing had one feeder per D C V but the
valve was mounted about 2 metres away from the feeder.)
T A B L E 2 - II
A3/AEDD SHOT SIZE COMPARISON

(ROOM TEMPERATURE)
Feeder Mean(g) Feeder-to-Feeder Shot-to-Shot Number Location

S.D.(g) S.D.(%) S.D.(g) S.D.(%) of tests
AEDD 1223 69 5.7 30 2.4 145 on-site
A3 1276 17 1.3 6 0.6 9 off-site
SEQUENTIAL 1270 7 0.5 1 off-site
A3/AEDD 25% 23% 20% 25%
Notes: 0) A E D D data from new pots testing (Table 2-IV) of A E D D feeders with inserts.
(ii) A3 from slowest dosing cylinder stroke speed off-site (10 shots).
(iii) Sequential from 2s dwell time off-site (10 shots).
Observations from the tests were as follows:
1. Standard deviation for the A3 feeders did not change over the complete range
of speeds possible from the D C V . S.D. was 0.6% of shot mass.
2. Shot size for the A 3 feeders was about l,280g, but shot size reduced by 40g at
m a x i m u m speed compared to slowest speed. A s fatigue and wear of parts is
minimized at the slowest speed, this setting was selected to provide
m a x i m u m feeder life for plant operation.
3. Shot size and S.D. for the sequential feeder were similar to the A 3 .
4. The type of D C V (Mac or Atlas Copco) had no effect, nor did the type of
feed cylinder (Ortman, Parker, K e m p e , Terry). This suggests that any of the
hardware on trial can be interchanged and not affect the feeder performance.
Clearly, the A 3 is superior to the A E D D with respect to feeder-to-feeder and

shot-to-shot shot size accuracy viz. four times better. These results m a y be
typical of other independent feeders as the D C V s and relative feed cylinder
area of the A 3 are similar to that of Pechiney and D 3 pots.
5. Note that the dosing assembly for the sequential feeder was the same as that
used on all the A E D D tests (#2). The air feed control system was different
but the dosing unit shape w a s identical. It w a s concluded that the main
reason for poor A E D D feeder accuracy compared to sequential or
independent feed is the pneumatics of the pot rather than the feeder dosing
shape.
2.4.6 Decision on Use of Spool Inserts
The main attraction of the insert is to minimize the effect of alumina particle size. If
segregation is a problem in a plant, then the plant should consider the benefits of
redesigning the dosing unit shape.
For T6 and Til feeders, shot size variability in almost all runs improved with an inse
(generally to half the standard deviation), however, shot size reduced by 15-20%. This
requires more air usage, more cycles of the feeder and more time that the plunger is
wet. A s shotsize accuracy is better than 2-3%, it is debatable if the reduction to under
1 % is worth the high cost of installation and maintenance due to increased cycles.
Note that the two Til feeders tested at Portland reacted quite differently. Maybe the
same could be said if other T 6 and Til feeders were tested as well. Other factors such
as cylinder, spool, and spring quality can affect shot size; the insert minimizes the
effects of these.
Due to the relatively good shot size accuracy of the T6 feeder and the difficulties of
shot size reduction for a plant which has a history of difficulties with lack of air, T 6
decided not to install inserts. For Portland, the benefits of inserts are quite clear and
justifiable from off-site testing results. T h e perceived benefits to pot operation were
considered to be worthwhile for the increased cycles required for a 2 5 % drop in
shotsize.
For independent feeders, the increase in air usage would have little effect on plant usage
as the feed cylinder is small. Angles of 60° at top or bottom of independent feeder
dosing units is recommended rather than the 45°commonly used. A s with T 6 and Til,
the improvement is probably not economic for a retrofit, but probably worthwhile on
n e w feeders in order to reduce variability for virtually no difference in cost for a n e w
part.
From May 1992, inserts were installed in Portland feeders with 81mm stroke.
However, it was found that in almost all cases a n e w spool and spring was required.
This took a $40 change-over to $240 which was too expensive. Further testing was
carried out on stroke length (refer to Section 2.5.7) and the stoke was changed back to
the original 5 3 m m stroke as of September 1992. B y June 1993, plant data confirmed
the benefits of the insert shape on pot performance.
Table 2-JJI shows that feeders with inserts had reduced anode effects by up to 17%.
Multiple anode effects per day ( A E P D ) by up to 3 1 % and S.D. within a pot reduced by
up to 1 5 % . Also, it appeared that the longer the insert was installed, the better the
results became as feeders installed by over 60 days gave better results than pots over 30
days old. This was due to better tuning of the computer interval between shots. Hence,
from July 1993 a program of rapid feeder changeout was employed to implement this
change as soon as possible.
T A B L E 2 - ffl
A N O D E EFFECTS FOR POTS WITH/WITHOUT SPOOL INSERTS

MONTH IN WHICH DAYS NUMBER OF UNSCEDULED AEPD MULTIPLE A E P D STD DEV WITHIN A POT
POT WAS STARTED IN POT WITHOUT WITH CHANGE CHANGE (%) WITHOUT WITH CHANGE CHANGE (%) WITHOUT WITH CHANGE CHANGE (%)
JANUARY >30 0.63 0.62 -0.01 0.13 0.12 -0.01 0.82 0.80 -0.02 I
>60 0.63 0 53 -0 10 0.13 0.11 -0.02 0.81 0.75 -0.06
FEBRUARY >30 062 0.61 -0.01 0.12 0.12 0.00 0% 0.78 0.74 -0.04 5%
>60 0.62 0.53 -0.09 0.13 0.09 -0.04 31% 0.77 0.67 -0 10 13%
MARCH >30 0.64 0.58 -0.06 I 0.13 0.11 -0.02 I 0.78 0.70 -0.08 10%
>60 0.63 0.52 -0.11 0.13 0.09 -0.04 0.78 0.66 -0.12 15%
Notes: (a) pots started after July 1992 had Inserts ("WITH") and pots with 8kg shot size did not ("WITHOUT").
(b) values listed in 'WITH" and " W I T H O U T " refers to average variability within a pot over the period selected.
(c) " C H A N G E " refersto" W I T H O U T " minus "WITH" values for >60 days worth of data.
(d)this analysis w a s calculated on 27th April 1993.
B y the end of 1993, virtually all Portland feeders had spool inserts and anode effects
were at record low levels. Since then, anode effects have fallen still further. The
inserts were an enabler to anode effect reduction but were not the sole reason for the
improvement due to considerable attention to better computer control and greater
awareness of the importance of reduced anode effect on pot performance. Computer
control could be fine tuned, knowing that the feeder shot size was n o w reliable.
2.5 SENSITIVITY ANALYSIS ON SHOT SIZE
A s noted in Section 2.4, the A E D D feeder design had quite varying results from feeder-
to-feeder and plant-to-plant. A sensitivity analysis was conducted to identify which
parameters affected feeder accuracy the most, so that design changes could be made to
minimize the variability.
2.5.1 Bulk Density of Alumina
Analyses of the bulk density of alumina (loose and packed) for different sizes showed
little significant change (Appendix 4). The average loose bulk density was 1,028 kg/m3
(S.D. 5 kg/m3) for normal reacted alumina and 1,044 kg/m3 (S.D. 13 kg/m3) for fine
reacted alumina. At a 9 0 % confidence level, there is no difference between standard
deviations of fines versus normal alumina for either loose or packed bulk densities.
However, the means are statistically different for each type of density.
As there was no vibration or any other disturbance to the hopper of the off-site te
during the tests (and little in an operating feeder) there should be no change to the
extent of compaction from one run to another. Hence, the density should be consistent
for all runs. It seems reasonable that loose bulk density is the more suitable estimate of
the density in an operating feeder. If so, one can see that the difference between fine
and normal reacted alumina is small (1.5%).
Hence, it was concluded that any change in shot size greater than 1.5% for different
feeder designs in the tests is a reflection of dynamic factors such as:
(i) the angle of flow into and out of the dosing unit;
(ii) the angle of repose in the dosing unit; and
(iii) the amount of material left in the dosing unit at the expiration of the dwell
time.
For almost all variables tested, the differences seen were greater than 1.5% so they
more important to shot size accuracy than bulk density.
It is also postulated that a volumetric feeder is an adequate meter for feeding fixed
masses of alumina to a pot. There appears to be little requirement to explore
gravimetric feeders. Shapiro (71) investigated such a device in 1981-83. This device
achieved standard deviations of 1.3% - 3.6%. Testing in Portland has identified
standard deviations of under 1.0% are achievable for A E D D feeders with inserts (Table
2-1) and standard A E D D feeders at T 6 and Til achieved this without inserts. Hence,
the significant extra cost of a gravimetric feeder is not justified.
2.5.2 Flow Through Inaccuracies in Dosing Systems
The spring in a Alcoa feeder pushes the spool downwards w h e n the cylinder extends
(Figure 1-9). It must do this as fast as possible so as to minimize unmetered flow of
alumina through the spool. Once the cylinder starts to m o v e down, immediately the
alumina starts by-passing through the spool until the spool stops. If flow through
occurs, this unmetered alumina adds to the variability of the shot size.
As flow is affected by alumina size and degree of aeration, spool movement in an
A E D D design is affected by the quality of cylinder and spring (which always touches
the spool). For independent feeders, flow through also occurs, but the spool is
controlled by the small pneumatic feed cylinder, cylinder design and air supply. A s the
D C V is mounted close to the cylinder, and there is only one D C V per feeder, there is
more repeatable control of piston movement. In both types of feeder, the speed of the
spool affects shot size accuracy by varying the amount of flow through.
2.5.3 Effect of Springs
Shot size testing was conducted with various spring configurations (Runs 113-129).
The observations from this testing included:
(i) the spring condition affects the standard AEDD shot size.
(ii) high pressure (Kempe) springs m a k e the feeder less sensitive to cushion
quality.
(iii)spools with inserts are less sensitive to spring condition.
All Alcoa, Alcan and Alumax integrated feeders until 1993 used 302SS or 304SS
springs of identical shape (until this study recommended to a change to high pressure
steel springs).
The normal free standing height of the standard AEDD stainless spring reduces due to
creep from exposure to heat (Figure 2-12). A s the spring reduces in length, it reduces its
spring pressure which in turn changes the characteristics of the spool movement. A soft
spring will not act as well as a brake for the spool on the upstroke, and will not force
the spool d o w n as hard on the downstroke. Less braking on the upstroke gives less
bypass and better shotsize variability. The net effect of the slower downstroke speed
(and deceleration) is a tendency to retain alumina. This causes a slow dribble of
alumina to flow out, so the shot size is affected by dwell time and the variability
increases. A sharp downwards action tends to throw the shot out of the feeder giving
less variability. The downstroke effect is considered to be more important than the
upstroke as witnessed by the effect of varying dwell time on shotsize.
The Kempe spring has been designed to cope with higher temperatures and to exert
high forces at the extended and compressed lengths it is exposed to in the feeder. Its
force does not drop to the original value of a n e w stainless spring until it reaches over
C H A P T E R 2 S H O T S SIZE ANALYSIS P A G E 90
500 C. The stainless springs start to lose strength at 250°C and to shrink in length at
300 C. Spring temperature reaches well over 300°C on some feeders (Section 6.5.2 and
8.4.2:2).
FIGURE 2 -12
EFFECT OF TEMPERATURE ON SPRINGS

300 r 600
p - KEMPE HEIGHT
i - STAINLESS HEIGHT
— KEMPE FORCE
— STAINLESS FORCE
25 200 250 300 350 400 450

TEMPERATURE (°C)
Note: Springs were subjected to 18hrs for each 50°C temperature increments
Force tests on old and n e w springs are illustrated in Table 2-iy From this table, the
following observations can be made:
(i) W h e n the spool is fully d o w n (cylinder extended), stainless springs drop in

force by 6 0 % after they have shrunk by 4 0 m m . A reduction of 4 0 m m is
c o m m o n in m a n y plants.
(ii) The closing force of stainless springs, even when new, is very small viz. 40-
120N. It does not take m u c h restriction by foreign objects or alumina to
prevent the spool closing off flow from the superstructure. This can lead to
by-pass of alumina into the pot on a continuous basis during the wet time.
This alumina is not measured and affects shot-to-shot variability.

(iii) K e m p e springs exert a 1 5 0 % higher force w h e n compressed than a n e w

stainless spring. However, the force is m u c h lower for both springs than the
force exerted by the cylinder, so this does not affect the speed significantly.
Hence, o n the upstroke, the spring does not affect alumina bypass
significantly.
(iv) The closing force of the K e m p e spring w h e n the cylinder is extended is

about double that of the stainless spring.
(v) Use of length as a criterion to reject springs (which is often done in repair
shops) is unwise. O n e must use force, as the force can drop up to 2 0 %
before the length reduces.
Portland started installing Kempe springs in July 1991 and has installed about 1,300 to
July 1994. (Refer Section 8.8.3 for discussion on spring failures)
2.5.4 Effects of Cylinder Design on Shotsize
2.5.4:1 Description of Cylinder Operation
From Section 2.4, one can see different cylinders (Parker, Terry, Lindberg) had
different effects on shot size of integrated feeders. This was due to cushioning, seal
type and seal condition. These factors are discussed below, but let us first consider
cylinder cushioning. Figure 1-20 shows the critical parts of a typical pneumatic
y
cylinder and Figure 2-13 shows h o w cushioning occurs.
A pneumatic cylinder operates by application of air to alternating sides of a piston in

barrel. Cylinder manufacturers design cushioning based on the calculated kinetic
energy of the mass being driven and the anticipated compression of the air to stop the
piston in a controlled manner. The clearances between the piston rod cushion bushes
and the fixed orifices of the cylinder mounting plates dictates the cushioning ability of a
given cylinder design; the closer the clearances, the better the cushioning. For many
cylinder designs, there are cushion adjustment screws to throttle the rate that air is
exhausted from the cylinder. O n e can increase cushioning by restricting exhaust air
flow.
CHAPTER 2 SHOTS SIZE ANALYSIS P A G E 92
FIGURE 2 -13
CYLINDER CUSHIONING OPERATION
CUSHION SPEAR
CUSHION BUSH
CUSHION
ADJUSTMENT
SCREW
The main* air supply to the rur and of the cylinder is rapidly distributed over the'entire
piston face with the aid of a non-return valve, built into the cylinder end cover. This drive*
the piston o n the out-stroke, and air forward of the piston is exhausted through the front
port until -die seal of the cushion boa* enters the forward recast The remaining trapped air
then escapes slowly through ai adjustable needle valve, thus providing a cushion affect ai
the and of the piston stroke.
O n the reverse operation, (in-stroke), the cushion effect la achieved in the same manner.
Adapted from Martonaire C I O 2.2.69-17 Brochure, P g 17,1.4.77
If cushioning is poor, piston rods or pistons can break due to excessive deceleration
rates. This has occurred at Portland, T6 and T17 (Section 8.6 and 8.8).
The presence of flow control valves at most plants helps correct for poor cushioning
settings during maintenance. These allow regulation of back pressure air to the
cylinders to control the speed. O n e will find that most operators or tradespersons w h o
look after feeders are aware that changing the settings on these flow control valves stop
the feeders from "banging". This is due to inefficient cushioning. At the above
mentioned plants which have piston rod failures, there are no flow control valves so the
cushioning m a y be ineffective. Clearly, there is a link between cushioning and fatigue.
In most situations, cushioning efficiency has not been perceived to be a problem for
maintenance, probably because cushioning is difficult to test in a workshop. Hence,
cushioning has not been examined in any detail. However, feeder cylinders have a
significant effect on shot size which few people (if any) have examined to date. This is
more important than maintenance, as this affects pot performance.
In mid 1991 a cushion test rig was installed at the Portland Kempe workshop to test
feeders after they were overhauled to ensure cushioning was adequate. A similar device
(though not as good a design) was installed at Boyne in 1993. Ensure that any cushion
test equipment is extremely anchored and that there is zero movement in the frame
when a feeder is tested. If the frame has any flexibility, it will affect the accuracy of
the measurement. This was discovered on the Parker production test rigs where the
deceleration for cylinders increased by a factor of 10 w h e n cushion feet were removed.
This can be assured by installing (for a once off test) an accelerometer on the frame.
2.5.4:2 Effect of Cushioning on Shot Size
Independent feeders have the spool driven by only a very small cylinder (40-50mm
O.D. x 4 0 - 5 0 m m stroke) that solely operates alumina dosing. Integrated feeders are
driven by a m u c h larger cylinder (100-150mm O.D. x 300-550mm stroke) but the
piston rod has to drive not only the dosing unit but also the plunger shaft and plunger.
The driven masses are about 10kg for independent feeders and about 30-50kg for
integrated feeders. In order to cushion the larger driven mass, integrated feeder
cylinders need m u c h more extensive cushioning otherwise the piston will impact on the
stationary metal stops and result in fatigue of piston and piston rod. Unfortunately, this
larger cushion has a deleterious effect on shot size accuracy.

For integrated feeders, as the cushioning is engaged, the speed of the driven mass is
affected. A n y change in speed in the cushioning area of the cylinder is directly related
to the spool speed as the piston rod is directly coupled to the spool of the dosing unit.
Hence, the speed of the dosing spool is affected, especially as the spool starts to m o v e
and w h e n it slows to a stop.
For the AEDD integrated feeder design, the piston is attached (by the piston rod) to t
plunger shaft which has a collar just above the plunger (Figure 1-11). W h e n the piston
is about 5 0 m m from the end of travel on the upstroke, the collar strikes the spool at the
bottom surface (the cup). The piston spear hits the cushion sleeve about 2 5 m m from
the end of travel for integrated feeders (Figure 1-20 and 2-13). Then the piston slows
d o w n progressively to a stop just before it hits the end of stroke. Hence, the cup is
struck before the cushion is engaged at a high speed...about 0.4-1.4 m/s (Table 3-1).
Independent feeders have m u c h smaller cushion lengths (10-20mm) and the velocity is
m u c h smaller (due to the short stroke length), so the cushioning effect on speeds is less.
In both independent and integrated feeder dosing units, alumina can flow unmetered
through the dosing spool as the cushions are engaged. The longer the cushioning, the
potential for by-pass is worse. Thus, the longer the cushioning length or the tighter the
air clearances between moving and stationary cylinder parts, the worse is the shot size
variability. Testing of the two Til feeders showed that the one with the tighter
cushioning (#2) had worse shot size variability than the one with sloppier cushioning
(#1). Refer to Runs 56-61 and Section 2.9. The "best" cylinder design (i.e. the one
which is least likely to fatigue) is one with good cushioning. However, this is expected
to give the worst shot size variability.
Shot size tests for Portland AEDD (Figure 2-14) verified that having no cushioning at
all on a cylinder gave more consistent shot size and had less variability than the
installation of an insert. Inspection of cushioning by the author at 13 plants during
1991-93 showed a variety of clearances. In the case of cylinders at Portland (Terry
cylinders), T 6 and T15, there was no cushion adjustment screw, so speed could be not
FIGURE 2 -14
EFFECT OF CUSHIONING ON
PARKER, 1.6s/1.9s
SHOT SIZE (g)

2000
»:CS*IWK!!.*l usssa. asyrtsâitftvcuf*
STANDARD S3 INSERT STANDARD, NO CUSHION S3 INSERT, NO CUSHION

RiPVntf if S V •*• ••
0.5 1.0 1.5 2.0 2,5 3.0 3.5 4.0 4.5 5.0
DWELL TIME (s)
STD DEV (%)
6
R U N S 129,127,123A,124
Note: 1.7/1.9 refers to 1 Js downstroke and 1.9s upstroke time.
be adjusted. The cushioning was poor. Lindberg cylinders at Tl 1 showed uneven wear
on the cushion spear of the two examined at Portland. The one with the most wear (#2)
was the one with the worst shot size results.
To minimize the by-pass effect in dosing systems of integrated feeders, the spool in
shape was designed to incorporate a shoulder at the base so that the dosing unit hardly
opens w h e n the cylinder is in the cushions (Figures 1-11). Also, the steeper angles (60°
versus 45 ) allow all the alumina to exit quickly from the dosing spool so there is little
cushioning effect at top or bottom of the spool movement. This has proved to be
successful as evidenced by discussion throughout this Chapter.
2.5.4:3 Effect of Seal Material/Quality
Six feeders (three Terry and three Parker) from off-line pots showed a spread of shot
size as illustrated in Figure 2-6. This illustrates the variation from feeder-to-feeder due
to wear differences.
A more detailed experiment was conducted on one feeder with old and new seals with
and without inserts (Runs 130,131,132,134). This showed up to 1 0 % greater shot size
variation for old seals, but this reduces to under 2 % if inserts are installed (Figure 2-
15). Thus, once again, inserts minimized variability.
The type of piston and rod seals affect spool speed. These seals may be Teflon (as us
in most Alcoa/Alcan/Alumax integrated feeder plants) or Viton (as used Atlas Copco
and C P O A C cylinders at several independent feeder plants (Table 1-H), and, until
recently, on Scheffer, Hanna, Terry and Parker cylinders at Portland, T8, T9, TIO,
T12). Teflon is self lubricating and hence allows faster movement than Viton which is
a rubber compound. Portland changed to Teflon rod seals and piston seals in 1992 for
all Terry cylinders overhauled. Trials are underway on Teflon piston seals on Parker
cylinders at Portland.
FIGURE 2 -15
EFFECT OF SPOOL DESIGN & SEAL AGE
NORMAL ALUMINA, PARKER #1248 EX PLANT, 1.6/1.9
]
SHOT SIZE (g)
2000
-
1800-
,/\. —
-*^
T— "" "• '"
1600 ^^^ —*
/ ^ —
f »**
1400 *r f v********9***^
' J /
1
1200 ¥•"'
' ''/''
t y/
H'/J
1000— ¥tf/
w ^
- ,'/ f
800 ft
1/
J
'/
_/
600
OLD SEALS,STANDARD SPOOL OLD SEALS WITH S3mm INSERT
400 NEW SEALS, STANDARD SPOOLNEW SEALS WITH 53mm INSERT

— ••» • •• "^ *• ^
200 l i t 1 J 1 1 1 1 J-
0.5 1,0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
DWELL TIME |s)
STD DEV {%)
R U N S 130,131,132,134
The flexibility of the material can have a significant effect. For example, the Viton
used for Parker cylinders is a hard material, but that used by Atlas Copco is very soft so
it can take a higher temperature excursion before becoming brittle.
The quality of the seals can also affect speed. If there is air by-passing the rod sea
effective air pressure on the bottom face of the piston is less, resulting in reduced piston
speed if the manifold air is limited. Note that rod seal failure is one of the most
frequent failures of feeders surveyed (as evidenced by ore leaks in Table 1-V). Piston
seals are one of the most c o m m o n causes of failure of crustbreaker cylinders on
independent feeders and are probably just as bad for Alcoa and Alcan pots (but they
have not been monitored in most plants). Reports by tradespersons at all eight A E D D
plants visited by the author in 1992 suggest a very high piston seal failure frequency.
A cylinder that has recently been overhauled will be slower than an old one (due to
tighter clearances on seals and bushes) and this consequently affects alumina bypass
through the integrated feeder spool. Shotsize testing of n e w and used feeders on start-
up pots at Portland (Table 2-TV) has shown that old feeders have a slightly more
accurate shotsize compared to n e w feeders viz. 50g S.D. versus 69g S.D. with inserts
(or 69g S.D. versus 147g S.D. for no inserts). This is expected to be due to wearing of
seals and increasing the velocity of the dosing unit.
TABLE 2 - IV
SPRING FORCE COMPARISON

RUN SPRING NOMINAL ACTUAL NEW FORCE (N)
TYPE FREE FREE OR SPOOL STROKE FROM TOP (mm)
HEIGHT (mm HEIGHT (mm USED CLOSE 10 20 30 40 OPEN
A 304SS 260 250 NEW 206 191 171 152 137 118
B 304SS 260 250 NEW 211 191 176 157 137 118
C 304SS 260 210 USED 132 113 93 69 54 39
D 304SS 260 215 USED 144 127 108 88 69 54
E KEMPE 190 180 NEW 700 600 500 420 330 260
F KEMPE 190 177 USED 554 470 392 314 240 171
G KEMPE 210 210 NEW 583 500 426 348 279 196
Notes: (i) "Closed" height is equivalent to when plunger is fully retracted (130mm spring height).
(ii) "Open" height is equivalent to when the plunger is fully extended (180mm spring height).
(iii) "Spool stroke fromtop"is distance in m m that the spring is extended from a closed height.
(iv) Forces for K e m p e runs E & F were using 3 0 m m spacer with the spring.
2.5.5 Effect of Dwell Time
Section 2.3.1 showed that increasing dwell time increases the shot size on most Portland
feeders. O n e would expect a change in shot frequency in the plant if the dwell time has
changed on all pots.
At Portland there are equal numbers of Terry and Parker cylinders. The shot size tes
of six (6) random feeders (Figure 2-6 and runs 40-45), showed that for a change of
dwell time from 2s to 4s the shot size increased from l,710g to l,795g; ariseof 5 % .
For a fixed current supply to a pot, the pot requires a fixed amount of alumina. If the
shot size increases by 5 % , then the shot frequency (which is automatically controlled by
the potline computer) will increase by 5%. In 1992, the dwell time on all pots was
increased from 2.5s to 4s as a result of slowing feeder speed to improve fatigue of
piston rods and spools. It was found that the pots required about 3 % more frequent
feeds to maintain stable alumina concentration in the pots.
Although the plant change was less than that expected, it was of similar magnitude.
The difference between the 5 % predicted and 3 % actual drop was probably due to the
different pneumatic controls for off-site and on-site feeders.
2.5.6 Effect of Height of Alumina in the Hopper
Level of alumina in the hopper did not affect shot size from 1.4m to 0.6m above the
dose unit (Appendix 1). N o testing was done for alumina height beyond these heights
because they relate to a Portland pot superstructure m a x i m u m and minimum alumina
inventory respectively.
2.5.7 Effect of Insert Stroke Length
The original spool stroke length is 5 3 m m for most A E D D plants. The original spool
insert design increased this to 8 1 m m as the insert had a 3 0 m m shoulder at the bottom
end that m a y restrict the discharge (Figure 1-11 and 2-3). Inserts were installed using
the 8 1 m m stroke from M a y 1992.

A s noted in Section 2.4.6, 8 1 m m insert stroke was too expensive and action was taken
to test if 53 m m stroke was accurate, and what was the effect on shot size. Comparison
of the two strokes in Figure 2-7 indicated that the 5 3 m m shot size was
(i) 23% smaller than the standard AEDD (versus 17% for 81mm stroke).
(ii) more repeatable.
(iii) less sensitive to alumina type .

(iv) less sensitive to cylinder type.
Hence, it was decided to change to a 53mm stroke, effective from September 2nd,
1992. B y December 1993, inserts were installed on over 9 8 % of Portland feeders. The
2 5 % drop in shotsize predicted by the off-site testing was evidenced by a similar
increase in feed frequency in the potlines.
2.5.8 Effect of Spool Manufacture Quality
Investigation of the quality of the spool dosing unit at Portland in 1991 showed that the
spools were not concentric. Similar results were observed in 1992 at T 9 and Til.
Figure 2-16 illustrates the Portland data on 60 random spools showing that 9 0 % were
not concentric. Non-concentric operation can result in intermittent or complete
jamming of the spool in the assembly due to the tight clearances. The spools should be
spun in a lathe to ensure they are built to original drawings if a n e w spool is installed in
a A E D D feeder. At Portland, about 1,000 spools were made of larger diameter pipe
that fouled in the assembly and had to be replaced.
Analysis of the dynamics of operation of the spool also detected that in all AEDD
feeders a partial vacuum forms in the spring chamber (Figure 1-11). W h e n the spool
moves downwards, a partial vacuum is caused due to the increase in volume in the
spring chamber. The only w a y air can generally flow into the chamber is from the
dosing unit itself. Alumina will also attempt to flow into the chamber through the tight
clearances between spool and feeder assembly which can lead to jamming of the spool
in the assembly causing:

(i) ore leaks,

(ii) not feeding, or
(iii) variable feed rates.
This can result in variability in shot size and intermittent alumina dosing. This
phenomenon will be generally undetected by operating personnel, as the feeder will
appear to be working correctly (unless the spool is totally jammed).
F I G U R E 2 -16
SPOOL W E A R POINTS
DISTRIBUTION OF WEAR POINT AT

THE TOP OF THE SPOOL PIPE
3% RUN OUT POINT

AVERAGE = 0.7 m m
STD. DEV. = 0.4 m m
no wear at
all = 10%
Note:
(i) sample size: 6 0 spools taken at random ex plant
(ii) measurements taken w h e n spool in lathe
(iii) tests carried out in January 1991 at Portland.
(iv) "Distribution" refers to the percent of spoors tested that had
wear at the points indicated at the top of the spool.
About half the feeders removed at Portland for "not feeding" up to early 1994 were
probably due to this cause as there was no mechanical fault with the feeder when the
feeder was overhauled after service. From early 1994, these factors were addressed at
Portland by drilling holes in the assembly to reduce the partial vacuum by 7 5 % (Figure
1-11). Immediately the frequency of j a m m e d objects dropped (Section 8.8).
2.6 ON-SITE TESTING O F SHOT SIZE
2.6.1 T i l - C o l d P o t s
Plant Til carried out shot size testing of feeders with and without inserts in cold pots to
verify the benefits (if any) of the insert design and to gain a better understanding of the
accuracy of A E D D feeders (76).
These tests on regular feeders confirmed the tests done at Portland on two feeders fr
Til. T h e shotsize for different feeders varied between l,640g and l,770g - an 8 %
change. Shot size was larger than Portland tests of l,500-l,690g.
Comparison of results of in-plant testing at Til and Portland shows that Til had a
better S.D. than standard Portland feeders viz. 1 % versus 3 % (76). This is probably
related to the fact that Til has only 3 feeders/DCV (100 x 3 5 6 m m cylinders) versus 5
feeders/DCV (125 x 5 0 8 m m cylinders) at Portland (refer Section 4.6). Both plants
have similarly sized D C V s so the Portland feeders are relatively starved of air. It is
easier to start the Til feeder spools, so the spool is quicker to open and close at the start
of the upstroke and the end of the downstroke.
Benefits of the insert were not evident at Til probably as the original design had a
good variability anyway due to less feeders/DCV. It is likely that other plants with 1-3
feeders/DCV would find the insert m a y not give a significant improvement to shot size
accuracy. Thus, the effect on pot efficiency m a y be insignificant.
2.6.2 Portland - Cold Pots
From August 1992, each feeder in n e w pots has had its shotsize tested prior to start-up.
A sample of this data is given in Table 2-V. Using data from 229 feeders, the benefits
of inserts are clear...61% improvement in standard deviation within a pot, and 3 7 %
improvement feeder-to-feeder for n e w feeders. Reused feeders showed a 3 8 %
improvement in standard deviation within a pot and a 2 0 % effect feeder-to-feeder. The
reason for the S.D. for all feeders having a larger value than that within a pot is due to
the variability of pot-to-pot changes because of different D C V and air supply
characteristics. This verified the off-site testing.

TABLE 2-V
SHOT SIZE TESTING ON NEW POTS
NEW FEEDERS
INSERTS N O INSERTS
MASS RANGE CALCD STD DEV S.D. AS % M A S S RANGE CALCD STD DEV S.D. AS %
WITHIN P O T OF M A S S WITHIN POT OF M A S S
A V E R A G E PER
1223 53 23 1.9 1661 190 82 4.9
POT
STD DEV F O R ALL

69 5.6 147 8.8
FEEDERS
N U M B E R TESTED 145 19
IMPROVEMENT FOR FEEDERS WITH INSERTS: STD DEV WITHIN POT

DROP IN SHOT SEE WITH INSERTS: I STD DEV FOR ALL FEEDERS
REUSED FEEDERS
MASS RANGE CALCD STD DEV S.D. A S % M A S S RANGE CALCD STD DEV S.D. A S %
WITHIN POT OF M A S S WITHIN POT OF M A S S
AVERAGE PER 1.6 102 44 2.6

1239 46 20 1670
POT
STD DEV FOR ALL 4.0 69 4.1
50
FEEDERS
NUMBER TESTED 43 22
IMPROVEMENT FOR FEEDERS WITH INSERTS: S T D D E V WITHIN P O T

DROP IN SHOT SIZE WITH INSERTS: ISTD D E V F O R ALL F E E D E R S
Notes:
(i) All mass in g.
(ii) "Range is the maximum minus minimum of individual shot sizesfor5 shots from 5 feeders.
(iii) Calculated standard deviation within a pot used conversion data from Janus (96) to convert a range to a S.D.
(iv) S.D.forallfeedersis calculated as the pooled S.D. of all thefeedersin the sample size e.g.forn e wfeederswith
inserts, the S.D. of each of 145feederswas squared and the square root of the average was 69g. This represents
5.7% of 1223g, which is the average mass taken from the 145 feeders.
2.6.3 Portland - Operating Pots
There is always an element of doubt over room temperature tests as they may not be
applicable at pot operating temperatures. T o verify the effect of the inserts and to get a
better feel of other factors affecting shot size at operating temperature, tests were
carried out on pot 1076 #1 and 1069 #1 feeders.
A special single outlet chute was installed to allow greater ability to catch all the
alumina from the shot. A large pan on a stick was used to catch all the shot and the
samples were weighed. Tests were carried out on a variety of feeders. A series of 20
shots was taken from each standard feeder. Then the feeder was removed and the spool
was changed from a standard to an insert design. The feeder was reinstalled in the pot,
and a further 20 shots were taken. A summary of results is shown in
Table 2-VI and detail comparisons are in Tables 2-VII to 2-VIII.
The percentage drop in shot size was similar to that found off-site and to that found
the plant on cold feeders i.e. 2 3 % drop in shot size off-site (Section 2.4.2 and Figure 2-
7) and 2 6 % on-site on cold feeders (Table 2-VI).
In almost all cases (8 of 9 runs), the insert gave less shot size variability with an
improvement of 2 3 % shot-to-shot. The S.D. from feeder-to-feeder comparisons was
4 3 % less for the insert. The drop in shot size from "without an insert" to "with an
insert" was 2 3 % . Once again, this verifies the insert improves shot size accuracy
(despite the fact that the shot size is less).
The magnitude (in grams) of the standard deviation was always better with an insert,
and the S.D. as a percentage of the shot size was also better. A s with on-site cold tests,
a S.D. of 3 % was found. Operating pot testing is very difficult due to restrictions of pot
operation, crane requirements and awkwardness of catching all the shot on an operating
pot. However, from these tests, one can conclude that off-site cold pot testing gives
comparable results.
The insert was designed to minimize the effect of alumina size and dwell time changes
A bonus was the observation that it improved variability feeder-to-feeder. This was
tested on the operating pot by using the same feeder on two different days (Table 2-VII)
and for different dwell times (Table 2 - V m ) . This showed that in both feeders tested on
different days, the insert has a better variability. The alumina type affected the standard
feeder by 1 0 % but only 3 % if an insert was installed.
ITi
o
rH
W
O
<
OH N
CO
cu U)
N Q- V)
CO E _
CANT (90% ji>
ISTICA LLY
ID O
a. tf) o
E a> o-
CO ro
CO
UJ
O
LU
O CO O
LU 2
O CO
LU
o ui S>">
z >-
2
z 2 cu ro TJ "a
H u. >- >- >-
> cu to ^
ra
^ co1
co CD
roro
 O
CO ^_
^_ £ io ^ csi
(A cu u_ CO
Ho
O) CO ID in CO CO
UJ O 3 1^ CO CN 5 CN
CD
co
CO
•*
CJ) o x:
(3 # roD) O •- CO
2 *-. 5 0> CD CO
X
< .—v c o " C — in
•S » in
CM h- cu £ CN
Ui IB CM d d d d d *- d d d -^ .S Q. y
H
t
o CO <0 CD
ro
k_
8£
co ^ CU
> «, D . C CO
•a
O 1- y v cu CO ™ <°
D-
£^
O £ co
•o "O CO
OH X a.
co O
CM
CN
CO m ID CJ) m o
CM CN CM CN CO
a> CN
T_ CO
CN CO en
co « ro £E
o a:
Q
cu * 8-g
o.
i-i CO CJ
TJ
ro "Z
O
^^ • g fi to c
z r- in CO CO CO *"! cp CN in CO O)
X
cu « t X
TJ E
ro OJ
a: CM CN -
CM CM CN CO CO CM d oCO o £ <o 0> CD
H UJ (0 "
o g « c
CO >
CO O
Cd 2
co a. Q-£ r-
.£ (D £ &
X , ^ c toen «- 3 ™ x B.
s I— 3 ^— CM *—
_i
O
CO CO CO CO
£! CO
• *
*-
CO CO
CN CM
CO- CN CO
" CN
3
o E
_3
o
ro
S ro O _ J Z d> * o
10
T^ 111 CN ^ E
< ID T—
*— |-~ co CM CO a> CO a> § _ tf) 0J
o
Q_
CO
-sr CM 05 CD T — 00 00
• t f
f
CMM
CO O «r o £ CL C CD CO g
CN o CO >§> CN CM CN CM CM CM
o E cu J? ^ ^8.0)
W ^= 35£ a^ c O) CO CO
H ., tf) cu .£ JO c ro
to CN m CO CO N; CO oo CO O ^
oc O ;£.
cu E o «
gE5
_i a. = <=
2 Cd
*C—
O
-tf CO x~ CO T — T-^
csi CM •tf CO T—
E < > CO <S
« O •-;
o
a. o o8 ,,8
c <» B 5 8.£ Q. O
H CO o mcu - a. OJ t tf)
Q 3 8 s<D 8-EC
Cd
OC
CO O
ID CM
ID
ID
CO
CN
CO
CN •J 5 a>
CO 5? 1^- tf) in E t c T3
CN
c CM co (0 — >- TJ 4S —
LSI <
<a o C c
•a •« c <5 TJ »p
a k—
CJ)
o .E 6 : <°
55 2 a. a) — ' CN O O ffll ^
in to <S JS
#££
cu • * CO CO CO CD CO CO 0) CO (A ocuCN
£!
co <*« s"Ec
• t f
O) OO CO in CO tf) CM i2 T -
H co
3
in CO CO Ln CO in co ro * CM CO 10
3 in in ID in in E
o
o .a
CJ) »•- <o E o
§8 $
o ^^ a.
o cu
c °
o- _
!fi 2
c^ c CN •§ ^0) »t .^10
10
8.
rvi
CN J"
LLl > > > > > > > > > UJ r> <£ S
i- 4fc L_ — - .—.
CO JZ CN tf)
o o o o o o o o o Q S-- -P to
TJ to
o O) H 2 2 z 2 2 z z z 2 > fc a
CM CO ID d>6 *— CO CO < co jz!
*; *
'>
ro CO .2
S
• ^
fif |i
rn ^_IS
CO CD<»n £
P
CM CN CN 4 c: co -"
O) < CO TJ
CN CN co to oo •-
TIt u
"- Q E 10 o c o
CO y-^ ,- •>f O H o CO o ^_ 1- «« •o
3 co S "O
-g
CO
r •— to <n
(0 CO TJ
0>
% 3 CM CM
co CO
UI 3 co
Ui 8 = §
Co r- b C
tf)"?| c?
CO
§o
H H
?l
tf)s"o_
E » co to » ^ t
o 2 w £ "
CYLINDER
a: OC >- OC >- > a: S = •COcu CU £ TJ ro oj a,

TYPE
E E Q-
Ul UJ OC UJ OC il
OC LU .!! ID c : X):
<
OJ
So •x.•xOC o: a: OC OC DC
X. (0 t o; 9> <o
><
UJ
Ql OC UJ < r- UJ UJ LU
01 a „ CO =
O g> CO
co Si 9 O ) o <"
CO - c o
r- r- < _i £ to co
< < r- a. _i < < <0 CD P
£ E?
5| Q-
< = > J=
K
a: a. a. < < CO
CO ~ CO
CM t*- o o
CM W t^- o— O
en
z
—, S5 CO i^ CO o 5 OO T
•<*•
«— T — o
•<*
to
m m OJ
H o z o
O
X 2 H
r/3 => O T—
CM CO T in CO t^ 00 O)
CN Ql -I
3
SO
o
I—I
•%
PL,
(90% ^-^,
LLY
1- < 0 (0
UJ
o O
co z >- z z
b Co
£
W S
O
CO
^lit
_
1*-
Ui CM to
CM in CO
o
O S?
z E _
I ^^ So
o # to a> IN CO Q E°!
Hi •—.. b *- ci d ft*
^CM
N ^, o
CO tf)
0) 9 a
CD
h- •5" 03 £ TJ
o 0.
I O) § re c tf)
o
CO T- m
CM T
CM _ a> <S a u
CO O re reOO
OC J=
CD To
d c
Q a> o «o
l-
JJ co StQC).
CSI ^ ,_: CO m O 9
oc CV
w
tsi Ui
>-
to
CO CO 5 <?
z .ggg ii
•D
W
HI tf)
.re o- o JZ
z S r e b) c
T—
T-
3 CO CO CO o CN CM o o (1)
CM •"J" B a n
• fe ra
« u (7)
© 8 re c
_l J) CM
* >«-s
8 s 00
OO CM
CM
to
CO
CO
to
Ui
O)
00
CM
CM CM O - £% S
-I
2
W
0-
Q
or
3 r- 1-
1
<? «? m
in
0)
to
in
CM
£ I
c 1Z HI
Q J re i? » 0) tf) o
<z
OO CD O) to 00 to » re § Si 2
1 o •>»•
ja c re
to 8 m ID
UI 8 c? ui O E Q) 5
3 g
> >-
SJS 8. °> °- S^
•^ sS •g OJ
^ c re T- tf) P
> > 8. Oof)
s
&•
- to "or
UJ
> > < r- > >
Ul
CM W
cil e « 8 «j? §. o
zo Zo O
z o z zo o
zo CO y ^ o O £
CO 4 Ui r- <
o d>CM
z
a.i- y E O D.
CM CM or
Ui
U-
1 u.
z
UJ O
CO u.
u. or b z
i
c
CM =
in E oS
CM as » in J
d M V CM ^5 o> & «-
# s^
Q c
u. QT^ w w Q
u. S ^ CO0
QS CO r £ o re d •= ?= to tf)
CM u>
S3 o
t-I-
V)
co 8^ S-g -Pre
Ui
r- LU
KB §1 E £ -2 re
•s
a. re Q .
3 .1 ° C) «"31
or o o
r-
re to 5
UJ T— § or or
Ui UI
5
a)
5 « .2 ol g o »
21 «
u. or or m
or or E si; |'l si- o 4j tS
O u. ra
UJ UJ -g-o •O TJ t» m o c p P
O CO H H O 2 Lf lis T3 I -o re t - T J i S S £
TJ -5
z z
tn p £ P
K
or or
CM
or
CM
ll re -o —' « !
— re jr
*= CO OJ
g>
o o U.
E^ « OJ tf)
fe • »
O 8 8 S = »•gm •£
CD
i Sa p.E-E
CO
|i Z •"-
*- z tn m
118.
CJ z O
CO
or
O
c
oo
r
»M £ .» >?2 t»p °o.s ogre
ta =
ro 6 <*
P £ TJ
< «
IS Lf r>- 00
Lf
O)
•S 2 -S :
?Ctf)P
2 ~ .tJ
re = re
p c"°
• »
o ^, © g ^^ s re C F ^
tf)-^
or -J
8 8 Z >=-' < < CO i
> C
re — re
c
H «^°
o g5«^
s
CN
oi
s
u
r-
o
i-H
O
- o tf)
U)
=i a p p
tf) CO
< i- (0 (0 a. a.
rt
o z UI 111 o E
0) Q.
E. i
s h- <
co cj
>• 5"
p
a>
ro
a
P ra
oj ra
E o
s^s
" • * > •
&o
0 M P
ro
£ «
O SZ
5 to o 3 ° S CO
ro .•= P ^ o
s S co ro cu
P > io Q.
ro P * 1
p I
" TJ TJ <o gco

tf) cp P ro
LU ,-
N #
cu a ± «
o. p -e x
p
x a) p 65 •A P
-5 LB
o CD m
_ o IT) in to • P "> S £
£ ra " I CO
CJ UI ° -= J=
csi co co ^ > • c
— +2
ra ^r Q. " . — —:
roj=£^_.T-tf)c
S O *= O TJ
a x— CSI CJ) en c J5 C (0 _ o ^r
N- 0
> Q.2 * fe
Ifllf
E OJ
ro J= ftO)
CQ
to£fe m i: to m tf)
p
< CO - O o p ® I P
p .£.
0
^^
z z •s ..
fe c fe c ra t
H g- fe •3 P O 3) T> o 5. o
o |
-J to 8 IS c i c t (0 „
-J co co CO CO en CO o tf) c !» iS fe Q.
1 UI O CM
W
co co ^r >- z 8 I o £ 5 i ~ i 11 (0 tf)
ft 8e O Q. i§ D-fe ^
CM tig
ffi P
ro .2
if) til
3-.E
P c"-S
B^cS
5 cc -C SI
<ocsi^
1 C -TJ
1* £ drad a - g1 COTJ
c.3?
s
60
T to ^r
co ? in
2
s in to
o
8 CM
to
Ui
>
&S
Q ra CD
•T CM j ; CM
8 8-2=
P pJS
S ra o
T)Js o
8 fe
•a tf)
•o P
8 ifsli — (0 jr fe ro fe
ti >-
-i
V
1
-, T J ra TO fe
E o
in III
CJ _J r- fe © to p J^§ £ E c » o ) r-
o z
O ro
2 LU < or ^ < Z = S3 o, TJ g y ro ^ ro — QJ •^ -£
7" h- < UJ T7 h- < IB O- E ca5 to JR t5 - Q . „ = 0)
» »~ lz
UL
CJ III (O u
LL
o CJ III o u
)ZU_
0 ( / O) CD £ P £ TJ CO c E
UI
< O tr Z b J; m P
Q tr m CO JJ D)^-.tf2^j-E fe
(9 X irrn l- o in I
ac ci-o CO
o p o U) tf)
_tf)
tM fe y
J7) p 0
Ife.
c/> z o u. to t/J o u LL
in ^S ra P ro
en <
N 3 5 5 5 8) o iS
HI UI
o EE
o s 3 3 o
o
=- ro
or Q
or co co UI
W UI
O J 0.
a CO
t/j
H
o
CN
s
04
CO
u
In addition, Table 2-VIII shows comparison tests of different dwell times and different
speeds. The Is dwell time change affected the standard feeder by 5 % versus 1 % for the
insert. The change is speed affected the standard feeder by 1 1 % but had no significant
effect if an insert was installed.
All these tests verified conclusions reached for off-site testing. The insert had achieve
better performance of shot size accuracy and shot size standard deviation for the worst
case testing i.e. on an operating pot. Also, the tests verified that cold testing is
satisfactory to give trends and that operating pot testing is not necessary unless one
wants to test the complete feed system.
Thus, in conclusion, the assumptions made at the start of the study (Section 2.3.4) have
proved to be valid.
2.6.4 Portland - A3 Feeders
Shot size accuracy for A 3 and A E D D feeders in operating pots was conducted to
compare their relative accuracies (Table 2-IX). This showed that the standard deviation
for the A 3 w a s only 21 % of the A E D D feeder-to-feeder and ws% of shot-to-shot on
h o t pots.
Note that the S.D. for hot pots was several times that of off-site testing, possibly due to
hold-up in the chute and inaccuracies catching the shot when anodes were close by.
The shot size for the A 3 varied little for different dwell times, but the A E D D was m u c h
more sensitive.
Thus, the independent A3 feeders were more accurate and less variable than the
integrated A E D D feeders in operating pots.
It is interesting to note that the accuracy of the AEDD with insert, the sequential and A3
feeders is better than the Pechiney and Boyne integrated feeders which are quoted as
lkg (+/- 50g) (ref 101) and 1.8kg (+/- lOOg) (ref 102) respectively for non-operating
feeders. Thus, the insert in the A E D D has taken a feeder with poor accuracy and raised it
of the best in the world.
TABLE 2 - IX
A3/AEDD SHOT SIZE COMPARISON

(OPERATING TEMPERATURE)
Feeder Dwell Mean(g) Feeder-to-Feeder Shot-to-Shot Number

Time (s) S-D.(g) S.D.(%) S.D.(g) S.D.(%) of tests
AEDD 2.5 1318 55 4.2 88 6.7 15
4.5 1299 112 8.6 64 5.0 15
Ave 1308 89 6.8 77 6.8
A3 3.0 1245 17 1.4 37 3.0 14

4.0 1250 16 1.3 36 2.9 14
5.0 1249 20 1.6 48 3.8 14
Ave 1248 17 1.4 40 3.2
A3/AEDD 19% 21% 83% 48%
Note: (i) Samples taken at weekly intervals over 3 weeks to include

variability due to alumina size (20 April to May, 1994).
(ii) All feeders had single outlet chutes.
(iii) All A E D D feeders had Atlas Copco cylinders and inserts.
2.7 EFFECTS OF PNEUMATICS O N S H O T SIZE
The pneumatics of the pot can affect shot size accuracy because of the effect on piston
speed. T h e pneumatics of the feeding system includes cylinder size, number of
cylinders per D C V , D C V size, air pipe length, air pipe size, exhaust muffler size and
condition. For Portland pots, there are 5 feeders operating off the one D C V yet off-site
tests used one feeder per D C V (as is the case with most independent feeders in
operating plants.
In order to quantify the effect of multi-feeder/DCV versus one feeder/DCV, one feeder
(Parker 5001 cylinder with #2 Portland assembly) w a s tested off-site and on-site under
cold conditions. T h e alumina that w a s sampled in the on-site operating pot w a s
retained and reused o n the testrigusing, once again, the same feeder. Equivalent stroke
times were used to cater for different response times of the valves. T h e results are seen
in Table 2-X.
^^
Z r-ZUl
-J Z
< < 9
ATI STIC
IGNIFIC
LU CO
UJ
O o o (0
UJ
O O CO O CO
z >- z z z > •
z z
LU
>- z
LU
>-
t- CO o
CO
to §.
Q
u w
UI *r
O
o o co
co CN CO
CO
CO
CO
CO
CO
CO m
m
CO
tf
CO CO
CO
to O #
to z

to
to
>
Q£ o T— "e— CN CN CN CN CN CN CN > co
cc CO H-
O Q
*- o
to O a)
CO CN CO a> oi
c» d"- o CO ro in T •tf ro
or • t f
CO CN ^ d CO CN CD
LU S 8
H Ul
CO
to >-
to to
to z to •£ ID
to
X
H H CO CN m in • t f ro co £ CN • t f
CO
8| c
to to CO CN •tf
CN CO O-.b
EZJ _l
CO >
H o
o E o
LU ,- 05 K CO CO m CO m ro CO
O a.
S in tn
o o
O) CO w
CN
•
CN
* CO
CN
00 CO
CN
in
CN
O
i
CO
CO
LU Ia
* CO >-
as (fl to
en ro
c/j
.-*. c c CO -->
H o o 00 oo CO CO • > -
• t f co CO CO o CO
LU % c "c ' • > ^
CN co CN CD >- c c TJ T-
O O CO
• t f
"" CO
8 8 pCO wo
io
OH o
a.
CO
Q
S ^CN_ CD CO CO • t f N- CO CO ro • t f
in in
c
o
•n
9„i
,_». <L> ID
ro .p
c ID
T3 •g 0>
TJ .P
r 0)
r co S
to OC CN CN CN CN CD CN g 0)o IDo
to • t f
CD D
ts |
CO CO
a. <a SP. '« >
(1) > ~ 8
z > (D
0)
"D T3
O P
< UJ h-
^_ ro O ro 00 r- CO ^ ro CO .£ to
"E "Se 8.
C/J CO
~tf CN m In CO
•tf in
o ro in co •tf UJ CO CO J3 CO
in CD i
i
to O •tf CD m >- •a T3
>
to
• * « c c
« !
Ig
CO (0
o
05
UJ
Ul
s
?
o
ro
o
in
to
to co 1- to to to to to to to to to to UJ
h_i i-
k_
1
_J in o m o m o m o o m CJ < z 2.
O LL
LL. UJ
_l T-^ CN CN cd CO tf • t f in • t f
co" z
Ul 0.
o <
H o CO
OC co LL a.
to o §
Q Ul O ro
O II
OC
1- z c
LL a CD
a £
CO CO or T-
to (0 to
CN 2 ^ > p
to to 5 s
o UJ
H h-
ra
$
05 Ul to
CO co 45
2
z l- 1-
Ul = r
S O _i
Z
_i
>> >
a o
to to
CN
to
CO
to
•tf
5
in
o
z
LU
r-
(Si
LL
LL
i
JtoU o
c
p
^-^z
f^r
1
o Ul
o O
<Si \-
—'
A-*
o
u UJ
r- CO
UJ
H z
r-
LU
z
co r- UJ _l
1 CO
u. LU
1-
o
LL Z o
LU
O
Both on-site and off-site tests improved with and without inserts. The variability off-
site was about 60-69% of on-site testing using exactly the same feeder and the same
alumina. It was concluded that the reason for the variability is the different pneumatic
effects of the "single feeder/DCV,r off-site versus "multi-feeder/DC V" for on-site tests.
Tests on sequential feeders showed that the shot-to-shot S.D. was 0.5% (Tabl
versus 1.1% for the same feeder with an insert (Table 2-X). In both cases the same
feeder assembly and spool were used, so the difference must be because of the
pneumatic control of the feeder. The valve used on the off-site test rig was about 2
metres from the cylinder yet the sequential feeder had the D C V mounted on the feeder
itself. Thus, the design of the pneumatics of a feeder has a significant effect on the
accuracy of that feeder (Section 4.6). This establishes that there is a major difference
between the speeds and reaction of the feeders with valves close and far from the D C V
and that one feeder/DCV is superior to multi-feeders/DCV
This also illustrates the effect that the proximity of the valve to the feed
size accuracy. The closer the valve to the feeder, the better the results.
In studies of the response time of feeders to DCV activation (Section 4.6), t

performance of five feeders in a pot was dramatically different from the situation where
there were only one or two feeders in the pot. Stroke speeds are much faster if less than
5 feeders are installed...about 0.5s for two feeders versus 1.6s for five feeders. Hence,
the operation of 4 or less feeders in a Portland pot will affect average shot size to the
pot. This suggests that "dead" feeders need to be changed as soon as possible after they
are detected, rather than just rely on the pot to adjust itself.
2.8 DIRECT FEED INTO THE HOLE
The standard A E D D feeder chute is an open design with no direction of flow towards
the hole (Figure 1-9). Study on the Alcoa feeder has found that up to 1 0 % of the shot
does not reach the liquid bath due to blowing of the alumina away from the hole (27).
The concentrated stream from a single or double outlet chute minimizes this loss. It is
inefficient to optimise the dosing unit to achieve under 3 % accuracy if there is up to
1 0 % loss after the dose is metered.
As discussed in Section 1.2.13, the preferred design for good dissolution of alumina i
fast dosing into the bath. This occurs in independent feeders where the plunger is out
of the w a y of the dose (Figure 1-12 and 1-13). The c o m m o n independent feeder
designs add a horizontal velocity component to the alumina which should help mixing
in the bath. It should be noted that despite all the activity on chute design, the hole is
often partly or fully closed in most operating pots. T 1 7 estimates few holes are open,
Til estimates less than 3 0 % and Portland had 3 0 % of holes open (refer Section
6.3.6:3). D 9 (which has A P 1 8 independent feeders) ensures that each hole is opened
every 32 hours, but m a n y holes are usually closed.
A single outlet chute aimed at the holes directly assists the alumina in the hole area
either to
(i) fall immediately into the bath (good mixing), or

(ii) allow the alumina to be directed into the path of the plunger on the next
crust break.
The latter results also in good mixing in the bath as the alumina is pre-heated prior
entry into the bath. In fact, at D l , the feed/crustbreak cycle is set so that the dose
occurs immediately prior to crustbreak so the crustbreak actually forces the alumina
into the hole. The A E D D feeder tends to spread the alumina around the feeder hole, so
alumina cannot feed directly into the hole (Figure 1-10). However, w h e n it does spread
around the hole, it is preheated which is good for dissolution w h e n it does slip into the
hole.
It is also desirable to have the chute outlet as close as possible to the hole to mini
dust loss. However, the closer the chute to the crust, the greater the chance of build-up
on the plunger which in turn m a y foul the chute. Unless the cylinder is strong enough,
the plunger m a y not fully retract. For integrated feeders this causes the alumina to leak
or not to feed. Pechiney has a "dag scraper" to cut off the dag. Examination of Figure
1-11 and 1-12 shows a shroud around the crustbreaker plunger to clean off dags.
However, this is still not effective for n e w pots where dags often occur (Figure 1-2 and
1-3). D a g cleaners in Alcoa and Alcan smelters have all proved ineffective and most
have been cut off. This is due mainly to the small cylinders (100-125mm O.D.) versus
the large cylinders (160-200mm O.D.) on independent designs. The latter cylinders
have about 4 times the force to knock off the dag.
At Portland there is insufficient space between the dosing unit and the plunger hole
fit a single outlet chute with angles over 60 (as suggested by Arnold (44)). A single
outlet chute with an angle of about 45 was tested but alumina tended to hold up in the
chute (Figure 2-1). Alumina discharge tended to stick on the wet plunger and buildup
dags which in turn fouled the chute.
About 5 different chute designs have been used at Portland to address optimum dose
delivery (Figure 1-22). The final double outlet chute largely solves all the above
problems by raising the chute above the high dag level to stop fouling and allow the
alumina to pass under the plunger into the hole (Figure 2-17). This chute fits the
A E D D , A 3 , sequential feeder and virtually all A E D D feeders installed at
Alcoa/Alcan/Alumax plants.
However, comparison of 10 pots with double outlet chutes showed no statistical

difference in performance to standard A E D D chutes. It could be that the preheating
effect m a y be more important than the direct feed that Reverdy (24) recommends.
If independent feeders break a hole every 2-3 feeds, the alumina tends to bridge the
hole (unless there is a large plunger penetration distance). This negates the desired
benefit of direct feed suggested by Reverdy (24), so Pechiney feeders have a deep
plunger penetration to ensure a hole is continually provided at regular intervals.
However, this contributes to greater plunger wear (refer Section 5.4.2). T o correct for
this, Pechiney use stainless steel plungers to ensure long plunger life.
FIGURE 2-17
A2 FEEDER WITH DOUBLE OUTLET CHUTE

2.9 FACTORS AFFECTING SHOT SIZE ACCURACY IN POTS
Based on the discussion in the previous sections of this Chapter, there are m a n y factors
that can affect shot size accuracy in pot alumina feeders; a summary is given in Table 2-
XI. Integrated feeders are significantly affected by 32 factors and independent feeders
by only 20. Clearly there is m u c h less chance of variability for the latter design.
The importance of these factors varies from plant-to-plant...even for the same type of
feeder design. For example, the results of Portland were worse than T 6 and Til,
despite all three being so called "identical" A E D D feeder designs. Even within one
plant, the variability from feeder-to-feeder can be very significant. In the case of two
Til feeders tested, they had quite different characteristics due to the quality of
cushioning and Portland had different results for Terry cylinders and Parker cylinders.
Thus, each factor needs to be addressed by each plant to determine if they have a
problem. Routine shot size testing needs to be done at all plants; this is rarely
undertaken (except for Portland since January 1993).
However, for both integrated and independent feeders, significant improvements can be
made by simple attention to quality of manufacture, operation and design.
TABLE 2-XI
FACTORS THAT AFFECT SHOT SIZE

INTEGRATED INDEPENDENT POT
- D C V condition
- variable supply pressure
- Mounting - loose fittings shake feeder
- Restricted hoses
- Muffler quality
- Non vertical mounting
- Hold up in chute under super
- Restricted flow control valve
- Alumina loss between feeder and liquid bath
ALUMINA
- Angle of repose i.e. 0 - 55 degrees
- Angle of discharge i.e. 60 - 90 degrees
- Alumina restricted to feeder - entry, kidney plate
- Insufficientflowtime to fill feeder before next shot
- Fixed computer dwell time yet variable feeder
response time
CYLINDER
- Different cylinder designs have different movement
characteristics
- Speed down over first 5 0 m m and last 5 0 m m of up stroke
- G o o d top cushion causes alumina by-pass,
- Seal type .... Teflon allows faster piston stroke speeds
than Viton
- A g e of cylinders ....New ones are slower
- Seal condition ...Leaking rod seals cause fast speeds
down and slow up
- Seals catch and release randomly
- Piston not perpendicular can cause vibration
- Foreign objects in cushion ports
- Tight clearances under heat
ASSEMBLY
- Spring tension... low tension increases variability
- Spool alignment not central
- Loose objects jam spool
- Non central movement
- Alumina jamming spool
- C u p jammed on assembly
- Tight clearances under heat
- Spool bottom face not horizontal causes fouling of spool
- Spool not central causes fouling of spool
MEASUREMENT
- Not collecting full shot
- Dust loss
- Weigher gauge capability
- Not consistent dwell time
- Hold up in chute causing carry-over to next shot
32 20
2.10 M A J O R F I N D I N G S F R O M S H O T S I Z E A N A L Y S I S
(i) The shot size from Portland feeders was affected by alumina particle size (13%),
different feeders (13%), type of cylinder (9%), dwell time (7%), and stroke time
(2%). U p to 3 7 % difference between one shot and another on different feeders
was possible.
(ii) Spool inserts reduced shot-to-shot standard deviation from 7-14% to 2 % at the
original Portland stroke time.
(iii) T 6 feeders showed little variability due to poor cushioning on the cylinder.
(iv) One Til feeder was similar to Portland feeders with Parker cylinders (poor) and
one similar to Portland feeders with Terry cylinders (good). T 6 operated like a
Portland A E D D with Terry cylinders.
(v) Spool inserts reduced shot size S.D. (generally by at least 2 5 % ) for Portland, T 6
and Til feeders.
(vi) 5 3 m m inserts had better results than 8 1 m m inserts but reduced the original shot
size by about 2 5 % (versus 17%).
(vii) Inserts reduced the sensitivity of feeders to spring and cushion quality.
(viii) Independent feeders are generally more efficient then integrated feeders for shot
size variability.
(ix) The A E D D with insert, the sequential feeder and the A 3 are more accurate than
the Pechiney and Boyne feeders.
(x) Seal life and feeder age affect feeder speed and shot size; old feeders are more
accurate.
(xi) 32 factors m a y affect shot size accuracy for integrated feeders including
mufflers, feeder location in pot, cylinder type, assembly, spring, and mounting.
Independent feeders m a y be affected by 20 factors.
(xii) Shot size testing on operating pots found that one feeder/DCV is more accurate
than multi-feeders/DCV.
(xiii) Shot size testing on operating pots verified the off-site testing.
(xiv) Cold new feeders on Portland pots showed 6 1 % improvement in S.D. shot-to-
shot and 3 6 % improvement in S.D. feeder-to-feeder if an insert was installed.
(xv) Chute design can affect effective shot size into the liquid bath.
(xvi) Independent feeders and integrated feeders with sequential dosing are superior
to the integrated multi-feeders/DCV feed systems with respect to shotsize
accuracy and standard deviation.
2.11 RECOMMENDATIONS FROM SHOT SIZE ANALYSIS
(i) Spool angles to be 60 at top and bottom to improve shotsize variability.

(ii) For multiple feeder operation with one valve, m i n i m u m dwell times should be
set for n e w feeders at the farthest position from the control valve such that the
movement time is complete to ensure complete shot is delivered.
(iii) Ensure that seals or bush clearances are sufficient to not hamper the speed of
newly overhauled cylinders.
(iv) Install high temperature springs on A E D D designed integrated feeders to ensure
quick operation of the spool, to avoid inefficient sealing of the spool and to cope
with higher temperatures.
(v) Overhaul feeder cylinder cushioning routinely on refurbishment to ensure no
effect on piston speed and to minimize fatigue.
(vi) Install cushioning testrigsusing accelerometers to test feeders after overhaul.
CHAPTER 3 DOSE DELIVERY PAGE 119
CHAPTER 3
DOSE DELIVERY
3.1 SUMMARY
A survey was carried out at Portland to establish the reasons w h y crustbreak plungers
do not break a hole in the crust. The cause of nearly 900 blockages were traced.
The major causes were new set carbon (28%) feeder repeaters (24%), low bath (19%),
pot repeaters (12%) and pointed plungers (7%). "Repeaters" were feeders or pots
where there was a continuing problem of blockages. Feeder repeaters comprised only
2 % of total feeders installed. Only 3 % of pots were repeaters but these comprised a
total of 3 6 % of blockages. N e w set carbon was also the main cause at other smelter
locations which conducted similar surveys.
For surveyed pots that had several anode effects per day, about half were due to
problems related to blocked feeders. The major cause of these anode effects was
turning off the feed and no one turning it back on in time. This resulted in about 2 0 % of
the anode effects.
Suggestions for improvement included longer dwell time after setting, better anode
cover practices and close control of pneumatics quality.
Investigations into pressures and kinetic energy of crustbreakers at 13 smelters sho

that high kinetic energy is not necessary to ensure the feeder hole is clear. Air cylinders
of 1 2 5 m m O.D. are adequate if attention is taken to cover practices, pressure at the
crust face and dwell time, but a 1 5 0 m m O.D. size is a safer option to cope with pot
variability in bath level and anode cover practice.
C H A P T E R 3 D O S E DELIVERY P A G E 120
3.2 I N T R O D U C T I O N
A basic requirement of a pot feeder is to ensure that a hole exists in the crust so that the
alumina can enter into the liquid bath. Generally, most plants standardise on a minimum
of 5 0 m m O.D. plunger width closest to the crust.
This basic question has largely been overlooked in the design of point feeders. Close
examination of material at the target area shows this is not a "crust". The area around
the hole is a hard fused crust, but the hole area is mostly a bed of loose particles of
anode cover or alumina in the path of the plunger.
If this really was a "crust", then how did it form so quickly... within 3 minutes? No
hard crust takes m a n y hours to form. For example, the hole which is used to tap metal
can be broken by a crow bar even after 24 hours.
The duty of the plunger is to penetrate this material in a vertical path. This
understanding is critical to design the crustbreaker adequately. High kinetic energy
(KE) plungers with pointed tips surfaces are not a good means of penetrating a granular
material. Pressure by itself also m a y not be effective if the material arches
significantly.
There are several techniques that the industry has used to "guarantee" a hole, but th
all have drawbacks (Table 3-1).
TABLE 3 -1
TECHNIQUES TO BREAK A HOLE IN THE POT CRU
Principle Method Drawback
High pressure large cylinder high air usage, cost

small plunger O.D. hole can bridge, short life
High K E fast speed fatigue, wear
high mass fatigue
Deep penetration dip plunger in bath plunger wear
O n e solution to all blockages is to increase the O.D. of the cylinder to such a size that
the pressure is so large that the plunger will always pierce a hole. Pechiney use large
cylinders ( 2 0 0 m m O.D.) versus 100-160mm in non-Pechiney smelters. The Alcoa
experience is that 1 0 0 m m is adequate but 1 2 5 m m gives a little more insurance (as
verified by Portland tests).
It is desirable to optimise breakage rather than increase cylinder size. The latter
an option at most plants. This Chapter examines w h y holes block and a comparison is
made of the high pressure and high K E plunger operation.
3.3 IMPACT OF ANODE COVER ON HOLE BLOCKAGE

Arnold and G u (79) examined the material flow properties of anode cover and the
factors that could affect the hole blocking. Conclusions from this study included:
(i) Cover Size

Large particles bridge easier than alumina, but a mixture of both is even worse;
even for holes up to 1 0 0 m m diameter.
Anode cover size is selected according to pot design and heat balance
considerations. M a n y smelters use alumina or minus 3 m m material as anode
cover; this material is unlikely to bridge easily. The minus 2 0 m m sized anode
cover used by Alcoa/Alcan/Alumax plants bridges more easily. Generally there
is little choice of a plant to change the size of anode cover as the mix is a
culmination of a material balance of returning anode cover from old anodes.
(ii) Gas Flow

Bridging will occur more easily if there is an up flow of gas.
The better the overall cover practice on the pot, the faster the gas flow through
the feeder hole, and the greater the chance of blockage. A "gas hole" would be
beneficial i.e. a hole always placed in the cover to let the gas exit away from the
feeders.
(iii) Cover Height

The deeper the cover and the finer the material, the more difficult it is to
penetrate the cover.
Once again, the more the pot is covered to reduce heat loss, the worse is the
effect on feeding. If alumina or fine bath dust is used as cover, it is more
difficult to break a hole.
(iv) Minimum Plunger Size

A 3 0 m m O.D. plunger penetrates the cover more easily than a 5 0 m m O.D.
plunger.
This observation verifies the reluctance of operators to remove pointed (50mm

O.D. or less) plungers. If it still breaks the hole, they prefer to leave the feeder
alone. This action has the drawback that the hole m a y block due to a size
change in anode cover or a drop in crust height resulting from a drop in liquid
level in the pot.
(v) Hole Size

The hole size required to stop bridging varied significantly based on cover
depth, cover size and gas flow. Virtually any hole will bridge if given the worst
of these conditions.
Clearly, the smaller the available hole area, the greater the likelihood of
bridging.
(vi) Vibration
Vibration significantly increases the chance of bridging.
This is usually not a problem in a pot as there is almost no vibration present.

(vii) D r a w d o w n angle
This angle adjacent to the hole w a s 40-45°(from horizontal) for all types of
cover tested viz. < 4 m m (mainly alumina size), > 4 m m , and 5 0 % mix.
This suggests that the hole will be unlikely to block if the angle of cover is
allowed to be at or below 40° so no material can free fall into the hole.
Integrated feeders have a crater over the hole. A n extreme example is that
shown in Figure 1-10. Independent feeders feed directly into the hole so there is
little to no crater (Figure 1-13). Thus, independent feeders are less likely to
block.
It is concluded that the mechanism of blockage is one of flow of alumina and anode
cover into the hole and bridging. The better the cover practices (deeper, less gas holes)
the greater the chance of blockage. Hence, the most likely times that blockage
problems will eventuate are
(i) directly after covering anodes when anodes are changed (set),
(ii) after pot dressing (general maintenance of cover material over the anodes),
(iii) after rapid feed (usually after anode effect), or
(iv) if the cover is vibrated by bridge movements (during tapping of metal or

resistance change).
If a granular mix is used as anode cover there is greaterriskof blockage. This theory is
verified by the observation at all Alcoa plants that the greatest frequency of blocked
feeder holes is directly after the anodes are set and covered.
Portland pots have some of the worst conditions for blocked feeder holes viz. large
anode cover size, 6 2 % of anodes are set next to feeders (versus 3 0 - 5 0 % for most other
pots), small gaps between anodes where feeders are located (due to high percentage of
carbon area in the pot). A survey of causes of blocked feeder holes at Portland
identified the situation w h e n blockages occur; the results of this survey are discussed in
Section 3.5. Results have since been proven to be similar at other Alcoa plants.
3.4 CRUST BREAKAGE TECHNIQUES

The Alcoa/Alcan/Alumax technique for crust breakage has generally been by use of
high kinetic energy. Hence, high plunger speeds have been a feature of their recent
plants except for T 4 which chose slow speeds. Plants such as those of Pechiney,
concentrate on high pressure (150-200mm O D cylinders versus 100-125mm for
Alcoa/Alcan/Alumax) and low velocities (3s stroke time versus 0.4-1.0s respectively).
Alcoa/Alcan/Alumax integrated feeder plants not only have smaller cylinder diameters
than independent feeder plants, but also often have lower mains air pressures. However,
these feeders still break holes successfully. This suggests the cylinder diameter
preferred by independent feeder plants is excessive. Use of large diameter cylinders
results in high air usage, so these plants needed to develop a feeder to compensate for
this...the independent feeder.
Which technique is better...high KE or high pressure?
Sixteen smelters participated in timing the stroke times of their feeders (75). This
used to calculate kinetic energy, to ascertain if there was a minimum level required for
feeders to break a hole. The mains pressure at each plant was combined with the
plunger diameter to calculate the pressure at the crust. Results are summarised in Table
3-II and Figure 3-1.
CN
I—I
W ^^
TJ
O a C
UJ (0
<! O •e
in in in <D -^ to in -v •*• \n in in inO 03 03 fl) s
OH §UJ o T- a>
a. O O O O O O O O O C J O O O »- »- 6 0 0 d r—
Q. ^ c
53
in w ra
> UJ a.
h- fK
o a. "TO o o o o o o o o o o o o o 0 0 0 _. 0 _ 0
UJ
LL. 0
LL a.
j^ n
0 (n0 n
(Oo Ni« o1 0n(nO o
t til oC I
i In) n( Dn (p! )) in a> 0 8 a> 9
UJ CM
>
* «
O
a: =5 a. , ,
* ^ « N ; « « p C M 0 q ^ ^ ' ' « J a > . C O 0 ID s n 0 v CO
UJ c 3 O O O O O O r ^ ^ o O O O * - d d ^ d T^ ^ d
z
UJ r«
UJ £
Hl
> t. * 0)
TJ CL
z E c
-j 3 q i n i n s f f l ^ s p i T - ^ ' t N C ) 0
Ul
§
O T - d d d d d d d d d d d r -
01 1-; CM *- 00 -<T T—
•t-1 d d d 0" T^0
0F |g
0
ro
0 in Q_
or Q ^ <*—
a) ro 0
> a r-0) 0 c
TJ 3
5^to C 0
< O r-© ra 3
,-v ra
>.
c
W O t
W N r i S O J ^ ^ t D a j f l O i - a ) O «> O T O *- 10 ra 0 ro > 0
a . -J rJ>
o ' d d r d c i ^ ^ d d d ^ r •^ 0 -^ d ^ ^ d ID
fcF ?a CO
In£ *§ in
UJ ^ c m ro Q.
O) .S
UJ
>
> OJ
is 5» CN
O * *= P a > 0 0 l ~ - ' < - ; t v . o > T ' ^ - C D 0 O 0 0 - < - 0 0 , >.CN

n 3 OJ a
OC : O CD O T O ^ m 1-
L_
c (D •*-• c
i-
w < o V ddd^dd^^ddd^-r^ T-' d ^ 0 v^ -^ d ID

£ 3 is -° O)
O
111 in
1-J
0_ S >o g * «
OJ
O
E
£ 0C -"Og TJ0)
a. -1 J,
"• L U ^
ID o> n ftTJ c >.
UJ o a q s o n n i O N O J O s ^ 0 LO 0 0 0 m co (0 r- 3 0)
0
in
5 6 c i ' - d r : d d d v : d d d d ^ ^ r N r ^ d CM OJ ra
h- 0 b 3 a-
0)
E
> * « 1
rac W O
Ê
J=
4-i
O
X 7= O
w Sot ^»(0(i>(ONO)(ointocoa)t 0 T m co tv. *-•q-
raCO b
TJ1-
OJ
CL »- JS
-J 1 ^ d d d d d d d d d d d r •^ d d d d T^ d
O tfJZ 9- !S E
0. - n£ II) i
3n 0)
3 in
Ul •*-
O O *
ts— 11 ro .82
PQ UJ O 0) o > CO
< X >-
> in 1/)
3 0
O * fc ~ a) ra iinn 3
in ?
—
c a £ 2 i8 e.
Of
h- < O Tt o o N p q w o a o o o i f f l : : :
n in q ^ q >t UJ
T - - « - ^ - T - T - O T - O O T - T - ^ - T - t- 0 d d d T- d ra ro Q. D.-0 0)
z S og a. ID ra JS
TJ
OJ
§ Q. >
-I E JJ J= c 3 1 •
O "- UJ —' * ^ >. E
•0 3 0 c £
Q 0) Q . a
fcn ® n OJ
s
UJ
E 0 > CO OJ
0) ii ro a.
inr~r--(Diôo<DO>T-iî^tD'>» q q N q q q m
5 o" o d ci o o d d •<-" o" did ci O r r N r d CM OJ ( A a It: OJ <D
l- m kO—J CD 3 TJ
ra m .— ror
TJ r m
0 fc 3 *0) CN
* ro Q.
co O 10 TJ fl) D >
5 CO > n TJ
CO LT
PQ oia>(i)<oooiDintDNoas O l~~ T- T- |V. CM a>
m "0
OJ
ID C
2 or d o ' d d ^ d d o ' d d d d d *" ° "" *~ "~ "•"" d = c
3 3= = Ul
=
H * 0- UJ <^ 0
C/J * = ^r .2 >
3
2 CO
w
m c B U J ' < - T - T - T - T j - ^ c o c O t - ,' T O CD CD ^ Iv. CM T
• > » - ' * ' * ' < J ' l f ) ' ^ ' ^ C M C O C 0 C 0 ' J C 0 LO co m in 00 <o •*r
5
u or
LU
<
w H SZ ci
H
o o o o o o o o o o o o o
o o o o o o o o o o o o o
m m in in in io 0
CM C M C M C M C M C M m
OJ
O
> -J IiO
>-
w o
Q LU ^•" 00 -
u z
_l
«-N«<33<«®5JS<< <<<<<< <
w 10 ;_;-
o
Q
Q Z
m n ffl s N
O IO » M S ^ " ° -i CM CO T 00 ° co
! -ht-hi-fpffpiifs H f- J- h- r- r.a
5! a. OC ^
00.
U
o
t/J
HH
tf
< E o
CU o ^_
0. o
§
O
U
u © © °- _,
S
C/3
t/5 3 — _ C5
w «
2 « *
-
.^ +•> —
a
<D
- *2 > c
>-
z
< E ~
=
=
a> 2 a.
i_ Q.
><
!-H
1
0
ss _l
O 1-
Q.E
>.2
Cl 1- « (9'91)9U <9'91)9U.
rt ui 0
U z a-
w * O
ai z
nts e
s u re
s u re
i. D
the
Ml "" <»-n)9u <»-n)9u
•£ 111
P w o c. « •O •U •^ :m
0 u r < (8'9'9"l)9U
—
- °- ,Q_-'°
a) •»
HH in -i (8'9'91)SU
HH *~^ ,_ j ; •> (a -
H 5?.
_ ~ o> *.
•2 „ •= = <=
— u i? (U
U4 0
Q UVnJsu. (z'»-n)su - o c
Z E
N
<» > £ a
° • = <o a
"" » O t3
cu
rt SS'^2
a. ™ > "
=
w E = „N
Ul 5 Q- > 0
"; 5™
ID ".£ oj c
«< D) O *- «
w 61 = ® - CT
rt E> <°
^-v " ~ >
co
H
P (n)si (ntei
rt
u (Z1)9±
(n)9i
H
Z <n)9i
(n)9i
HJ
CH
3.4.1 Kinetic Energy

The downstroke kinetic energy values were widely different - some up to 4 times that of
other places, and the ratio was no different for 1 0 0 m m O.D. or 1 2 5 m m O.D. plants.
Hence, the cylinder size was not significant and clearly the K E is not significant. For
1 0 0 m m O D plants, the variation was a factor of five, and for 1 2 5 m m plants the factor
was 12.
Pechiney's specification for crustbreak cylinders is 0.5-0.7 m/s between stops.

However, most Pechiney smelters operate at m u c h slower stroke speeds. (In Table 3-II,
the peak velocity is double the average velocity.) Stroke speeds of this study varied
between 0.2 m/s and 0.9 m/s. Note that the slowest rates are for 1 2 5 m m cylinders
(except Portland and T10).
If one considers KE values on 125mm O.D. plants alone, Portland could drop KE by a
factor of 10 which relates to a stroke time 3.2 times the original stroke time. Thus, a
2.5s downstroke appears acceptable for Portland to break a hole based on data from
operating plants.
One would surmise that high KE is not the most critical factor in breaking the crust.
it were, then all plants could slow their feeders d o w n to that of the lowest K E plant, and
benefit from lower friction wear on components.
Generally, wear rate of seals increases by the cube of the velocity, so any piston sp
reduction is expected to have significant benefit if friction is the root cause of seal wear.
Most plants would be very nervous about reducing cylinder speeds for fear that feeder
holes would block and anode effects would increase.
3.4.2 Pressure
The pressure applied by most of the integrated feeder plants is similar and is the lowest
of the plants surveyed. Using a 1 2 5 m m (not 1 0 0 m m ) cylinder greatly enhances crust
pressure; by 5 6 % . In addition, the plants using 1 2 5 m m cylinders (and T7) also have
higher supply air pressure. The combination raises applied pressures to double that in
integrated feeder plants which use 1 0 0 m m cylinders. Once again, there is quite a
difference between plant pressures and yet all plants survive. However, the smaller
cylinders will be expected to have more problems with crustbreakage from pressure.
If increasing air pressure is not possible, then reducing plunger diameter may ensure
crust breakage for lower pressures, but at the expense of plunger life.
3.4.3 Stroke Time to Break the Crust

In 1991, w h e n Portland was testing slow crustbreak speeds, there were very strong
opinions within Alcoa and other companies that this was very dangerous due to the
belief that high kinetic energy was needed to break holes. For the previous 30 years,
feeder design and operation was based on m a x i m u m K E in Alcoa and plants that used
Alcoa feeders e.g. Alcan and Alumax.
Initial tests were undertaken almost in secret until there was clear evidence that sp
was not important. Only when stroke speeds as slow as 8s (versus 0.9s normal stroke)
had been proved successful was senior management informed that slow speed was
possible. Even then, it was only the fact that test had been successful for 3 months that
enabled them to be convinced. It is hard to disagree when the answer is "We've already
done it".
As an extreme, tests were conducted at zero velocity (Table 3-HI). In the tests, 29 po
had the feeder turned off for periods of 2-7 hours. After the stated period, the plunger
was gently lowered to the top of the crust. Then the air was turned on and the time to
break through the crust was timed. Instantaneous breakthrough occurred for 9 0 % of
feeders after 4 hours. Crustbreakers on most potlines operate about every 3 minutes. It
seems likely that there is tremendous over capacity in operating pots.
Note that in this test there was no alumina or anode cover placed over the hole during
the period. This factor is quite significant as it is postulated that these materials are the
TABLE 3 - m
TIME TO BREAK A HOLE AT ZERO VELOCITY

TIME TO
DOWN
BATH POT FEEDER BREAK
DATE POT CaF2 TIME
LEVEL TEMP LOCATION CRUST
(hrs)
(s)
01.06.91 4012 160 927 3 3 0
T.
4014 180 972 3 2 0
4018 170 976 3 3 0
4024 150 940 3 3 1
4032 170 954 3 4 1.5
4033 180 960 3 4 0
02.06.91 4012 150 952 3 7 0
4014 170 955 3 7 1
20-26.06.91 4018 200 962 3 7 0.5
2041 4.9 200 958 1 4 0
2042 4.8 150 966 5 4 0
2043 4.9 160 945 5 4 0
2044 4.7 230 935 5 4 0
4049 5.0 120 985 4 0
2010 4.7 200 993 4 0
3088 5.0 200 1005 4 0
3091 5.0 210 951 4 0
3101 4.9 200 968 4 0
2040 4.7 190 960 4 0
2011 4.9 200 972 4 0
2004 4.9 230 958 4 0
2060 4.8 190 945 4 1
4075 4.9 270 970 5 4 0
4026 4.7 200 994 4 0
4020 4.8 240 969 5 1
2009 4.8 180 961 5 5 0.5
1033 4.9 150 955 5 0
1037 4.9 200 955 5 5 0.5
1034 4.9 200 950 5 0
Notes:
(0 Method is to turn feeder and kidney plate off for stated period with
plunger above crust. Then slowly rest plunger on crust, put air on at
zero velocity and time period to break hole.
(ii) The lower the CaF2, the harder the crust. A 4.7-5.0% C a F 2 is typical
for smelters.
reason w h y plungers mainly do not maintain a hole. It was concluded that it is not the
hard crust that is the problem with blocked feeders, but compaction of the loose anode
cover that falls into the feeder holes. The plunger drives the material into the base of
the pot until the anode cover packs into a solid blockage. This test enforces this
argument as clearly it takes little effort to break the crust per se.
It was concluded therefore, that specification of KE or pressure in themselves are not

solutions to the task of breaking a hole successfully in a repeatable fashion. The stroke
time (plunger speed) of Portland feeders was reduced late in 1991 to a speed that would
largely eliminate fatigue of components viz. a change in stroke time from 0.9s to 1.6s
for downstroke (Section 8.6.2). There was no change in anode effects due to blocked
feeders. This proved that the above analysis was correct (even for a 7 0 % reduction in
kinetic energy). T 1 7 has also slowed d o w n their feeders due to the success of Portland
and no effect has been seen in anode effects or blocked feeders.
There is no commonality between KE or pressure for all the plants surveyed, so these
parameters do not appear to be the main criteria for selecting a feeder design that does
not block. Yet, blocked feeder holes are a c o m m o n problem in all plants. It is
postulated that the causes of blockage are not due primarily to the design of the feeder
but to other independent factors.
The reasons for unsuccessful breakage are more complex, as the reasons are related to
things other than hard crust. A blocked feeder survey was conducted at Portland to
identify w h y blocked feeder holes occurred and h o w to prevent them.
C H A P T E R 3 D O S E DELIVERY P A G E 131
3.5 BLOCKED FEEDER SURVEY

A survey w a s conducted from mid December 1991 to mid January 1992 to try to
establish reasons for blocked feeder holes. S o m e 1058 blockages were logged, of
which for 9 % the cause could not be identified. Table 3-IV shows the results for the
9 1 % of blockages for which the cause could be identified.
F I G U R E 3-IV
CAUSES OF BLOCKED FEEDER HOLES AT

PORTLAND
CAUSE OF BLOCKAGE NUMBER PERCENTAGE
NEW SET CARBON 250 28
FEEDER REPEATERS 207 24
LOW BATH 168 19
POT REPEATERS 106 12
POINTED PLUNGERS 60 7
UNKNOWN 41 5
ORE LEAKS 23 3
OTHERS 21 2
TOTAL 876 100
Notes:
(i) 44 feeder repeaters (2% of feeders) had 2 4 % of blockages.
(ii) 11 pot repeaters (3% of pots) had 1 2 % of blockages.
(iii) Observations were taken in December 1991 and January 1992.
3.5.1 Examination of Blocked Feeder Survey Data

Before the survey was conducted, several operators said it was a waste of time. They
already knew the main reason w a s "new set carbon"; that is, the holes block shortly
after the anodes are set and covered. While this proved to be the case, it was generally
not predicted by most engineering and management personnel across many smelters.
Most engineers thought the main reasons would be pointed plungers (where the
diameter is small) or w h e n bath level is low (when the plunger does not reach the crust).
All nine (9) Alcoa plants have since done similar surveys to Portland and have also
found the major cause was "new set carbon". Pointed plungers and low bath level were
well d o w n the importance list.

This is an important anecdote as it illustrates h o w important it is for engineers to

consult operators. Operators have a feel for problems, even if definitive data does not
exist.
Examination of the blockages shows some feeders blocked frequently; whatever the
reason. It seems reasonable to group these blockages as being due to a faulty feeder.
"Feeder repeaters" were defined as where a feeder had 3 or more blockages during the
survey period. Also, groups of feeders in a pot had frequent blockages. "Pot repeaters"
were defined as where 3 or more feeders had 2 or more blockages during the survey
period.
The prioritising of blockages were (in order with most important first);
(i) feeder repeaters,

(ii) pot repeaters, then
(iii) the stated cause viz. "new set carbon", "low bath", "pointed", "leaks" and
"others".
Any blockage that had none of these was listed as "unknown".
Below is a discussion of the causes and proposed actions for each of the common cause
of blocked feeder holes.
3.5.2 New Set Carbon

W h e n the anodes are set and covered, the plunger is actually below the top of the
carbon and is generally buried by anode cover. At most, there m a y be 1 5 0 m m of space
between the bottom of the plunger and the top of the anode cover. Hence, the plunger
hardly starts moving before it hits the anode cover. So, speed is not the major cause of
blockages after setting.
Anode cover pours onto the anode and flows d o w n into the open hole. If one covers
immediately after setting, the anode cover fills the hole from the bottom of the pot to
above the anode. W h e n the plunger is extended, it compresses the material as far as it
can, but generally there is little movement. It is only when the metal/bath washes the
material away at the bottom does the plunger have any chance of "extruding" the anode
cover through the hole.
In the meantime, every 3 minutes or so, another kilogram of alumina is added to the
hole area as there is nothing to tell the feeder to stop operating. Hence, the hole starts
to bridge up, being assisted by the compression action of the plunger each crustbreak
cycle. In a couple of hours, the pile is up to the chute and there is virtually no plunger
movement.
From mid 1993, Portland changed from covering anodes immediately after set, to
covering 4 hours after set. Blocked feeder holes dropped by 8 5 % after this change.
Covering later lets the hole crust up a little so that the cover material only fills up above
the crust. Hence, the next crustbreak has somewhere to push the cover material into the
pot and the hole has a greater chance to remain open. Note that this technique also
reduces the amount of unmetered alumina dissolving in the bath and reduces the
amount of bath generation from the anode cover.
In the Portland survey, the worst blocked feeder frequency was where operators
covered the anodes with the highest anode cover depth. This is a potential problem if it
is decided to increase the amount of anode cover to minimise airburn on anodes.
T7 found 40% of blocked feeders to be due to newly set carbon in 1992. Now they have
the people w h o set and cover the anodes, ensure that the feeder hole is clear before they
leave the pot. They have more than halved blocked feeders since this was done. This
appears to be the cheapest and most effective solution to this problem. D 2 dig out the
hole after covering. (D2 have also achieved 0.03 anode effects per day which is the
benchmark for anode effect rate.)
Increasing the dwell time after set (say for 2 hours) will allow the crustbreaker to
reduce the angle of the anode cover to its natural angle of repose, then holes will not
block as often. Since mid 1993, this practice has been used at Portland with reduced
anode effects resulting. A similar program could be used whenever a feeder blocks i.e.
increase dwell time for a few hours to ensure the hole is penetrated. This feature is best
installed on plants that have the ability for dwell time to be controlled separately on
each pot. Older plants do not have this ability.
When a blockage occurs, it is wise to stop feeding until the hole is clear. For integra
feeders this is done by shutting off the alumina valve (the kidney plate at Portland). A
number of anode effects are due to people forgetting to open the kidney plate after the
hole is clear. T h e computer will automatically increase frequency of feed to
compensate for insufficient feed, however, anode effects occur until the pot feed rate
matches demand. T 1 7 has seen the situation where the kidney plate was shut for a
month before it was detected! The pot was operating satisfactorily, but the frequency of
feeding was detected as being very biqh.
Whenever anodes are set next to a feeder there is more risk that the new anode cover
will block the hole. Hence, the pattern of setting anodes in the pot and the number of
feeders installed in a pot will have a large bearing on blocked feeders. It is preferable
to have few newly set anodes next to feeders. Every time a n e w anode is set at Portland
it is next to a feeder. In fact, 2 of the 5 feeders have recently set anodes on each side of
the feeder at the same time. This is most unusual in prebake pots, and m a y be the only
pot design that has this feature. S o m e smelters set one anode at a time in almost a
random pattern to avoid this problem e.g. Pechiney. Choice of setting style can assist
the prevention of blocked feeder holes. (In addition to the effects of the setting pattern
used at Portland, Portland also has the highest number of feeders of any pot in the
world. A s such, it has greaterriskof blockages.)
Reducing carbon height or reducing carbon area near the feeder has been used by
several plants to minimise bridging across the hole e.g. T6, T7. This has proved very
successful. The carbon horizontal area around the feeder hole often airburns anyway
and is a loss of carbon (which costs about $250/t). Thus, not having the carbon there in
the first place will save blocked feeders and cost of carbon production. The latter cost
can be significant if the plant is making 180,000 anodes per year as is the case for
Portland.
Thus, there are several methods that can reduce the frequency of blocked feeder holes
(i) longer dwell time

(ii) ensure hole is clear after covering
(iii) stop feeding after blockage
(iv) different setting pattern for n e w carbon
(v) reduce carbon height
(vi) reduce carbon area around the feeder hole.
3.5.3 Repeaters
S o m e 2 % of the feeders surveyed had 2 8 % of blockages and 3 % of pots had an
additional 1 2 % of blockages. S o m e 4 0 % of blockages occurred on less than 3 % of
pots. From the previous discussion, maybe Portland was operating at high speed to
correct for the inadequacies of only 2 % of feeders, and wearing out the other 9 8 % as a
result.
Mechanical factors ("feeder repeaters", "pot repeaters", "pointed" and "leaks")

comprised 4 6 % of blockages. The three basic flaws of repeaters are dwell-time time-
out, lack of supply pressure and back-pressure.
3.5.3:1 Dwell Time Time-out
Both "feeder" and "pot" repeaters can be related to too low a speed causing the dwell
time to run out before pressure is applied. This is not lack of kinetic energy, only lack
of time. Although all pots have the same computer dwell time, the individual pot m a y
be affected by D C V integrity, local low pressure points in the air lines, restricted hoses,
CHAPTERS D O S E DELIVERY P A G E 136
tight rod/piston seals, hose leaks, air leaks in fittings, or alumina in the cylinder. All
these factors affect variability of speed and hence the effective time that the cylinder is
fully extended.
Operating at an inflated dwell-time may achieve full shot size, but results in low
plunger life and "dags" and doesn't address the cause of the problem. Chapter 4
addresses these factors in more detail.
The solution to "feeder repeaters" is to change out feeders that frequently block -
whatever the reason. The repair people can trace the cause and report back to the
initiator via a tracking system for follow-up.
3.5.3:2 Lack of Supply Pressure
It is useful to monitor air leakage rate so that rod and piston seal leakage can be
detected before blockages or ore leaks occur.
The most important air leaks to address in order to stop blocked feeder holes are when
the crustbreaker is in the d o w n position. Air leak checks are often done when the
plunger is up, but this does not detect some piston seal leaks or leaks on some fittings.
Pistons often have seals pointing in opposite directions and it is indeed possible for a
cylinder to leak in one direction only. O n e needs to check air leaks on the downstroke
if one wants to see the effect on crust breakage. Section 9.4 discusses in depth air
leakage detection and the costs associated with leaks.
3.5.3:3 Back Pressure
Another major factor causing lack of pressure is blocking of mufflers on the exhaust of
the D C V . These block due to grease and deposits from the cylinders and air lines. A s
they block, they cause a back pressure on the cylinders, slow them d o w n and reduce the
pressure on the surface of the crust. Portland had a muffler life of 6 months with nearly
all these being badly blocked to the extent that the plunger would not go back up into
position. Hence, the 1 2 5 m m cylinder could not even lift a 35 kg mass! There would be
virtually no pressure available on the d o w n stroke. This problem is more serious at
plants which use a c o m m o n muffler for several cylinders than designs where there is
one feeder per muffler. Til found that blocked mufflers halved the pressure at the
crustbreak cylinder. A t Portland, the back pressure can get so large for blocked
mufflers that the muffler can explode (Figure 3-2(a)).
To achieve no back pressure, one needs to:
(i) direct the exhaust to another place that does not need a muffler, or
(ii) use a "straight-through" muffler which does not block, or
(iii) reduce the air flow rate by having several small outlets.
It is necessary to not only solve the blocking problem, but also to reduce noise level
There were several possible solutions to this problem:
(i) Redirect flow

S o m e 68 pots at Portland had the exhaust directed to the pot fume duct as an
alternative to a muffler. This design is costly (about $300/pot) and although
able to achieve less than 85dBA, the noise level is a nuisance to people working
in the area as the exhaust tends to echo in the building. S o m e plants (T8, TIO)
direct the exhaust into the pot cavity which is a very good idea as there is no
need to muffle the exhaust which m a y affect back pressure on the feeder.
(ii) Straight-through or non-blocking mufflers

At least a dozen different muffler types have been trailed as part of this
investigation ranging from customised units to car mufflers. All have failed due
to noise or blockage. Dr. B o b Randall of Unisearch (University of N e w South
Wales) designed a straight-through silencer similar to ariflesilencer (Figure 3-
2(b)). It achieves under 8 5 d B A and does not block. These were installed on all
CHAPTER 3 D O S E DELIVERY P A G E 138
pots at Portland in mid 1993 and none have failed in over 2 years versus an
average life of 6 months for the original Allied Wotan fibre unit. Costs are
about $200 each.
FIGURE 3 - 2
PORTLAND FEEDER MUFFLERS
(a) Allied Wotan M 1 2 exploded and new
(b) Portland silencer

(iii) Reduced air flow

This is the method used on independent feeders. Each feeder is controlled by its
o w n valve mounted on, or near, the cylinder with its o w n exhaust. The reduced
air flow allows the use of small off-the-shelf silencers with little back-pressure,
but the installation cost adds up for the large number of feeders on a modern
plant.
3.5.4 Low bath

This appears to be an obvious cause of blockages. The solution also appears simple -
don't over-tap bath or metal. Plants such as T 1 7 and D 1 3 do not have the plungers
touch the liquid yet still successfully break holes (refer Section 5.4.2). There is no
reason to get the plunger wet if one can predict the location of the crust. Operating to a
total liquid level is desired, rather than work to a fixed bath level. Total liquid level
control ensures a relatively fixed distance from plunger to liquid and crust. Fixed bath
level alone does not guarantee this relative distance as the metal level m a y change from
pot to pot or on changing the metal depth targets of the pot.
3.5.5 Pointed Plungers

Note that pointed plungers were only responsible for 7 % of blockages in Table 3-IV.
Despite all the activity and discussion on "pointed feeders", this is not a major
contributor to blockages. Operators often suggest that pointed feeders seldom block. A
pointed plunger ( 5 0 m m diameter) concentrates the pressure to 4 times that of a 1 0 0 m m
plunger. A s long as the crust is high enough, a pointed plunger will work.
Portland installed 30 Inconel 60mm plungers in 1991. After a period of 6 months, the
Inconel pipe had corroded away and the plungers were operating on the 5 0 m m steel
vertical plunger shaft quite successfully. Hence, 5 0 m m is sufficient for alumina flow as
long as cover practice is adequate. Clearly, the wider the plunger the better it is to
break a hole and allow alumina entry (Section 3.3(d)).
3.5.6 O r e Leaks
Ore leaks on integrated feeders that block holes are generally caused by very bad air
leaks from the air cylinder that aerates the alumina in the feeder. This floods the pot
with alumina; generally from the kidney plate. Providing better rod seals and/or
allowing a better air passage for venting air leaks are possible solutions (Section 8.4).
3.5.7 Feeder Location

Feeder repeaters were less c o m m o n on #5 feeder at Portland. This m a y be due to less
cover near the tap hole. Pot repeaters had no blockages at #5. This is interesting as # 5
feeder is farthest from the D C V and could have been affected by slower speeds. So it
appears that the cover is more important than the manifold layout (Figure 1-21).
3.5.8 Minimum Dwell Time Determination

Dwell-time is chosen to achieve two aims:
(I) to release all alumina from the feeder (which does not apply for independent
feeders), and
(ii) to break a hole.
Shot size testing has been undertaken on Portland, T 6 and Til feeders over a range of
conditions (Chapter 2). For the feeders tested, shot size was consistent if the dwell-time
was greater than the movement time for full stroke.
Tests were conducted on the pressure profile of the inlet and outlet air lines to the
feeders at Portland (Chapter 4). They showed that full pressure was not achieved until
well after the plunger reached the end of the stroke. These factors need to be
understood for each plant to optimise dwell-time. If one chooses too short a dwell time,
there is insufficient time to reach m a x i m u m pressure to break the crust and all the shot
m a y not be delivered. O n e gets no feed at all or not enough. Anode effects will result
for both situations.
Independent feeders can reduce wet time to about 0.2s because they are not affected by
the time to discharge the alumina. This compares with 0.5-1.5s for most integrated
feeder potlines, and up to 4s in some potlines.
3.6 RESULTS OF IMPROVEMENTS TO BLOCKAGE PREVENTION

From mid 1992, considerable action was taken at Portland to address blocked feeder
holes based on the findings of this survey. Covering was changed from immediately
after set to the shift after set to reduce the amount of cover in the holes. Dwell time was
increased to 5s for 2 hours after covering to accommodate slump of anode cover in the
holes. These changes resulted in reduction of blocked holes after set by about 8 0 % .
Air leak surveys identified hardware problems which could have contributed to lack of
pressure and variability in shotsize accuracy - leaking feeder valves, hoses, fittings and
by-passing pot valves being c o m m o n . B y 1993, air leak rate per pot was under
301/minute at Portland which is the lowest in the Alcoa system and possibly a
benchmark level.
By November 1991, all feeders at Portland had their KE reduced by 70%. There was
no change in anode effects due to blockages. This verified that high K E is not
necessary to break a hole and suggested that the large cylinders used by some
independent feeder plants (e.g. Pechiney) are unnecessary. Smaller cylinders would be
satisfactory, with significant savings in capital and air usage. Although 1 2 5 m m O.D.
cylinders are adequate, they m a y be a little small to guarantee a hole under all plant
conditions. Instead of the 2 0 0 m m O.D. size used by Pechiney, a 1 5 0 m m size is
recommended for n e w plants. It achieves its task and is considerably cheaper.
The reduction in blocked feeders was a major contributor to halving anode effects at
Portland from mid 1991 to mid 1993.
Although initially it was thought that holes were broken only by high kinetic energy
is n o w widely accepted that this is not the case. Thus, this research has completely
changed a theory held by m a n y smelters for over 30 years.
3.7 MAJOR FINDINGS FROM DOSE DELIVERY INVESTIGATION

(i) Holes generally block from anode cover bridging the hole.
(ii) Bridging is aggravated by increased granule size, cover height, hole size and
vibration of anode cover.
(iii) Blockage frequency does not relate directly to the kinetic energy or pressure
from the feeder crustbreaker but to a range of factors related to pot operation
(such as anode cover practices).
(iv) A t Portland, newly set carbon was the major factor causing blocked feeders
(28%), followed by feeder repeaters (24%), low bath (19%), pot repeaters
(12%), and pointed plungers (7%). Similar trends were seen at other plants.
Repeaters were due to expired dwell time before full pressure is achieved, lack
of supply pressure, and/or back pressure.
(v) Mufflers blocking cause back pressure on feeders, reduce effective pressure at
the crust, and D C V failures.
(vi) A 1 2 5 m m O.D. cylinder is adequate for crustbreak cylinders, but 1 5 0 m m O.D.
is the recommended size for n e w plants.
3.8 RECOMMENDATIONS FROM DOSE DELIVERY INVESTIGATION

(i) Dig out the hole after covering n e w set carbon or after dressing the pot.
(ii) Provide a long dwell time after set and after blockages.
(iii) Consider alternative setting patterns to minimise the number of n e w anodes next
to a feeder.
(iv) L o g all blockages with reason for occurrence to trace patterns of failure.
(v) R e m o v e feeder repeaters if no other cause is found.
(vi) Regularly monitor air leakage rate for each pot to identify feeders with piston
leaks on the downstroke.
(vii) Install silencers that do not block or redirect exhaust into pot exhaust duct.
(viii) Control total liquid level (rather than bath level) to ensure reasonable
penetration depth through the crust (not necessarily to get the plunger wet).
(ix) Install 1 5 0 m m O.D. cylinders on n e w crustbreaker units.
CHAPTER 4 PNEUMATICS PAGE 143
CHAPTER 4
PNEUMATICS
4.1 SUMMARY
Testing was carried out on the pneumatics of a Portland pot for a range of feeder age,
speeds, location of feeder in the superstructure and type of muffler. In all, 144
combinations were tested to see the response time before the feeder moved, stroke time
to reach full supply pressure, and time to reach m a x i m u m pressure (both for
downstroke and upstroke). Calculations were then carried out on the time the plunger is
potentially in the bath ("wet") to understand better the causes of erratic plunger wear.
It was found that new feeders operated slower than old feeders. Paper type mufflers
affected the time it takes to reach full pressure but had little effect at reduced speeds.
For a multi-feeders/DCV pot, the feeder closest to the pot DCV initially moved fastest
but then the others caught up by the time the feeders reached the crust (except that a
n e w feeder always was the slowest). Variable wet time was traced to an interaction
between mufflers and feeder age. Waiting time on the up stroke was the major factor
on wet time for the plunger. The transient stalling of the downward movement just after
the feeder starts to m o v e indicated insufficient air supply and identified a major
contributor to shot size variability for Portland.
Having a DCV for each feeder significantly reduces reaction time of feeders so that we
time can be reduced by suitable selection of crustbreak dwell time. Location of the
D C V s close to the feeder(s) reduces air consumption significantly.
4.2 INTRODUCTION
Shot size testing showed that the speed of the feeder affected the size and variability of
the shot (Chapter 2). The dose delivery investigation showed that 4 0 % of blockages are
related to repeating blockages on the same feeders or pots (Chapter 3). A s will be seen
in Chapter 5, plunger wear is related to the time that plungers are in the liquid. Clearly,
feeder pneumatics are a major contributor to shot size accuracy and feeder failure.
Air usage for feeders is usually the highest single consumer of air in a smelter. At
Portland, the feeders account for 3 0 % of the plant air usage which is not unusual for
smelters. O n e can see that attention to air usage is a major factor in the efficiency of
not only the pot feeders, but also the plant as a whole.
To gain a better understanding of the dynamics of the feeding system, an investigation

was held into the pressure trace of inlet and exhaust air to feeders at Portland (Figure 4-
1). Factors considered were:
(i) different feeders (runs of 2 new and 2 old Terry feeders),

(ii) location of the feeder in the pot (tap and duct end),
(iii) paper muffler design and degree of blockage (Allied Wotan M l 2 muffler
and Portland Aluminium silencers, Figure 3-2),
(iv) speed of feeders, and
(v) multi-feeders/DCV versus single-feeder/DCV.
Two pressure transducers were used to measure pressures in the air line of off-line po
at several locations. M o v e m e n t of the piston rod was sensed by a special test unit
mounted on the top of the cylinder (Figure 4-2). A piston was pressurised by a 420kPa
chamber that forced a small piston to follow the large crustbreaking cylinder piston.
The test unit had a piston rod travelling through the top flange of the crust breaking
cylinder and pressing against the cushion spear of the piston. Limit switches were
mounted to sense end of stroke by use of an extension of the piston rod above the test
unit.
FIGURE 4 -1
POT PRESSURE SENSING POINTS
PRESSURE TAKE-
OFF POINTS
WIDE O R
TAP AISLE
RESTRICTOR.
(12mm)
1Z^
L A
NOTES:
(1) In closed position, ball valve ^ly' BALL
VALVE
has a 12 m m hole. NARROW
(2) Line A supplies air to force AISLE
plunger up.
(3) Line B supplies air to force
plunger down.
RESTRICTORS
(12mm)
GATE VALVE
IR SERVICE UNIT
-BALL VALVE (BLEEDS
TO A T M O S P H E R E )
-FILTER
-LUBRICATOR
FIGURE 4 - 2
STROKE TESTING UNIT
< = 0 TOP LIMIT SWITCH
BOTTOM LIMIT SWITCH

EXTENDED
PISTON
ROD
CANISTER
TRACKING ROD
FEEDER
CYLINDER
CRUST BREAKER
PISTON
The piston diameter of the test unit was 3 2 m m . This relates to a 4 % change in force on
the crustbreaking piston. It was not considered that the unit significantly affected the
performance of the crustbreaking cylinder and, as such, could be used to sense the
tatter's performance.
The pressure transducers were wired to a high resolution chart recorder and traces were
obtained of the response; Figure 4-3 shows typical curves. A n off-line pot was selected
(#1015) for most of the final testing after experimentation on several pots showed that
the location of the pot in the rooms was not significant i.e. a feeder in one pot had a
similar profile to that in another pot. T w o old and two newly overhauled feeders were
tested in location 1 and 5.
Supply pressure was 800+/-20kPa at 140kPa per division on the vertical axis of the
pressure traces. Unfortunately, the figures do not always show each division clearly
due to the quality of the copying of the instrument graphs.
4.3 GENERAL OBSERVATIONS ON PRESSURE TRACES
Looking at the traces in Figure 4-3, one will see locations noted as location A-L. The
cycle occurs as follows:
A. Start of cycle: air was applied to top of cylinder.
B. Start of cycle: air at bottom of cylinder was opened to vent line of the D C V .
A-C. "Waiting time".
Pressure rose in the supply line to a stage that it was larger than the exhaust air
(B-C) until point C is reached. This usually took 0.8-0.9s for restrictors
installed (in) or 0.4s if restrictors are not installed (out).
C. Pressure across the piston exceeded the friction (called "sticksion") of the seals
and bush holding the piston stationary. Sticksion was about 40kpa. The piston
starts to move. There was a rapid drop in curve A C due tofillingthe cylinder
up with air as the piston starts to rapidly free fall. Then the pressure difference
between the two traces dropped to be too small to overcome the friction and the
OS
u
Q rtt'--
w
!_.-.;..:« o
ti
Z
CH
CH -j-
>
—/%-";!
o
o
CH
u
H
ti
CH LU
CH
ti
tf
0- J.»tl-
r»_-
C3- —-:i.
5 co
UJ Ul
NOISIAia/«d>(0»l
NOISIAia/w'MOtrl
NOISlAt07!*dO?
NOISIAia/nOQZ
piston slows. Hence a bouncing effect occurred for a cycle or two until the
piston starts to m o v e smoothly. Note that this double movement m a y be seen in
the movement of the plunger on slower moving feeders in an operating pot.
This "stalling" of the feeders occurred on all tests (even at the original design
speed) and suggested that the air supply was not sufficient enough to maintain
full pressure at the piston. A larger air manifold across the top of the pot is
needed, or the D C V needs to be positioned closer to the feeder. It was less
noticeable for n e w feeders, slow speed (Figure 4-4), and off-site (Figure 4-5).
Also at C, one can see the step change in the second line from the top of the
graph. This is the trace of the limit switch for top end sensing.
C-D. The time that the piston moved the length (stroke) of the cylinder. The pressure
trace was almost horizontal.
A-D. "Hit crust" time i.e. the time from start of cycle to end of movement.
E. Piston reached full travel (as evidenced by the third top trace from the limit
switch). The pressure increased at this point.
The rate of build up depends on back pressure and type of feeder. Sometimes D
& E occur together (Figure 4-3, 4-4). This is because this feeder was the
slowest in the pot.
E-F. Pressure rapidly rose and fell at point E when the throttling effect of the muffler
and the orifices does not apply. At F the m a x i m u m supply pressure was
achieved.
A-F. Time to reach " m a x i m u m supply" air pressure.
G. " M a x i m u m force" at the crust. Supply air was at m a x i m u m and exhaust air was
at zero.
F-G. The time between " m a x i m u m supply" and "maximum force" on the crust took
0.4-0.6s in most traces.
A-G. Time to reach " m a x i m u m force" since the energising of the D C V .

s
-*- .' —". .,
o
-I
(0
•11 1
-LJ
LifJ—~rl
CH «0
ti coiai» z
O triifc-'il »* r t i n »«
Tl Tf' T r 0
o
or
:::±_.-So J_JS»C0
^
|.D o
u • 4-IL • •
,, . !«v
•jr.
M.
Tim
- — * (fl
Q te~; 1"I*1°IS
ti "IJLI:
ti
OM
j> 2
CH JL B
i
- _ L. — 1 — _ ,
CH
ti
- * •
" S ± " "~S
U
** Ht"a ::-:tx
J l .-±S
- 11
JL -rH,
'>. t
_J. J_U T IT ti
ti
tf d NL "
ir "i/l^C-- ?
* ~" y i
.P
v' J
CH
CH
at J-
.:a.Lwi-ur ' f*" -J
ri~
3. _2
55^
ti ~T" . .!
tf 1
• -H-HI ™J~ I•
j__r""it* 3
: ± : "'I..
J- i
X I J-.J
zi-it •" Of
f*H- UJ •*••-i i i - ._ s or UJ
•g
:szx^
:_:±i LU -I
\ Y "9 m ••••» UJ =
±Kfcj±tfi J--3*-:*L I LL S
rf-i. §§
• •4-4 • - -
sizfe*; LU LU
IS .dh
NOISIAia/w*10»l NOISIAia/™»10t>l
NOISIAtd/pdOZ NOlSIAia/!*dQZ
;AT MJ
H
CH
ti
H
ti jq:îl
H J
CH •IT""iIS j
i
ti
- •
5-
ti • •
1
< «8
o i
i
ti H/ •z
H ' 1
i ;
12
O NH
CH
i
i
!
\*\
tn
ts. i
i i
•>-
!1 j i
; ..... SUJ
z r i ,* .
•atn
i_u.
i j •-
o r "^LIZ - L
"kr
CH
ti
_ .
• >
•F>
J-l/ i •'
u j. !\L__/a.j
;-«•/
: i
- i
H
CH
LU- f i
\
ti i.i •H ,•
;
CH ! -'
ti .U.Ji o
•!• L i-
tf • i
ti
- lr
J.J
ILi.4—:
II.
or
I-
co
LU
h-
LU
0.
LU
* ' T
•.
:
I
• N:
:! •*.'
1 \r
I
s
UJ
UI
h-
CBT\
••-
;Â—- .—
NOlSIAia/^>IOfl 05
NOISIAia/!"iOZ u.
LL.
NOis.Aiayrtwt'i
NOISIAia/lsdQZ
o
D-G. Time that the plunger was potentially in the bath for the downstroke; the "down
wet" time.
G-H. This time changed from curve to curve depending on when the valve was
reversed. Care was taken that the traces came to steady state before the feeder
was activated to go upwards.
H. Start of upstroke.
I. Pressure traces crossed over with a bounce in the supply pressure.
This was not so pronounced as for point C due to the retarding effect of gravity
on the feeder mass.
J. Piston started to move. The up cycle was similar to the down cycle except there
was less "stalling" when the feeder started to move. The piston did not start to
m o v e until possibly half w a y along the I-K stretch. The pressure required to
start the piston moving was about three times that needed for the down stroke
i.e. 140 kPa.
J-K. Upstroke movement time.
L. M a x i m u m supply pressure upwards.
M. M a x i m u m force upwards.
The "total wet time" is calculated as the end of downward movement to the start of
upward movement i.e. "down wet" plus "wait time" on the up stroke. This is the total
time that the plunger is stationary and it is potentially in the bath (wet).
Note from Table 4-1 and 4-II that the d o w n wet time was 40-70% of thetotalwet time.
People generally think of dwell time as the main factor affecting the time that the
plunger is wet, but seldom consider the response time of the plunger when it is
retreating from the fully extended position. O n e needs a quick response time of the
piston on the upstroke in addition to the conventional interest in dwell time.
In addition to the times listed, the plunger wet time will be affected by the penetration
in the bath. If there is no penetration, "wet time" is a misnomer and the plunger is not
affected. Most plants have penetration, and the degree of penetration will affect the
time in the liquid in addition to the times listed here where times are specified for
change of direction only. Refer to Section 5.4.2 on discussion on plunger penetration.
Testing was done using Terry cylinders so that the type of cylinder was not another
variable. However, as the Parker cylinders n o w have Terry rod bushes and rod seals, it
is expected that a similar performance would apply. There was little to no difference in
the performance of any similar pair of feeders. The values of times in Tables 4-1 and 4-
II are the average of a pair of similar feeders.
Note that at X and Y , the gap suddenly became larger on several pressure traces. This
is believed to be due to other cylinder(s) in the group offivereaching the end of travel
before the rest. Once the fastest cylinder piston reached the end of travel, the rise in air
pressure in the manifold caused the remaining cylinders to develop a greater pressure.
This phenomenon illustrates that the cylinders were being partly starved of air; not
"pressure limiting" but "flow limiting" which also shows that one cylinder can affect
another. This is a problem for pots using multifeeders per D C V .
4.4 METHOD USED TO SLOW FEEDERS
W h e n the decision was m a d e to slow the feeders at Portland in 1991 to reduce fatigue.
Three alternative designs were tested and rejected.
(i) Flow control valves - block with time due to annular seat and small aperture
diameter.
- poor experience in Alcoa plants.
- anybody can adjust them and no one knows when or
h o w much. Uncontrollable.
- expensive as needed on all 2,200 feeders ($300,000)
(ii) Pressure regulators - anybody can adjust. Not controllable.
- iffixadjustment, cannot easily over ride all pots on
major plant upset e.g. power off for extended period.
- expensive as need two per pot ($150,000)
(iii) Reduce air line size - could not go back if unsuccessful.
- expensive ($200,000)
TABLE 4 -1
EFFECT OF MUFFLER & SPEED CHANGE - LOCATION 1

TRACE N E W FEEDER OLD FEEDER
DIRECTION SPEED TIME SECTOR PA NEW OLD PA NEW OLD
DOWN ORIGINAL WAIT AC 0.4 0.4 0.4 0.4 0.4 0.4
MOVE CD 0.9 0.8 0.9 0.6 0.6 0.8
HIT CRUST AD 1.2 1.2 1.3 1.0 1.0 1.2
MAX SUPPLY AF 2.0 2.0 2.1 2.0 2.0 2.1
MAX FORCE AG 2.4 2.4 2.8 2.4 2.6 3.4
DOWN WET (2-AD) 0.8 0.9 0.8 1.0 1.0 0.8
UP ORIGINAL WAIT HJ 0.5 0.6 0.6 0.5 0.5 0.6

MOVE JK 1.0 1.1 1.0 0.9 1.0 1.4
HIT TOP HK 1.5 1.6 1.7 1.4 1.5 2.0
MAX SUPPLY HL 2.2 2.3 2.4 2.2 2.4 2.5
MAX FORCE HM 2.2 2.3 2.9 2.2 2.4 3.5
TOTAL WET TIM LOC1 (2-AD)+HJ 1.3 1.4 1.4 1.5 1.5 1.4
DOWN SLOW WAIT AC 0.8 0.9 0.9 0.9 0.8 0.9

MOVE CD 1.9 1.8 1.9 1.1 1.1 1.2
HIT CRUST AD 2.7 2.8 2.8 1.9 2.0 2.0
MAX SUPPLY AF 4.7 4.6 4.7 4.7 4.9 4.9
MAX FORCE AG 5.5 5.4 5.5 5.4 5.2 5.6
(4-AD) 1.4 1.2 1.3 2.1 2.1 2.0
UP SLOW WAIT HJ 1.3 1.2 1.3 1.5 1.6 1.6

MOVE JK 1.5 1.6 1.6 2.2 2.1 2.1
HIT TOP HK 4.0 4.0 4.1 3.7 3.7 3.6
MAX SUPPLY HL 5.1 5.0 5.2 5.1 5.1 5.1
MAX FORCE HM 6.0 5.7 6.1 5.8 5.6 5.9
TOTAL WET TIM LOC1 (4-AD)+HJ 2.6 2.4 2.5 3.6 3.6 3.6
Note: (i) All figures in cells are seconds (s).

(ii) Times are the average of 2 old and 2 n e w feeders respectively.
(iii) "New" and "old" mufflers referton e w and blocked Allied W o t a n mufflers respectively.
(iv) "PA" muffler refers to Portland Aluminium straight through silencer.
(v)" T O T A L W E T TIME" is the time the plunger is fully extended. During this time there is
the greatest risk of plunger corrosion. It is calculated as follows:
T O T A L W E T TIME = dwell time - hit crust time + wait time
(vi) "Trace sector" is the time between points in the pressure traces Figures 4-3 and 4-4.
TABLE 4 - II
EFFECT OF MUFFLER & SPEED CHANGE - LOCATION 5

TRACE N E W FEEDER OLD FEEDER
DIRECTION SPEED TIME SECTOR PA NEW OLD PA NEW OLD
DOWN ORIGINAL WAIT AC 0.4 0.4 0.4 0.4 0.4 0.4
MOVE CD 0.8 1.0 1.0 0.9 0.9 1.0
HIT CRUST AD 1.2 1.4 1.4 1.5 1.3 1.4
MAX SUPPLY AF 2.0 1.9 2.0 2.0 2.0 2.1
MAX FORCE AG 2.4 2.4 3.2 2.4 2.5 3.2
DOWN WET (2-AD) 0.8 0.7 0.6 0.6 0.8 0.7
UP ORIGINAL WAIT HJ 0.6 0.6 0.9 0.5 0.5 0.6

MOVE JK 1.1 1.2 1.2 0.9 1.0 1.3
HIT TOP HK 1.8 1.7 2.1 1.4 1.5 1.9
MAX SUPPLY HL 2.3 2.3 2.5 2.2 2.4 2.6
MAX FORCE HM 2.3 2.3 3.3 2.2 2.4 3.3
TOTAL WET LOC5 (2-AD)+HJ 1.4 1.2 1.5 1.0 1.3 1.3
DOWN SLOW WAIT AC 0.9 0.9 0.9 0.8 0.8 0.8

MOVE CD 2.0 2.0 2.0 1.2 1.2 1.3
HIT CRUST AD 2.9 2.9 2.9 2.0 2.0 2.1
MAX SUPPLY AF 4.7 4.4 4.6 5.0 4.9 5.0
MAX FORCE AG 5.3 5.9 5.4 5.5 5.4 5.6
(4-AD) 1.1 1.2 1.1 2.0 2.0 1.9
UP SLOW WAIT HJ 2.6 2.6 2.6 1.1 1.2 1.2

MOVE JK 1.6 1.5 1.6 2.3 2.4 2.5
HIT TOP HK 4.2 4.1 4.1 3.4 3.6 3.7
MAX SUPPLY HL 5.0 5.1 4.9 5.0 3.1 5.1
MAX FORCE HM 5.7 5.8 6.0 5.6 5.8 5.5
TOTAL WET LOC5 (4-AD)+HJ 3.7 3.8 3.7 3.1 3.2 3.1
Note: (i) All figures in cells are seconds (s).

(ii) Times are the average of 2 old and 2 n e w feeders respectively.
(iii) "New" and "old" mufflers refer to n e w and blocked Allied W o t a n mufflers respectively.
(iv) "PA" muffler refers to Portland Aluminium straight through silencer.
(v)" T O T A L W E T TIME" is the time the plunger is fully extended. During this time there is
the greatest risk of plunger corrosion. It is calculated as follows:
dwell time - hit crust time + wait time
(vi) "Trace sector" is the time between points in the pressure traces Figures 4-3 and 4-4.
It was decided to use orifice plates in the air lines which were also considerably cheaper
than the other three alternatives ...about $20,000. Installation on the exhaust side of the
D C V was unsuccessful due to back pressure causing the valves to stall and by-pass (or
stop working completely). Orifices ( 1 2 m m ) were placed in the airlines to increase the
movement time from approximately 0.9s to 1.6s.
Initially a restrictor was placed in the supply line at the pot (line B in Figure 4-1
the exhaust line for the downstroke (line A in Figure 4-1), the restrictor was produced
by drilling a 1 2 m m orifice in a closed ball valve. In this way, if unsuccessful, the valve
could be opened and there would be no back pressure and the feeders would operate
close to the original design of fast speed. It was found that there was an annoying noise
from the orifices, so another restrictor was placed in both pipes near the D C V . This
solved the noise problem.
As the cause of anode effects is often difficult to determine, operators go to any le

to stop them. It was found that over a period of time some of the ball valves were
opened in a (vain) attempt to get better pressure at the crust. In October 1993, a
comparison was m a d e of the number of anode effects for pots with valves open to those
with valves closed. This showed no difference in anode effects. In November, handles
of the ball valves were removed to avoid the temptation of operators opening the valves
and causing the feeders to fatigue.
The problem with this cheap design is that it meters air into the cylinder as well as
It is best to only meter out so the piston speed was smooth and has repeatable stroke
times. Pressure sensing showed full pressure is attained, but there was an extension to
the exhaust time.
4.5 RESULTS
S o m e 144 combinations were tested. It would be too cumbersome to explain every one,
so only the critical ones are discussed here; a more detailed analysis is in Kissane (76).
4.5.1 Location
The feeder at location 1 generally moved faster than location 5 (Table 4-1 and 4-II).
Time to hit the crust was about the same and so was the wet time. This cross checks
with the observation in the plunger wear investigation that there is no trend of higher
wear at any location in the pot (refer Section 5.3.7).
Some 60 pots in both lines were checked to see which plungers move the quickest from
a rest location. Location 1 was the fastest 8 5 % of the time and location 5 was the
slowest 7 0 % of the time. However, there was no trend as to which plunger hit the crust
first. The fastest was #1 for 1 9 % and #5 for 1 3 % . Thus, the pattern for movement was
not the same as hit the crust. A s they hit the crust about the same time, then wet time
would be constant and hence plunger wear (from dissolution in bath or friction from
hitting the crust).
On average, the slowest feeder was 0.5s behind the fastest. Feeder-to-feeder variati
speed was more important than location in the pot. However, as location 5 starts to
m o v e the slowest, it will be affected more if the cylinder piston and rod seals have tight
clearances or there is drag from assembly component friction. For plants that have
bushes to keep the dosing unit tracking correctly, be aware that this m a y have an effect
on stroke speed.
4.5.2 Feeder Age
At location 1, older feeders were faster than n e w feeders by 0.2-0.3s at the original
speed and 0.7s at slow speed (Table 4-1 and 4-II, Figure 4-3 and 4-4). At location 5
there was still a 0.7s spread in stroke time for slow speed, but no difference in speed for
the original pneumatic design. Hence, the action of slowing the feeders down caused a
greater variability in speeds of feeders across the pot. Both the old and n e w feeders
attained full pressure at the same time; only the travel speed into the bath was different
at slow speed.
4.5.3 Mufflers
Feeders using old mufflers took 0.4-0.8s longer to get to m a x i m u m force on the
downstroke for the original speed (Table 4-1 and 4-II). They have only a 0.2s effect at
the slower speed. Therefore, slower speeds are not affected as m u c h by partially
blocked mufflers in time to reach m a x i m u m pressure.
On the upstroke, the time to reach the top was very similar for old and new mufflers,
but location 5 was slower by 0.2s for the original speed. Generally, the old mufflers
affected the starting speed on the up stroke as there was a greater waiting time before
they received sufficient pressure to accelerate, however, they still reached the top in
comparable time.
It has been found at Portland that a blocked muffler will cause so much back pressure
that the piston will not rise at all. This was verified in 1994 by Tl 1 when it was found
that blocking of mufflers can increase the back pressure from lOOkPa to 350-400kPa.
Thus, pressure at the crust could halve if a muffler is blocked by a reasonable amount.
This m a y not be obvious to an operator assessing the cause of repeated blockages as the
plunger goes up and d o w n "normally". Installation of pressure gauges at each muffler
is indeed a solution, but the better solution is to install silencers similar to that designed
for Portland.
Mufflers may block because of cylinder lubricant exhausting when exposed to excessive
temperature. Selection of high temperature lubricant minimises this effect (Section
7.4).
For both integrated and independent feeders, the effect of partly blocked mufflers on
feed cylinder exhausts is to change shot size, however, the effect is less on the small
cylinders used for independent feeders (due to the smaller volume of air vented) and
feeders with slow speeds. Note that blocked mufflers also contribute greatly to air
consumption. A n y blockage on a D C V generally results in the internal spool or poppet

not seating correctly and letting air bypass internally through the D C V . At Portland,
this was found to be the cause of the largest air leaks on pot feed air systems. This can
be detected by a continual hissing noise at the D C V or muffler.
Plants can minimise muffler effects by any or all of the following:
(i) routine muffler changeout or clean,
(ii) using a straight through silencer (such as the Portland design) (Figure 3-
2(b)),
(iii) re-routing D C V exhaust into fume duct or pot cavity, and/or
(iv) using high temperature lubricant in cylinders.
4.5.4 Speed Change
The installation of restrictors increased stroke time by about Is on new feeders bu

0.4s for old feeders (Table 4-1 & II, Figure 4-4). This difference m a y be due to minor
seal leaks that m a k e the piston less sensitive to backpressure.
The time to reach maximum force increased by 3 s for location 1 and 2s for location
which was m u c h larger than the change in the movement time. This was found to be
the case in the potline w h e n the feeders were slowed d o w n in November 1991. Dwell
time had to be increased by 2s due to increased frequency of unscheduled anode effects
from blocked feeder holes after the speed change. Hence, for any crustbreaker speed
change, the dwell time must cater for the change in exhaust as well as supply pressures.
The wet time changed by 2.2s for old feeders and 2.5s for new feeders; a rise of 14
and 1 8 9 % respectively. Hence, restrictors have a very large effect on wet time. Plunger
wear was an unfortunate side effect of the speed change. The disadvantage of increased
plunger wear would be compensated by dual dwell time or material change (Section
5.6).
From original conditions, the Portland speed change dropped kinetic energy by 7 5 % for
downstroke and 4 8 % for upstroke for n e w feeders, and 6 0 % downstroke and 8 0 %
upstroke for old feeders. This benefit for upstroke is the main aim of Portland's speed
change as this is when the highest deceleration rates are experienced. Following the
speed change, fatigue failure of piston rods reduced by 6 0 % (Figure 4-6).
FIGURE 4 - 6
PISTON ROD FAILURE HISTORY
MECHANICAL FAULT HISTORY REPORT . C-CONTROL CHART
Ftradfc>ra Installed between 1 January 1939 and 30 September 1 995
fauHType: 212 <PISTON BOO BROKEN)
1<T LlmH&
For n=1i
40
SLOW
3D
UCL=24.D
9
C=t€.6
1D LCL=B.O
0 H •p.. ,. .., , ,.
T T ,r ( ., r.TT T -[—r i i i i I i | i i i i i i i ' i ' i n t t T i I'IIII i r i | i i i i i r-i-fr t i r t i i | i i i i i i i j i I i i r 1 r •[ •
JAMB9 SEPBB MAY30 JANS1 5EPS1 MAYS2 JAN93 5EP93 MAYS4. JANa5 5EPS5
Dote In aha Ned
Subgroup Sizes: n=1
Total number at F&flders In stalled : 1+772 Date Created: 2 October f 995
Total number af fauHs: 40429 Numb«r matching aelaction: 1155 ( 35)
4.5.5 Dual Dwell Time
One way of reducing the impact of the longer wet time for slow speed is the use of dual
dwell time. A dwell time of 2s is long enough for the alumina to be fully discharged (as
shown in the shot size testing experiments in Section 2) but not long enough to hit the
crust. For a "hit crust" time of 2.0-2.9s, a 2s dwell time will mean that the plunger will
not reach (or barely reach) the crust. The subsequent dwell time (say 4s) every 2-3
cycles ensures that m a x i m u m pressure is attained to guarantee a hole. The net effect on
wet time is shown in Table 4-IJT.
One can see that, even with the slow speed, one can reduce the wet time to the same
period as before the speed was changed if a long shot every 3rd time is used. Dual
dwell time wet time is 5 0 - 7 0 % less than that of a conventional single dwell time system
for slow piston speeds. For fast speeds, the effect is 2 0 - 7 0 % but it is more difficult due
to accuracy of timers for small time intervals. Hence, dual dwell time is more effective
for slow speed crustbreakers. It is a very cost effective idea, as the only cost is a
computer program change and somefinetuning to determine the best dwell times.
Note that dual dwell time is also useful for independent feeders as these usually o
at slow speeds. The dwell time is already less than for integrated feeders (because they
do not havetoallow for alumina discharge) but the principles are the same.
There has been no change in anode effects since dual dwell time was introduced in
Portland late in 1992. A s will be seen in Figure 5-9, there was also a drop in the wear
rate of plungers after the change.
4.5.6 Pressure Required to Break a Hole
The pressure required to break a hole can be estimated from the traces by comparing
pressure difference between supply and exhaust at typical plant dwell times. For slow
speeds, a pressure of 630kPa for n e w feeders and 560kPa for old feeders was seen for
slow speeds (Figure 4-3). This is effectively 880-890kPa at the crust for a 1 0 0 m m O.D.
plunger. This estimate is probably high as the 4s dwell time used in this calculation is a
conservative estimate. However, it does illustrate the problems experienced in
blocked feeders in plants using 700kPa plant air supply pressure.
TABLE 4 - HI
DUAL DWELL WET TIMES.

FEEDS NEW FEEDER OLD FEEDER
PER MUFFLER DESIGN MUFFLER DESIGN
CONDITION CRUSTBREAK LOCATION PA NEW OLD PA NEW OLD
TIME TO 1 1 2.7 2.8 2.7 1.9 2.0 2.0

HIT CRUST 5 2.9 2.9 2.9 2.0 2.0 2.1
@ SLOW SPEED
DOWN WET 1 1 1.4 1.2 1.3 2.1 2.1 2.0

@4.0s DWELL 5 1.1 1.2 1.1 2.0 2.0 1.9
DOWN WET 1 1 0.0 0.0 0.0 0.1 0.0 0.0

@2.0s DWELL 5 0.0 0.0 0.0 0.0 0.0 0.0
AVE WET TIME 2 1 0.7 0.6 0.6 1.1 1.0 1.0
@2/4 DWELL 5 0.6 0.6 0.6 1.0 1.0 1.0
3 1 0.5 0.4 0.4 0.8 0.7 0.7

5 0.4 0.4 0.4 0.7 0.7 0.6
DUAL/SINGLE 2 1 50 50 50 50 50 50
%DROP 5 50 50 50 50 50 50
3 1 70 70 70 70 70 70
5 70 70 70 70 70 70
ORIGINAL SPEED 1 1.3 1.4 1.4 1.5 1.5 1.4
WET TIME 5 1.4 1.2 1.5 1.0 1.3 1.3
DUAL/ORIGINAL 2 1 50 60 50 30 30 30
% DROP 5 60 50 60 - 20 20
3 1 70 70 70 50 50 50
5 70 70 80 30 50 50
Notes:
(i) Figures in cells are seconds (s).

(ii) D o w n wet time = (dwell)-(hit crust time).
(iv) Average wet time at 3 cycles/tong shot = [(wet time at 2.0s)+{wet time at 2.0s)+(wet time at 4.0s)]/3
(v) Data is average of two feeders.
(vi) Assume dwell times of 2.0s (short) and 4.0s (long).
(vii)"% drop" is the percentage reduction in wet time with dual dwell time.
4.6 Am SUPPLY OPTIONS
Comparison of performance for off-site and on-site tests on the same feeder and stroke
times shows m u c h lower wait time for the off-site situation (Figure 4-5). This is due to
the fact that there is only one feeder per D C V off-site, versus multi-feeders per D C V
on-site. Response time dropped by 0.2 - 0.6s and time to m a x i m u m pressure dropped
by up to 3s. This illustrates the advantage of the better control for one feeder per valve
design.
Figure 4-7 shows the response time for one and five feeders operating at one time in a
pot for 1.6s stroke time. This shows a faster response time and faster time to reach
m a x i m u m downstroke pressure for one-feeder/DCV versus 5-feeders/DCV. The results
were similar to comparisons of on-site and off-site tests (Figure 4-5). The difference in
the one-feeder/DCV off site (Figure 4-5(b)) and on-site (Figure 4-7(b)) illustrates the
effect of proximity of the valve to the feeder. The off-site testrighad the D C V only
about 2 m from the feeder versus 1 0 m on-site. The closer the D C V is to the feeder the
quicker the response and the less variable are the stroke time and m a x i m u m pressure
time.
The dwell time chosen for a pot line is generally that of the slowest feeder as dwell
is mostly chosen by people reacting to blocked feeders across the potline. It is usually
extended to satisfy the worst pot rather than select a time based on the majority
requirement and addressing the problems with the worst pots. Thus, plants generally
penalise the good feeders to satisfy the bad ones. This also affects the life and degree of
dagging that will occur for feeders that stroke fast. The net effects are plunger wear,
dags, cylinder seal failure, and assembly failure.
For a multi-feeder pot, what is required is to have all the plungers go down together
with little variability across the pot so that the dwell time can be reduced. However,
this is not good for the alumina concentration stability of a pot (Section 1.3.13).
>
>
a
LL
w
Q j-iJ-i-.l
• j j i i
w ,..J.jJ..i
1 ! i ! (9
w L U . Li
. JLLLi
O f i t !
tf
Li.j-L.1.
w i"i * r.L/i
PQ 14-] y!
.1.. i. —_.-.A ..--
_I ... t. - >
a
a
UJ
•0
o
Aia/!*d oz A|Q/!*d 0 2
M
a
a.
vS A
In the pressure traces, point C often shows a "stalling" effect as the piston starts to
move. This was better at slow speed (Figure 4-5).
This cross-checks with the shot size testing on operating pots where the variability was
better for slow speed w h e n changing from 2.5s to a 4.5s dwell time (Table 2-X) and
indicates that the pneumatics of the pot are one of the main contributors to shot size
variability.
Portland has one of the worst pneumatics for a pot. Compared to Til for example,
Portland has 5 feeders (not 3), 1 2 5 m m bore (not 1 0 0 m m ) , 5 0 8 m m stroke (not 3 5 6 m m ) ,
longer pipe length (20%), and larger pipes ( 3 2 m m not 2 5 m m ) . This amounts to about 7
times the air volumetom o v e the pistons full stroke.
The capacity of a DCV or valve is rated in a dimensionless unit called CV. The CV
rating is a linear relationship to the m a x i m u m air flow rate. The higher the C V , the
higher the possible air flow rate. Portland has a 3 2 m m valve (versus a 2 5 m m valve at
Til) which has a C V of 16 (versus a C V of 14 at Tl 1). Thus, Portland has only a 1 7 %
higher air capacity compared to Tl 1 yet it should have a valve 7 times larger to obtain
similar air flow characteristics. Clearly, Portland is relatively starved of air supply.
Note that the DCV size at Til is similar to that at most Alcoa/Alcan/Alumax plants
with equal or less feeders/DCV. Hence, Portland the worst situation of any of these
plants.
The off-site test rig had a CV of 5. Hence, for the same control on a 5 feeder pot,
ha*
of 25 would be needed. The Portland valve only 6 0 % of this capacity. This is w h y
the feeder "stalls" on the pot and does not on the testrigoff site. The ideal situation
would be one feeder/DCV so that many of the pneumatics problems are minimised.
Feeders that utilise separate crustbreaker/feeder have the feed cylinder pneumatica
operated directly from a cylinder with no spring and little cushioning as the stroke is
only 5 0 m m and the cylinder is only 50-80mm in diameter (Figure 1-12). The
cushioning is not a major issue due to the small size and masses. The dosing cylinder
valve is also usually positioned directly above the cylinder or close to the cylinder,
providing better repeatability as there is no variability due to compression of air in

pipes between the valve and the feeder cylinder.
Pechiney API8 pots have two hoppers with 2 feeders and 2 crustbreaking units on each
(20). The valves for each hopper are on the side of the hopper. The pipe/hose length is
about 3 metres for the closest and 4 metres for the farthest crustbreaking unit. (This is
about the same as that used on the Portland off-site unit viz. about 3 metres). In the
A P 3 0 , the valve is on the feeder itself.
In the case of Portland operating pots, the farthest cylinder on the five cylinder
manifold being controlled by the same D C V is 19 metres from the valve and it takes
about one second to pressurise sufficiently to m o v e the valve. The closest feeder is 10
metres from the D C V . This large difference in distance from D C V to feeder reports as
a slower initial spool speed and different feed rate. In addition, each plunger/spool m a y
start moving at slightly different times each cycle. The performance of one feeder can
affect the variability of the others in the same pot which indicates that plants with 1 or 2
feeders per D C V m a y be less affected by pneumatics comparedtoPortland.
Clearly, both pneumatic testing (Chapter 4) and shot size testing (Chapter 2) indicate
that the pneumatics of feeder design have a significant effect on shot size accuracy, and
that independent feeding is superior to the integrated multi-feeder/DCV design.
4.7 GRADUAL INCREASE IN DWELL TIME
Note also that the dwell time for a plant needs to be revised objectively from time to
time to ensure that the dwell time does not increase gradually over time. Often, if
anode effects occur, management will "err on the safe side" by increasing dwell time.
This can happen on several occasions spaced over years and people do not realise the
gradual change.
An anecdote of this gradual trend of increasing dwell time is an experience the aut
had when visiting a plant with a 4s dwell time when a 2s dwell time would be sufficient
to do the job (T9). In the middle of a presentation on optimising feeders, a discussion
was held on minimising dwell time. A shift supervisor stood up and apologised for the
interruption by saying "I'll be back in a moment after I have dropped the dwell time on
m y Potline".
She halved the dwell time and there was no effect on anode effects. Within a few
weeks all potlines on this plant had halved dwell time. S o m e feeders and mufflers had
to be changed as they were in poor condition but were being protected from detection
by the long dwell time. The observation was made that everyone noted that the time
appeared long, but only when someone outside the normal operating group asked basic
questions did people decidetodo something about it.
We all need to be aware that when we get close to the process we often cannot see t
wood for the trees and getting an outsider to audit the operation can be worthwhile.
4.8 MAJOR FINDINGS FROM PNEUMATICS INVESTIGATION
(i) If not all feeders are operating in a multi-feeder/DCV pot, the speed of the
remaining feeders is faster (which affects shot size for integrated feeders).
(ii) The time that the plunger is wet and the speed of the feeder are affected by
feeder age and muffler quality.
(iii) About half the wet time of plungers is due to the waiting time to begin the
upstroke.
(iv) Operation of one feeder affects the other feeders for a multi-feeder pot.
(v) Partially blocked mufflers cause D C V s to bypass internally and can cause large
air leaks.
(vi) Piston rod failures reduced by 7 5 % after reducing feeder speed at Portland.
(vii) Dual dwell operation saves "wet time" of plungers by 20-70%.
4.9 RECOMMENDATIONS FROM PNEUMATICS INVESTIGATION
(i) T o minimise back pressure on feeders, consider:

- routine muffler changeout or cleaning
- straight through silencers rather than mufflers,
- redirect D C V exhaust to the fume duct or pot cavity
- high temperature lubrication on cylinders.
(ii) Plants should examine their air supply system to minimise the variation of the
speed and wet time of individual feeders and to optimise air usage.
(iii) All plants should consider slow crustbreaker speeds and fast response time for
upstroke to reduce seal wear and plunger wet time. In addition, integrated
feeder plants should consider dual dwell time and regular high pressure strokes.
CHAPTER 5 PLUNGER WEAR PAGE 169
CHAPTER 5
PLUNGER WEAR
5.1 SUMMARY
Plunger life varies from 5 months to 6 years over a range of smelters. The causes of
wear of plungers was examined by routine measurement of O.D.'s of over 10,000
plungers into and out of the workshop, plus about 2,500 in pot measurements of
plungers of various materials, mass, length, and shape, since 1992. Tests on 19
materials were conducted.
A large range of factors was detected as being significant to plunger wear, with the
main ones being (i) penetration distance into the bath, (ii) cumulative wet time and (iii)
material of construction.
The wear mechanism of plungers is a combination of erosion and corrosion, with

corrosion being the predominant factor. Use of high nickel/chrome materials and less
penetration in the bath can reduce wear rates by up to 15 times that of cast iron.
Plunger life over 5,000 days (15 years) is possible.
5.2 INTRODUCTION
Plunger wear is one of the most frequent causes of crustbreaker overhaul for both
independent and integrated feeders (Table 1-V). If the plunger diameter at the base is
.P 4VVS D.UOQB4- 'Yb
too small orAtoo short, bridging m a y occur at the crust and the alumina does not flow
into the bath.
This chapter will examine the pattern of plunger wear and will provide evidence to
support or refute the c o m m o n theories of the plunger wear mechanism. The results of
an intensive experimental program on different plunger materials and designs will be
documented.
5.2.1 Pattern of Plunger Wear
Plungers are cylindrical in shape with a bottom diameter of 80-100mm. The plunger
wears faster at the bottom and generally the shape of a worn plunger is conical as
illustrated for a range of feeders removed from the pots at Portland in Figure 5-1.
FIGURE 5 -1
TYPICAL WORN PLUNGERS
In some plants, the shape can have a sudden drop in diameter 10-30mm from the
bottom e.g. T9, T15, T16. This is generally due to penetration in the metal pad that
causes extreme wear. Plungers m a d e of cast iron can have variable roughness of the
original surface but stainless steel bar is very smooth from new. Irrespective of the
initial surface condition, the worn surface of a plunger is very smooth.
All Alcoa and Alumax plants and most Alcan plants use Alcoa Heat Resistant
specification ( H R spec) cast iron plungers with beveled end similar to plunger C in
Figure 5-2. T h e bevel acts like a chisel to break the crust. Most other plants use
stainless steel bar with flat bottoms.
Often plant trials of plunger wear are not concluded due to the long period required t
assess results and to poor tracking systems for tracing plungers once they have entered
the plant. Little is published on the causes of plunger wear despite a great deal of
experimentation by m a n y smelters. The only paper of significance is one by Kaiser in
1984 (54).
Estimation of plunger life is very subjective for most plants. Little information is
usually kept on plunger life. Life between 5 months and 6 years is quoted, with little
data being available to substantiate the statements. O n e would expect that all feeders
and all pots are the same, however, this is far from the truth. Even within a plant there
is a wide range of life for individual feeders.
5.2.2 Wear Rate Measurement
It is important to specify a parameter that satisfactorily measures plunger wear.
Consideration was given to use of mass loss but it was felt that this is impossible to
measure on operating pots, and extremely difficult in the workshop as the plunger is
generally destroyed in some w a y in order to separate it from the plunger shaft. Metal
loss m a y give a false indication of bottom edge wear, but is difficult for people to relate
to the ability of the plunger to break a hole. A physical measurement is more desirable
FIGURE 5 - 2
TRIAL PLUNGERS
A FLAT BOTTOMMED PA SPEC CAST IRON
B HIGH MASS HR SPEC CAST IRON
C BEVELLED ALLOY COATED CAST IRON
SQUARE SILICON CARBIDE (FIRST TRIAL)
SQUARE SILICON CARBIDE (SECOND TRIAL)

INCONEL 601 (PIPE) 1. NEW
2. END FALLEN OFF
3. PEELED UPWARDS AND SHAFT EXPOSED TO CORROSION
G 80mm O.D. CAST NIRESIST D4
H CAST 31 OSS
O n e of the most important factors in delivering alumina to the liquid bath is a suitably
sized hole in the lowest edge of the crust. The plunger diameter O.D. at the bottom
edge is the most important feature of a plunger. It was also considered that, whether the
attack was corrosion or erosion, the wear will probably be a one dimensional attack, so
diameter would be a simple and reasonable parameter. Throughout this report, wear
rate has been expressed as the "reduction in O.D. of the bottom edge of the plunger over
time". This has been measured in mm/month.
From 1990, Portland has measured plunger O.D. into and out of the workshop to
investigate the wear rate of plungers, and for tradespersons to have a simple gauge to
decide if the plunger needs replacing. Over 10,000 measurements have been made
from 1991 until the end of 1995.
In addition to these routine data, three monthly measurements of plungers in operati

pots has been conducted on about 400 plungers every three months from September
1992 to September 1994.
5.2.3 Theory of Plunger Wear Mechanism
Possible causes of plunger wear include
(a) gas attack when the plunger is raised,

(ii) erosion as the plunger passes through the crust, and
(iii) chemical attack (corrosion) while in the bath.
As there is little evidence of attack of superstructures or the fume system, gas att
was rejected as being insignificant. This leaves erosion and corrosion as the potential
causes.
Three independent assessments of the cause of cast iron plunger wear were conducted
by metallurgical laboratories. Metalab (80) concluded erosion was the cause despite
noting no wear grooves on the surface. It was assessed that the Metalab conclusions
were based on only a few plungers of unknown life in the pot, so their conclusions were
discarded, but their observations were accepted as being valid though limited.
CETEC (81) concluded it was corrosion after making similar observations to the
previous study. D u e to the opposite conclusions from the same types of studies, it was
decided to conduct a third assessment, but this time using more samples and doing a
"blind" assessment where the laboratory was not aware of the history of the samples.
Hans K e m p e (Kempe Engineering), a metallurgist of some note, sourced Greg Hillard
from E T R S as an expert in this type of analysis. K e m p e prepared samples for E T R S to
do microscopic analyses and chemical analyses of 14 plungers of different batches and
ages. Hardness tests were conducted on different locations on the sides and top of the
resultin lon
test plungers as it was thought that high hardness should ger plunger life. The
results are presented in Kempe's report (82) and are discussed below.
Assessment of hardness can be measured using Rockwell or Brinell measurement units.

The Rockwell method uses a spike that presses into the test material and the depth of
penetration indicates a rating of hardness. The Brinell method uses a small ball that is
forced onto the test material and the diameter of the indentation indicates the hardness
rating. It w a s found that the Brinell method was more repeatable for plunger
measurement because the spike of the Rockwell method is affected by the grain
boundaries of the cast iron. The Brinell ball tends to be more robust for the location
where the test is conducted. Hence, all hardness measurements in this study use Brinell
units (BHN).
The main observations from these studies were as follows:
(i) no microscopic differences between the edge and the middle of the
plunger (80,81),
(ii) no sign of grooves from erosion (81),
(iii) no grain spheroidisation and dissolution of pearlite which would indicate
a heat affected zone (82),
(iv) comparison of 14 plungers showed wear rates of 1.7 to 6.5mm/month,
(v) the longer lasting plungers had molybdenum contents of 4 % and copper
content of 0.2% (versus 2 % and 0.5% respectively for average plungers
tested),
(vi) hardness was over 2 0 0 B H N for the better wearing plungers versus
1 6 0 B H N for the worst (82),
(vii) high wear rate plungers had less chrome (82), and
(viii) there was no difference in hardness between worn surfaces and surfaces
that had not contacted the bath (82).
The final conclusion was that the cause of wear was mainly chemical attack with
flaking off of the reaction deposits occurring on each movement into the crust.
Abrasive attack was present but was not considered to be the main factor. This rusting
theory was supported at T 9 where analysis of the material that fell off some plungers
had Fe203 (rust products) on the inner face which previously had been touching the
cast iron plunger.
To test the corrosion/erosion theories in a simple way, plungers made of Inconel 601
pipe were placed in service at Portland in February, 1991 (Figure 5-2 plungers Fl, F 2
and F3). Inconel has a hardness of 1 3 0 B H N which is about half the hardness of H R
cast iron. If hardness is the main factor restricting plunger wear, then the Inconel would
not wear as well as cast iron. However, the Inconel proved to have a wear rate of about
1 4 % of cast iron. This clearly showed that corrosion is a dominant factor.
Following the metallurgical investigation, a specification for cast iron ("PA spec"
Portland Aluminium specification) was developed by K e m p e , Hillard and Kissane (82).
This was introduced in January 1992 (Figure 5-2 plunger A ) . B y November 1992, it
was not considered that any significant improvement in life was attained by changeover
from H R to P A spec, so investigations commenced into alternative materials and
operating parameters tofindwhat are the causes of plunger wear and what materials are
better than H R cast iron. Figure 5-2 illustrates several of the plungers tested.
In hindsight (3 years later), the mistake made in the metallurgical studies was that,
although the life of the test plungers was known, there was no knowledge of the pot
where the plungers were used. Later it was found that the variability of plunger wear
and hardness was within normal variability, so to compare "good plungers" with "bad
plungers" was inappropriate as the cause of this large variability was the duty in the pot,
CHAPTER 5 PLUNGER W E A R P A G E 176
rather than the material of construction (Section 5.3.2). This is one of the problems of
analysis of what statisticians call "happenstance data" where one does an analysis of
data that just "happens". It is far preferable to design a controlled experiment where the
experimenter knows all the parameters and measures their changes in a log as the
experiment unfolds. The reason this was not done at the time was the long time to get
data out of plunger experiments and the lack of understanding of what parameters are
important in plunger wear. It turned outtobe a costly mistake as the P A spec
proved to be 4 0 % worse than H R spec (Section 5.6.3).
5.2.4 Procedure for Analysis of Plunger W e a r Mechanism
If corrosion or erosion are joint causes of plunger wear, then some factors which m a y
contribute to plunger wear would include those itemized in Table 5-1. The effect of
these variables can change from plunger-to-plunger both within and between pots.
T A B L E 5-1
PARAMETERS POTENTIALLY
AFFECTING PLUNGER W E A R
CORROSION
Electrolysis
Wet time - Wet time per crustbreak
- Crustbreaker frequency
- Cumulative wet time
Penetration depth in bath
Temperature - Bath
- Plunger
Anode effect control
Plunger mass
Plunger buildup
Plunger material
EROSION
Plunger speed
Crust hardness
Plunger hardness
Plunger shape
Crustbreak frequency
T o determine evidence of plunger wear, extensive trials were conducted on different

plunger materials in a range of pot conditions from February 1991. Routine plunger
measurements into and out of the workshop (from M a y , 1991) plus three monthly
measurements on operating feeders (October, 1992 to September, 1994) gave early
indication of successes and failures on different plunger materials and plunger designs.
To establish the factors affecting plunger wear and the relative wear rates of altern
materials tested at Portland, plungers were measured by a pair of calipers each time a
feeder came into the workshop irrespective of the cause of failure. This gave a range of
life from days to years so a good assessment of wear rates resulted. The tools made to
measure plunger diameter on operating pots was a large pair of calipers on poles. A
gauge capability study on these tools showed the in-plant tool had a standard deviation
of 0.7mm, and the workshop tool had a standard deviation of 0.5mm.
This Chapterfirstlyexamines w h y the wear rates of plungers vary so much, then
establishes a theory on the wear mechanism. Data from test work have been compiled
against plunger wear parameters raised in Table 5-1 and this is discussed in the
following Sections 5.3 to 5.6. Following this assessment, in Section 5.7, conclusions
are m a d e on the main parameters causing plunger wear. This then leads to
opportunities for life improvement and different materials of construction.
(This Chapter only discusses plunger life. Section 9.3 discusses the economics of
different plunger materials and plunger designs.)
5.3 VARIABILITY OF PLUNGER WEAR RATE
Before one can consider methods to improve plunger life, one must understand the
causes of variability of plunger wear. Using mainly Portland data, measurement of
plunger wear is discussed below.
5.3.1 Overall Observations
Figure 5-3 shows the "shot gun" nature of plunger O.D. at repair for 1100
measurements. Plungers that last a very long time are probably not touching the bath
and/or crust, so have no corrosion or erosion. Clearly there is a great deal of scatter of
wear rates which is not u n c o m m o n with most plants. The causes will be discussed
below.
FIGURE 5 - 3
PLUNGER DIAMETER VERSUS LIFE

REPORT PLUNGER DIAMETER VS LIFE: SCATTERGRAW
Date Repaired fram 1 January 1991 to 30 Sept amber 1995
Pari No: 351
A
M
E
T
E
R
50
T o gain any understanding of plunger wear, one has to identify the causes of this
variability. This has to be done before contemplating changes to material of
construction. It is best to optimize what one has, before changing it.
5.3.2 Effect of Single Batch Variability on Plunger Wear Rate
It is possible that the cause of variability is at manufacture. It was decided to do a trial

of similar plungers and follow their performance in the pots over a period of years.
These plungers were examined prior to installation in the pots to establish if they were
different, and then assessed the wear rates w h e n they were removed.
A set of 60 HR spec cast iron plungers were tested. Cast iron plungers are made in an
inductive furnace. Each batch (heat) was poured into a ladle which in turn was poured
into a group of sand moulds which contain five plunger shapes. The plungers were
coded individually with heat (4 were made), mould (3 were m a d e from each heat) and
position in the mould (5 positions).
Each furnace heat was analyzed separately using a spectrometer and a potential gauge
capability study was held on the spectrometer. This technique is a statistical analysis to
determine the accuracy and repeatability of a gauge or measuring process. The
spectrometer analyses were statistically "in control" for all elements except silica
(which is not considered an important element for wear). Hence, the plungers were
identical chemically before service in the pot.
It was considered that if erosion was important in wear, then hardness should be an
important variable. Firstly, the measuring unit had to be tested for capability and a
testing location had to be selected that gave repeatable results. Brinell hardness was
tested at m a n y locations on several plungers. It was established that they could be tested
at any location on the plunger and would have a reading that was within 20 units (on a
9 5 % confidence level for an average reading of 1 7 0 B H N ) .
Hardness results for the 60 plungers are represented in Table 5-II.

TABLE 5-II
PLUNGER HARNESS TRIAL DATA
HR SPEC CAST IRON

M O U L D HUSI1IUN 1 2 a 4 b
7 RIAL #1 - H K SPLC BHN W t A P T BHN W t A K BHN W b A K BHN W b A K BHN WEAK
15-May-1991 RATE RATE RATE RATE RATE
HbAI MOLL) (mm/m) (mm/m) (mm/m) (mm/m) (mm/m)
A 1 217 1.5 207 5.3 229 207 5.8 217

2 197 200 4.4 217 207 2.7 207 2.0
3 207 207 179 1.9 207 0.2 201
B 1 212 2.5 197 3.4 207 4.1 217 197

2 212 217 2.6 207 2.0 223 212
3 212 2.0 197 217 3.8 217 7.8 201 3.4
c 1 197 1.8 217 3.1 223 3.0 207 207

2 197 5.3 197 1.1 197 4.4 191 2.2 201
3 187 2.8 217 207 207 2.6 187 2.7
D 1 217 2.5 212 6.4 217 207 1.6 212 4.1

2 201 2.5 207 212 4.3 217 1.5 212 3.8
3 212 207 4.1 212 3.3 212 217 2.1
Average 206 2.6 207 3.8 210 3.4 210 3.1 206 3.0
Standard Deviation 10 1.2 8 1.6 13 1.0 8 2.5 9 0.9
Minimum 187 1.5 197 1.1 197 1.9 191 1.5 187 2.0
Maximum 217 5.3 217 5.5 229 4.4 223 7.8 217 4.1
Number of data points 12 8 12 7 12 8 12 6 12 6
MOULD No. #1 #2 #3 M O U L D No. #1 #2 #3

W E A R RATE (mm/month) HARDNESS (BHN)
Average 3.0 3.1 3.3 Average 211 207 206
Standard Deviation 1.1 1.4 1.6 Standard Deviation 8 8 11
Minimum 1.5 1.1 1.9 Minimum 197 191 179

Maximum 5.3 5.5 7.8 Maximum 229 223 217
Number 11 13 11 Number 20 20 20
HR SPEC OVERALL HARDNESS W E A R RATE

(BHN) (mm/month)
Average 208 3.2
Standard Deviation 9.5 1.4
Minimum 187 1.1

Maximum 229 7.8
Number of data points 60 35
Note: G a p s in the table are due to missing wear rates or wear rates over 10mm/month (which are probably
due to errors in dates in pots).
Tests found that the position in the mould (1,2,3,4,5 on the top of Table 5-Et) and each
of the four moulds (A,B,C,D) had statistically similar wear rates. Hardness varied
between 1 8 0 B H N and 2 3 0 B H N . Hence, manufacture of cast iron plungers w a s
repeatable for hardness. Note that this variability was similar to plungers of different
ages tested in the K e m p e / E T R S investigation in Section 5.2.3.
Thus, the 60 plungers were identical chemically and had the same hardness. Any
variability in service would be due to pot variability and not to h o w the plungers were
made.
These plungers were placed in service within a two week period and the wear rate was
determined w h e n the feeders were removed over the next 3 years. Table 5-II shows
that there was no statistical difference in wear rate between heat, mould or position in
the mould. The plunger wear rates were variable but not statistically different. Plunger
wear rate varied from l.lmm/month to 6.4mm/month (which was similar to the
variation seen in samples tested by K e m p e / E T R S (82) mentioned in Section 5.2.3).
In 1992, a similar trial was conducted using PA spec cast iron (Table 5-ffl). This
showed similar results in wear rates and no pattern between heat, mould or position.
Once again, the manufacture process was not the cause of plunger wear variability for a
cast iron of different analysis.
In both trials, the standard deviation was 30-50% of the average for these "identical"
plungers. The variability was thus very large. S o m e plungers could last 5 times longer
than others despite being identical w h e n placed in the pot. This suggests that variability
from pot-to-pot (such as bath level, temperature, penetration distance into the bath and
time in the bath) is the main reason for this plunger wear variability.
TABLE 5-HI
PLUNGER HARDNESS TRIAL DATA

PA SPEC CAST IRON
M O U L D POSITION 1 2 53 4
TRIAL #2 - PA SPEC BHN WEAR BHN WEAR BHN WEAR BHN WEAR BHN WEAR
25-May-1992 RATE RATE RATE RATE RATE
HEAT MOULD (mm/m) (mm/m) (mm/m) (mm/m) (mm/m)
A 1 321 341 5.3 363 4.1 341 331 1.2

2 341 3.9 341 4.9 341 6.0 321 5.3 363 3.9
3 341 2.8 321 321 3.1 331 5.2 341 5.4
B 1 363 331 341 5.8 363 363

2 363 7.9 341 4.1 341 4.9 341 1.5 341
3 375 5.0 341 6.0 341 6.3 341 3.8 341 2.1
C 1 362 341 4.3 341 4.8 341 5.0 341

2 363 331 341 352 341 5.2
3 341 4.1 341 352 7.6 341 5.3 363
D 1 363 3.6 363 3.1 341 3.1 341 4.5 375 6.9
2 341 3.0 321 5.9 363 341 363 3.3
3 363 5.4 321 388 6.8 341 341 8.9
Average 353 4.5 336 4.8 348 5.3 341 4.4 350 4.6
Standard Deviation 16 1.7 12 1.0 17 1.5 10 1.4 14 2.5
Minimum 321 2.8 321 3.1 321 3.1 321 1.5 331 1.2
Maximum 375 7.9 363 6.0 388 7.6 363 6.3 375 8.9
Number of data points 12 8 12 7 12 10 12 9 12 6
MOULD No. #1 #2 #3 M O U L D No. #1 #2 #3

WbAK RATb (mm/month) HARDNESS (BHN)
Average 4.3 4.6 5.5 Average 348 345 344
Standard Deviation 1.4 1.5 1.6 Standard Deviation 14 13 17
Minimum 1.2 1.5 2.8 Minimum 321 321 321

Maximum 6.9 7.9 8.9 Maximum 375 363 388
Number 12 13 15 Number 20 20 20
PA SPEC OVERALL HARDNESS W E A R RATE

(BHN) (mm/month)
Average 346 4.8
Standard Deviation 15 1.6
Minimum 321 1.2

Maximum 388 8.9
Number of data points 60 40
Note: Gaps in the table are due to missing wear rates or wear rates over 10mm/month (which are probably
due to errors in dates in pots).
CHAPTER 5 PLUNGER W E A R PAGE 183
5.3.3 Effect of Batch-to-Batch Variability on Plunger W e a r Rate
To establish if there was a batch-to-batch variability in wear rate from the plunger
manufacturer, the wear rate of a random batch of plungers each month for 20 months
was assessed for wear rate. A control chart of these data is presented in Figure 5-4.
The batch number relates date, month, and year e.g. 13F90 was made on 13th June,
1990 and 11K91 was 11th October, 1991 (where "I" is not used as it looks like "1").
FIGURE 5 - 4
CAST IRON PLUNGER W E A R BY BATCH NUMBER

WEAR RATE (mm/month)
10 10
HR SPEC I PA SPEC
2s DWELL 4s DWELL 1
FAST
" SLOW '
SPEED (SPEED 1 (JCL
.^^yt^^-^Â^^
I LCL _
1 1 2 0 1 2 2 0 0 2 0 2 0 1 1 2 1 1 0 0 1 0
3 4 3 9 5 3 4 1 8 2 1 4 5 4 6 5 1 5 7 4 4 6
F F G K K L A B B B c E G H J J K L M B B C
9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9
STANDARD DEVIATION
HR SPEC DA CDCJ**
- 2s DWELL 4s DWELL ' r M 0rC
°
FAST SLOW '
SPEED I SPEED j (JCL
1 1 2 0 1 2 2 0 0 2 0 2 0 1 1 2 1 1 0 0 1 0
3 4 3 9 5 3 4 1 8 2 1 4 5 4 6 5 1 5 7 4 4 6
F F G K K L A B B B C E G H J J K L M B B C
9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9 9
0 0 0 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1 2 2 2
A s the 20 month data are inside the control limits, despite having a poor variability, the
batches are statistically in control i.e. they are predictable 9 5 % of the time. All the
batches have statistically the same wear rates. Manufacture of cast iron plungers has no
statistical effect on plunger wear rate.
As was found in the single batch trials, the standard deviation is about half the avera
The average wear rates of batches of plungers were between 3 m m / m o n t h and
7mm/month, so the spread of wear rates for individuals was m u c h wider than was
experienced for individuals in the single batch trial. This illustrates the compounding
effects of within-pot variability and across-pot variability. In the single batch trial, all
feeders were installed at the same time so had the same potline conditions viz. current,
potline changes in bath level, potline changes in resistance. Here the plungers were
installed at different times, so there m a y be time-related changes in pot process
parameters in additiontothe variability seen before. This m a y explain w h y individual
wear rates were m u c h more variable than in the single batch trial.
Hence, trying to optimize plunger material seems a bit pointless until the reasons for
influence of pot conditions are understood and all that is possible is done to minimize
this variability.
These results also highlight how important it is to control experiments of different

plunger materials to be within-pots to minimize factors that m a y influence the
significance of results. The bigger the variability, the harder is to determine whether
the differences in wear rate are due to natural variation in the process or to the plunger
material itself.
5.3.4 Effect of Diameter on Plunger Wear Rate
In order to compare wear rates, it is possible that plungers m a y wear at different rates
over time. Is the wear rate the same w h e n the plunger is 1 0 0 m m O.D. as it is at 6 0 m m ?
If there is a significant relationship, one cannot compare wear rates shortly after
installation to those w h e n the plunger is well worn.
To detect if this is important, P A spec cast iron wear rates were extracted from three
monthly measurements on operating feeders. T o avoid pot-to-pot variability and
changing pot conditions, the wear rates of different plungers in the same pot were
ratioed to each other, then the data were sorted by the original size (Table 5-IV).
For example, consider two plungers (A and B) that were measured at the start of a th
month period at 1 0 0 m m and 5 0 m m respectively. If they were measured again at 9 4 m m
and 3 8 m m respectively at the end of the three month period, the wear rate would be
2mm/month for A and 4mm/month for B. The wear ratio A/B would be 0.5.
In a similar way the data were sorted and grouped into 90-100, 80-89, 70-79, 60-69,
and 50-59mm starting sizes. These were called groups 90,80,70,60, and 50
respectively. Each of these were then sorted by their ratio to each of the other sizes.
Hence, a matrix was developed of median wear ratios of "original size" on the Y axis,
and "wear ratio" on the X axis. The final matrix is shown in Table 5-IV with a
composited graph, which ratios all data to a 90-100mm starting size. A total of 395
wear ratios were used to generate this table.
There appears to be a reduction of wear rate as the plunger gets smaller. For exampl
for cast iron plungers, the wear rate at 1 0 0 m m cast iron plunger is about twice that at
5 0 m m . The effect for more passive materials is less. A similar analysis has been
conducted using Niresist and H R spec cast iron plungers with similar results, however,
the wear rate does n6t drop as quickly for the Niresist plungers possibly due to its
passive nature as a result of high chrome/nickel content. It is reasonable to assume that
there is probably a different curve for each material, with the more passive materials
having minimal dependence of wear rate on plunger diameter.
A trial of twenty 80mm Niresist plungers (Figure 5-2 plunger G) showed that the 80mm
wear rate was 1 8 % lower than 1 0 0 m m plungers. This reinforces the theory that plunger
wear rate reduces as the plunger diameter reduces.
TABLE 5-IV
CAST IRON WEAR RATE MATRIX

COMPARISON
STATISTIC GROUP ORIGINAL PLUNGER O.D. (mm)
O.D. (mm) 90 80 70 60
Median Wear Ratio 90 1.0

80 0.9 1.0
70 0,9 0.8 0.9
60 0.6 0.7 0.8 0.9
50 0.3 0.6 0.8 0.9
Pooled Median Wear Ratio 90 1.0

80 0.9 0.9
70 0.8 0.7 0.9
60 0,6 0.6 0.7 0.5
50 0.5 0.5 0.7 0.5
Sample Size 90 40
80 35 32
70 23 37 35
60 11 20 55 26
50 15 7 16 23
Notes:
(i) Median wear ratio compares median wear rate of a Y axis value with an X axis
value eg. for cell 70 vertical/90 vertical, the median wear rate of the 70-79mm
plungers is divided by the median wear rate of the 90-100mm plungers.
(ii) "Pooled median wear ratio" uses the mean wear ratio weighted by sample size for
each cell, eg. 60/90 cell is calculated as follows:
(0.6 x 11) + ( 0.9 x 0.7 x 20) + (0.8 x 0.9 x 55) + (0.9 x 0,6 x 26) =0.6
(11 + 2 0 + 55 + 26)
(iii) Median wear ratios of 395 comparisons of P A spec cast iron plungers of different
starting diameters in the same pot for the same 3 month period.
POOLED MEDIAN WEAR RATE RELATIVE

TO 100mm STARTING WEAR RATE
0.8 —
2
ne
0.6-
2
S
z
0.4
o
Ul
s
111
r- 0.2--
5
iu
VC
0.0
90 80 70 60 50
STARTING O.D. (mm)
S o m e reasons for reduced wear rate as the plunger gets smaller could be:
(i) the plunger surface is farther from the crust so is cooler and the corrosion
rate is slower,
(ii)the hole is smaller so there is less radiant heat, a cooler surface, and lower
corrosion rate, and
(iii)the plunger is not affected by erosion due to the wedging shape as it hits the
crust.
Hence, comparisons of plunger wear should only be calculated when the original O.D.
of the two different plungers are comparable; say, within 1 0 m m of each other.
5.3.5 Effect of Between-Pot and Between-Plant Variability on Plunger Wear Rate
Consider Figure 5-5 where plungers were placed in the pot at the same time. The life to
get to 5 0 m m varied from 450 days for pot 1002 to 250 days for pot 2016. Note that the
duty of plungers in these pots is theoretically identical - same line, current, dwell time,
and (for 8 of the 10 plungers) the same plunger batch (15K90). This illustrates h o w
m u c h wear varies from pot-to-pot. It is best to compare rates within a pot rather than
across pots.
It is even more difficult to compare across plants. Here there are many factors that m
affect wear rates. These include:
number of feeders per pot, type of feeder (independent or integrated), DCV size,
crustbreak frequency, pot size, penetration depth, muffler type, dwell time,
current, shot size, air pressure, plunger mass, plunger manufacturer quality
control, original diameter, anode effect kill strategy.
Thus it is unwise to read too much into what one plant can achieve compared to
another. O n e needs to do comparisons within one plant (preferably in the same pot) to
avoid these variables.
FIGURE 5 - 5
POT-TO-POT PLUNGER WEAR VARIATION
Pot 1002
Plunger size (mm)
120
H R Spec
-*- -i- -4- -i- -4- -i- -4- -U

0 50 100 150 200 250 300 350 400 450 500
Days
Plunger No/Location
- -
-"-15K90 ~r~09K90 * 09K90 •""15K90 "*"15K90
Date Installed : 16 Nov 1990
Pot 2016
Plunger size (mm)
120 HRSpec
100 *
80
60
40
20
1 1 1 i 1 — i 1 1 1
50 100 150 200 250 300 350 400 450 500
Days
Plunger No/Location
"•"16K90 -T-16K90 -*T16K90 -""15K90 -*16K90
On Line November 1990

5.3.6 Effect of Within-Pot Variability on Plunger W e a r Rate
Figure 5-6 illustrates variation within a pot. Pot 2090 had one plunger reach 5 0 m m in
400 days and the rest reach 5 0 m m in 600 days. Pot 4087 had a scatter of wear rates to
give lives between 150 and 300 days. Note also the difference of pot-to-pot as well. In
the case of 4087, four of the plungers came from the same batch and the best and worst
plunger in the pot were from the same batch.
Wear rate can be up to 100% different between plungers in the same pot despite bein
the same original size from the same batch and installed at the same time.
FIGURE 5 - 6
WITHIN-POT PLUNGER WEAR VARIABILITY
Pot 2090
Plunger size (mm)
120
HR Spec
100 *
80
60 - ---^
40 -
20
1 i 1 1 1 i 1— 1 1 1 1 1 i
0 50 100 150 200 250 300 350 400 450 500 550 600 650
Days
Plunger No/Location
""-/1 -r~/2 -"-/4 -X-14F90/6

Date Installed: 22 Aug 1990
Pot 4087
Plunger size ( m m )
120
100
Plunger No
-•-22B91 +121391 *K'22B!jVt "•-22B91 "*-22B91
Date Installed : 19 M a y 1991

CHAPTER 5 PLUNGER W E A R P A G E 191
5.3.7 Effect of Location Within a Pot on Plunger W e a r Rate
Table 5-V shows plungers of H R and P A spec cast iron in the same pot for the same
length of time. The wear rate of each was divided by that of location #5 and the
locations were compared in the table. Although the average increases towards location
#1, the minimum and m a x i m u m highlight that this is not significant as the lowest wear
rate is in fact a #1 plunger and the standard deviations for each location are 20-60% of
the average.
There is no statistical evidence to prove that wear rate varies with location.
TABLE 5 - V
EFFECT OF LOCATION ON PLUNGER WEAR

HR SPEC CAST IRON
LOCATION
1 2 3 4 5
MEAN 1.3 1.1 1.1 1.0 1.0
STD DEV 0.6 0.3 0.2 0.6 0.0
MIN 0.6 0.6 0.8 0.1 1.0
MAX 2.5 1.7 1.4 2.2 1.0
COUNT 10 10 8 10 18
PA SPEC CAST IFION
LOCATION
1 2 3 4 5
MEAN 0.9 0.9 1.1 1.1 1.0
STD DEV 0.4 0.4 0.4 0.5 0.0
MIN 0.3 0.4 0.6 0.3 1.0
MAX 1.5 1.5 2.0 1.9 1.0
COUNT 8 7 9 10 18
Note: Values in table are calculated as follows:

relative wear rate = (wear rate/wear rate in location 5)
e.g. for H R spec, location 1 has a mean wear rate 3 0 %

higher than location 5.
5.3.8 Recommendations for Data Analysis of Plunger W e a r Rates
It has been seen that there is a large variation in plunger wear data due to across-pot and
across-plant variation. In this type of situation, assessment of different materials cannot
be m a d e in a statistical comparison unless
(i) the mean wear rates are different by several hundred percent, or
(ii) the sample sizes are very large.
Clearly, as it is desirable to detect small differences in wear rates of different mat

and with the high cost of different exotic materials, one cannot tolerate very large
sample sizes in experiments of different plunger materials. Several recommendations
are m a d e in order to reduce the sample size.
(i) Sample period

A s seen in Section 5.3.2 and 5.3.3, wear is affected by pot condition. This variability is
minimized by comparisons in the same pot over the same time. It is not possible to
avoid within-pot variability (Section 5.3.6), but it does not matter which locations one
chooses as it has been established that location is not significant (Section 5.3.7). T o
avoid short term changes to pot conditions or timing errors, a period of 3 month
intervals is wise. The 3 month period gives a trend relatively quickly, so that one does
not need to wait years to see the effect of different designs.
(ii) Original O.D.

A s mentioned in Section 5.3.4, plunger wear reduces as the plunger gets smaller.
Hence, comparative trials must be between plungers of similar original O.D. e.g. within
10mm.
(iii) O.D. Measurement Accuracy

Comparisons of diameter will be affected by the accuracy of the measurement. From
gauge capability studies, it was established that a wear rate should not be used for mean
(average) wear calculations unless there was a diameter reduction of at least 5 m m .
However, median wear rates can use all data because medians are not affected by
extreme values in a population. For example, averaging a sample of nine values at 0.1
and one at 3.0 gives a mean of 0.4. A better estimate of this sample would be 0.1
(which is the median). Medians are more robust to outliers than means, so will be used
in future for comparisons. Most comparisons also table means for information
purposes.
(iv) Sample Size

For statistical significance calculations, it is preferable that the sample size is at least 8;
the higher the better. Standard deviations can be used for values of 8 or more.
Standard deviations are a useful measurement of plunger wear variation as the
populations of the sample sizes generally have a normal distribution.
Hence, the following conditions are recommended for plunger wear data of different
plunger materials or feeder conditions:
(i) compare plungers in the same pot over the same time (3 months minimum),
(ii) plungers to be within 1 0 m m O.D. of each other at the start of the trial,
(iii) use medians, and
(iv) use sample sizes of at least 8 and calculate standard deviations.
5.4 TESTING CORROSION THEORY
It is unlikely that electrolysis is a major factor in plunger wear as it would be expected

that the whole bottom end of the plunger would dissolve. This would mean a reduction
in length. This does not occur on plants except those with insulation breakdown and
concurrent long periods of the plunger in the bath. Plants with arcing or with deep
penetration into the metal are the only cases detected that could possibly have
electrolysis problems. This source of corrosion is therefore not discussed further,
except in Section 8.7 where arcing is discussed.
5.4.1 W e t T i m e
Given that plungers generally cycle regularly in the bath, any time that the plunger is in
the bath there is more risk that the temperature and shedding of rust layers will be
aggravated by immersion time.
Consider the effect of wet time which comprises
(i) wet time per crustbreak,

(ii) frequency of crustbreak cycles, and
(iii) cumulative wet time viz. (i) x (ii).
5.4.1:1 Wet Time per Crustbreak
Dwell time is defined as the time that the air extends the crustbreaker to when it retracts
the crustbreaker....DCV "on" to D C V "off. It includes the reaction time of the valve,
down stroke time and "wet time" (where the plunger is potentially stationary in the
liquid).
For the purpose of this discussion, wet time will be considered as the time in the bat
priortoD C V operation to retract the plunger, and assume that the retract delay time is a
fixed ratio to the dwell time. This m a y not be exact, but it is probably a reasonable
assumption. It will be more inaccurate for pots with 3-5 feeders than one-feeder/DCV
pots as the delay time is greater due to large air lines. Portland, Til, D 1 2 and T 7 m a y
show conservative wet time values as times m a y be up to double that indicated for these
multi-feeder/DCV pots.
For most situations, the reaction time is consistent if the valve is in good condition
there are no air leaks and there is no back pressure. Stroke time is consistent if there is
constant friction on the drive mechanism and cylinder seals. Hence, the main criterion
considered in most plants is changing dwell time to vary the wet time. Section 4.6
shows, for pots with multiple feeders per D C V , other factors can affect wet time. Each
feeder and pot m a y have quite different response characteristics due to D C V efficiency,
number of feeders on line, air line length and feeder type. Thus, there are potential
errors introduced by using dwell time as wet time.
Figure 5-7 shows the plant average wear rate for different wet times per crustbreak
similar plunger specification for 13 smelters. There appears to be a general trend of
reduced wear for less wet time.
FIGURE 5 - 7
CAST IRON PLUNGER WEAR RATE VERSUS

WET TIME PER CRUSTBREAK
WEAR RATE
(mm / month)
T9
7 at
~m FEED ON ANODE EFFECT
6
A DO NOT FEED ON ANODE EFFECT
T6(1)
•
T6(3)
5 •
- T6(2) D12
3
2 T3
T11
•
1 T16
£6
1 T2
•
Y
f) 1 1_ — I — I 1 1
0.5 1.0 1.5 2.0 2.5 3.0
W E T TIME / CRUST BREAK ( s )
Note: Cast iron specification is the s a m e at all plants and T1 is Portland.

The requirement of selecting a dwell time is to have sufficient time, after the normal
stroke time, to cater for such things as:
(i) alumina discharge (for integrated feeders),

(ii) piston seal leakage,
(iii) variability in mains pressure,
(iv) crust hardness variability,
(v) D C V condition,
(vi) cylinder friction from n e w seals, and
(vii) air leaks.
For pots that have multiple feeders per DCV, the dwell time is also related to the
historical slowest feeder on the manifold (Sections 4.6 and 4.7).
Independent feeders do not have to wait for alumina discharge so have lower dwell time
and wet time compared to integrated feeders. They only need to reach full pressure. A s
the D C V is usually close to the feeder itself (within 2 m ) , the response time is much
faster than for an integrated feeder which m a y have pipe lengths of 2 0 m (Section 4.6).
integrated feeders usually have a wet time of 1.2-2s versus about 0.2s for independent
feeders. Clearly there is potentially a greater effect on dagging and plunger wear for
integrated feeders.
Tests at T9 have shown that very long dwell time does not necessarily increase the
chance of breaking a hole. Experience in each plant will determine the optimum dwell
time, but results at T 2 and T3 suggest times of about 1.0s are sufficient after completion
of the stroke for integrated feeders. From A 3 tests at Portland, independent feeders
could be retracted within 0.2s of the end of stroke.
The reduction of dwell time is usually at zero cost, with potentially significant
improvement to plunger life. Following recommendations by the author, all Alcoa
smelters reviewed dwell time. T 9 and T15 halved dwell time in 1991/92, with many
other plants reducing by 10-30%. Most plants found that the long dwell times were
covering other errors in the hardware...such as blocked mufflers, air leaks, deteriorated
air hoses, faulty valves. All plants have improved feeder life since 1992 (Section
10.3.2).
Another observation of time in the liquid is the pattern of wear. Plants that have a
long stationary time (T9, T15, T16) m a y experience a "bottle" shape wear which is
often referred to as a "short" plunger. This shortening is one of the major causes of
removal (Table 1-V). Plants with less stationary time each crustbreak show a more
uniform conical shape. Short plungers m a y be due to rapid corrosion from the end of
the plunger penetrating the molten aluminium in the pot.
FIGURE 5 - 8
CAST IRON PLUNGER WEAR RATE VERSUS

CRUSTBREAK FREQUENCY
WEAR RATE
(mm / month)
T9
7 •
m FEED ON ANODE EFFECT
6 T6J1) A D O NOT FEED ON ANODE EFFECT
T6(3)
5
T6(2) D12
3 • T4 •
• T15 T1
T11 A
2 • T"6 T3
A17 T2
1
0 .. J I 1 I. 1 -J 1 1 1 1
100 110 120 130 140 150 160 170 180 190 200
CRUST BREAK FREQUENCY ( s )
Note: C a s t iron specification is the s a m e at all plants a n d T 1 is Portland.
5.4.1:2 Crustbreak Frequency
Figure 5-8 illustrates the effect of crustbreak frequency of wear rates of 13 smelters.
Despite the variability from plant-to-plant, there is a general trend that the less frequent
the breaks, the lower the wear rate. This could be relatedtoeither erosion or corrosion
(or both)
An example of the effect of crustbreak frequency is for pots that were converted fro
A E D DtoA 2 feeders. Figure 5-9 shows the change from a normal A E D D operation to
an A 2 feeder operation when the frequency of feed is 5 0 % more frequent than the
A E D D duetoa smaller shot size. The A E D D would have reached 5 0 m m in about 400
days, yet the A 2 would have failed at about 250 days - roughly a similar ratio to the
shot size ratio between normal and A 2 feeders.
FIGURE 5 - 9
EFFECT OF CRUSTBREAK FREQUENCY

CHANGE ON PLUNGER WEAR RATE
Pot 2009
Plunger size (mm)
120
A2 Inst. 26-M1 HR Spec
1005
AEDD ^ " " " ^ ^ N l l t > ^ r ~ " ^^s^v^
A2 6.0mm/month
80
AEDD 3.8mm/month
60
40 -
20
. 1 1 1 1 -I 1 1 u 1
50 100 150 200 250 300 350 400 450 500
Days
22B91/1 01O91/2 "^"l8B91/3 22B91/4 - * 22B91/6
Date Installed 25 May 1991
This increase in wear rate was a similar effect to the 2 5 % increase in wear rate when
there was a 2 5 % reduction in shot size on installation of spool inserts.
Figure 5-10 is a "box and whiskers" graph of plunger wear rate versus month repaired.
This type of graph shows a box with "tails" at top and bottom with a horizontal line
inside the box. The "whiskers" indicate the highest and lowest value of the population.
The top (and bottom) of the "box" are the values that have 2 5 % of the population
greater than (and less than) that value. The horizontal line in about the middle of the
"box" is the median value in the population. After July 1992, there is an increase in
wear associated with the introduction of inserts. Median wear rates increased from
4mm/month (between points 2 and 3) to 5mm/month (between points 3 and 4).
Thus, there appears to be a general relationship between the crustbreak frequency and
wear rate.
Safji»L
Q
UJ
H
CO
LU
H
Q LU
or
W CO o
h- in
-J a 0_
LU: CO
CO
-J £
H z
ro CO
haa*
CO
fK
H LO III
LL O
LU
"/J
o 1 CD _1
ad z
D n
0- CO
c ron
PQ u i
§ as o
5
LU _i
CO. z
a <r
< i-
0_ co
CM m
OS UJ
o CO
Q
I-
_i
p LU
LU
-I
Uj
-a
-J D- §
CM CO g
c
"Cl
a)
Ig
CO Q
0)
ID
O CM
CL r--
5 ffl (J J rl fi « ») it M 6 00
Ul N a- T- T- T- t-
5 • O I- CKO-H-B v££\I OC-.Er-%
5.4.1:3 Cumulative W e t Time
W h e n crustbreak frequency and wet time per crustbreak are jointly considered (Figure
5-11), the pattern of plant plunger wear rates versus cumulative wet time shows a
clearer relationship than wet time/crustbreak (Figure 5-7) and crustbreak frequency
(Figure 5-8). The correlation coefficient (r2) was 0.7 which means that 7 0 % of the
variability of wear rate can be explained by cumulative wet time. Considering the graph
is based on "whole of plant" data, this is a good correlation given the circumstances.
FIGURE 5 -11
EFFECT OF CUMULATIVE WET TIME
ON CAST IRON PLUNGER WEAR RATE
WEAR RATE
(mm/month)
8 -
7 FEEDONA.E. "T9
5 FEED ON A.E.
2
T3 B
1 T2 B
0 1
10 15 20 25 30 35 40
T I M E (min/day)
Note: Cast iron specification is the s a m e at all plants and T1 is Portland.
This relationship applies to all plants except T19, T15 and T16 which achieve about
three times better wear rate than the others. This is probably due to the fact that these
plants do not feed on anode effect when there is elevated temperature and major mixing
in the pot which would lead to higher corrosion.
5.4.2 Penetration Distance in Bath
Table 5-VI shows the penetration of plungers into the bath for average metal and bath
depths with n e w cathodes in 2 2 plants. It is indicative only, as each pot in a plant can
be different depending on liquid level, distortion of the cathode, or erosion of the
cathode. It gives a general trend of what penetration distances exist in different plants.
T A B L E 5-VI
PLANT PENETRATION COMPARISON

BATH METAL PENETRATION
PLANT LINE DISTANCE T O C A T H O D E ( m m ) (mm)
(mm) (mm)
T13 All 356 150 100 -100
T11 All 273 158 135 -80/+20
T17 All 273 170 65 -38
D13 All Vary bath level for penetration <0
T3 All 284 190 100 6
T15 8(P100) Vary bath level for penetration 12
T2 All Estimate 15
T6 3 274 190 100 15
D7 Prebake Move feeder vertical <20
T16 All 228 159 90 21
T6 2 268 190 100 21
D3 All 295 185 150 40
T7 All Estimate 30
D1 All Estimate 36
T1 All 295 172 165 42
T10 All Estimate 50
D12 All 260 170 150 60
T15 Vary bath level for penetration 64
T6 1 226 190 too 64
T4 All 250 200 120 70
T8 All Estimate 125
D9 D,E 210 175 190 150
D2 L4 210 200 190 150
D10 All Estimate 150
T9 P100 N e w Fixed 183 76 140
T9 T51 N e w Fixed 183 76 183
Note: "Fixed" refers to the plunger is at a fixed height relative to the base of the pot for all pots.
These plants target different bath levels for different penetration depths.
Penetration varies from - 1 0 0 m m to + 1 8 3 m m showing the wide diversity of operations.

If plunger wear does vary with penetration, there are m a n y opportunities for plants to
improve plunger life by reviewing this parameter.
M a n y plants with minimal penetration recognise its importance and targettotalliquid

level height so that the plunger does not get wet (or only by a minor amount). These
plants generally achieve very good plunger wear rates. Hence, it is possible to run a
plant on virtually zero penetration (on average).
Based on these data, trials were conducted of different lengths of plungers at Portl
1993. Initially a few plungers had 5 0 m m removed from the bottom edge before
installation. Then a number of pots with the lowest liquid base were converted to
shorter plungers. N o problems were experienced with blocked feeder holes.
Comparison of different plunger lengths in the same pot showed that 5 0 m m less
penetration reduced the wear rate of P A spec cast iron by 6 0 % , and 7 0 m m less
penetration reduced the wear rate by 8 0 % . Similar improvements were achieved for
other materials as well. For example, 5 0 m m shorter plungers improved life of cast
310SS by 1 5 0 % and 304SS by 3 0 % (Section 5.7).
These results showed very clearly the importance of penetration on plunger wear. Even
zero bath penetration often still breaks a hole as the crust bottom edge is higher than the
liquid bath. T h eriskis that bath level control m a y become unstable, liquid levels will
drop and plungers do not reach the crust. Each plant needs to explore zones of safe
operation over a long period of time to determine the optimum penetration depth.
Portland has seen at times a relationship between high anode effects and low liquid
level. O n e needs to control the bath to avoid blocked feeder holes.
It was decided to further reduce penetration at Portland from July 1993 by reducing
penetration from 4 3 m m to - 7 m m by using 5 0 m m shorter plungers. This change has had
no effect on blocked feeder holes, thus verifying the tests. Together with the change in
length, the plungers were changed to cast 31 OSS. These changes had excellent results
with 8 times the plunger life; a rise from 360 days to 3,000 days (Figure 5-12 and
Figure 5-13).
FIGURE 5 -12
PLUNGER WEAR RATIOS
(RELATIVE TO PA. SPEC CAST IRON @ STANDARD LENGTH)
WITHIN-POT DATA
(a) STANDARD LENGTH
(30Snim)
PA SPEC
INCONEL
HR SPEC
NIRESIST
HI CHROME (25-27%)
310SS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH

(255mm)
PA SPEC CAST
304SS BAR
"• •y^-yy "•<•• • &
310SS CAST
JKS CAST
NICROFER BAR
WORKSHOP DATA
(a) STANDARD LENGTH
(305mm)
PA SPEC
HR SPEC
HR SPEC (HEAVY)
NIRESIST
HI CHROME (27%)
HI CHROME(25%)
31 OSS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH

(255mm)
304SS
31 OSS
JKS
NICROFER
W E A R RATIO (relative to P A Spec cast iron @ standard length)

FIGURE 5 -13
CALCULATED PLUNGER LIFE
(RELATIVE TO PA. SPEC CAST IRON @ STANDARD LENGTH)
WITHEV-POT DATA
(a) STANDARD LENGTH
(305mm)
PA SPEC
INCONEL
HRSPEC
NIRESIST
HI CHROME (25-27%)
310SS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH

(255inin)
PA SPEC
304SS
31 OSS
JKS
NICROFER
WORKSHOP DATA
(a) STANDARD LENGTH
(305mm)
PA SPEC
HR SPEC
HRSPEC (HEAVY)
NIRESIST
HI CHROME (27%)
HI CHROME(25%)
31 OSS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH
(255mm)
304SS
31 OSS
JKS
NICROFER
1000 3000 5000 6000

PLUNGER LIFE AT PORTLAND (days)
From early 1994, additional tests were initiated on plungers which were a further 5 0 m m
shorter than the previously cut plungers. Hence, these plungers had an average
penetration o f - 5 7 m m for a potline. The trial pots had a penetration more like -100mm
as these were chosen as being the lowest liquid level pots in the potrooms. There were
no problems with these pots for m a n y months, but by the end of 1994, it was considered
that it was too risky to expand to full pot operation as it was likely that the liquid level
in all pots m a y need to be dropped in the future and there m a y be a danger at some time
in the future that there m a y be problems with blocked feeder holes. Hence, the trial was
stopped.
The trials of reduced penetration proved that crustbreakers with negative penetration
would operate successfully provided that the net distance between the plunger and the
crust was not going to drop over time e.g. if surface of the top of the cathode (bottom of
the pot) eroded. If the cathode generally rose due to the cathode "heaving" (swelling
upwards), then negative penetration would be a low risk method to improve plunger
wear rate.
A further attraction of less penetration is a reduction in plunger temperature due to

greater distance that the plunger tip is from the hole. A s will be discussed in Sections
6.3.6, every 1 0 0 m m of reduced penetration reduces the temperature of the plunger and
crustbreak cylinder by about 70°C. Cooler cylinders improve seal life which is one of
the main causes of crustbreaker cylinder failure.
5.4.3 Bath Temperature
Chemical corrosion is generally accelerated by temperature increase in either the metal

or the liquid/gas. A study of the temperature of the plungers at Portland showed a range
from 60-330°C existed (Table 5-VTI and Section 6.5.6). There w a s a median
temperature drop of 100°C at the plunger for 100°C drop in the pot, but the variability
hole-to-hole (S.D. 500°C) was so large that significance could not be seen across the
pots. In addition, the plunger temperatures for a single pot changed randomly over a
period of 3 weeks (Table 6-ffl). Most plants have pot temperatures averaging 950-
960°C, but temperatures as low as 920°C and as high as 1,000°C are possible.
t-
©
rs
CO
h-
o
CO Q n N 0 ) 0
in J5 CO to $ £!
OT-OOC*>T-
CNÔOOJCO
III
<6 &
3l M
S8 E
111 -c =0> S S
0. 1-2 in ;? o o ° in n m c N N N
t 88 CD
t ^ o> 93 m tOLoaicnoo
. M
w '
E
ui
go
x
CL
2 i-
ce
gco CO 0> ^- ,-, CD _. CO o IO IO CO
ui CC O pp
(9 O D- cu
O 2
D CO
-J r- CO ^ (OCM ^ in CN CO 1^ CD CN
§ 82
STD DEV
Ills
Z NUMBER
LOCn2
LOCn4
L0Cn3
L0Cn5
LOCnl
MEAN
E <
MAX
MIN
3 D
10 LU
w 5
-J CO •* CM O CO •* CM O
pa
H
CD ^f CM O
o N o in •* n cj IN o m ^ n « -
3
The frequency of high temperature periods for pots will cause extra corrosion. Hence
plants with large temperature variability m a y experience worse plunger life.
Plunger life at Portland was graphed against bath level but no trend was evident as th
was a great deal of variability in the data.
5.4.4 Plunger Preheat
Bath temperature affects the surface temperature of a plunger but has little effect on
conduction up the shaft (Section 8.4.2:2). The temperature of the plunger affects dags
(Section 6.5.7).
If the plunger is hot prior to entry into the bath, wear rate may be accelerated. Surv
of temperatures near the plunger hole showed a range of 60°C to 330°C (Table 5-VII).
The hotter the pot, the hotter the plunger. There was a large variation in individual
plungers. This variation in preheat adds to the variability of plunger wear within and
between pots. The temperature will change depending on the percentage of gas that
vents under the plunger. The less holes open, the hotter the plunger.
The shape of the plunger may also affect plunger temperature. A flat bottomed plunger
will heat up more than a conical worn plunger as the former will tend to create more
turbulent air at the tip which will promote heat transfer.
The effect of this variability depends on the type of plunger material. A highly passi
material such as 31 OSS or Nicrofer will probably be affected less than less passive
materials like cast iron. A s preheat changes by location, pot, pot design and pot
condition, no reliable testing has been conducted to identify the significance of preheat.
It is concluded that plunger preheat has some effect on plunger wear.
5.4.5 A n o d e Effect Control Strategy
Anode effects occur w h e n the alumina concentration drops below a bottom limit
(Figure 1-6). The bath starts to break down, large high resistance bubbles occur below
the anodes and the bath violently agitates. During this time, entry of the plunger into
the liquid will result in high corrosion rates. Agitation tends to remove any corrosion
products quickly, leading to exposed fresh metal. The higher temperature is an enabler
of high corrosion rates.
Anode effects can be controlled during this high agitation period by two methods (or a
combination of both).
(i) feeding the pot (to increase the alumina concentration), or

(ii) push the anodes into the bath (to squeeze the bubbles out).
The former will result in high plunger wear rates as generally several rapid feeds are
necessary. The latter will result in possible bath wash onto the top of the anodes which
m a y dissolve the cast iron connections where the rod meets the carbon anode block.
Iron dissolution will drop the quality of the aluminium product which is most
undesirable. The effect on metal quality m a y be small if old anodes are sufficiently
large, if the cast iron is properly poured into the join or if carbon paste is used as a joint
rather than cast iron. Each plant has a different philosophy on anode "kill" strategy.
Figure 5-11 shows that plants that do not feed on A.E. and have similar stationary tim
have m u c h longer life - about three times that of other plants that do feed on anode
effect.
The cost of plunger wear is not as significant as the benefit of killing anode effects
quickly, however, plunger life will suffer if it is decided to feed on anode effect. There
are m a n y plants that achieve very good anode effect frequency and do not feed on
anode effect e.g. Pechiney (0.1-0.3 anode effects per day) and D 3 (0.2 anode effects per
day). Thus it is concluded that plants should consider the possibility of not feeding on
anode effect in order to maximize feeder life and balance this against the risk to metal
quality.
CHAPTER 5 PLUNGER WEAR PAGE210
5.4.6 Plunger Mass
In February 1990, some 20 "long" H R cast iron plungers were installed in several
Portland pots (Figure 5-2 plunger B). These had the same distance from bottom edge to
cathode, however, they were taller. They had the gap between the top of a normal
plunger and shaft collar filled up (Figure 1-11). This increased the mass by 2 0 % yet
did not change penetration in the bath.
Of the 12 plungers traced, average wear rate was 4.7mm/month with an S.D. of
1.5mm/month. H R cast iron plungers placed in service at the same time averaged
5.5mm/month with a S.D. of 2.7mm/month. There was insufficient evidence to suggest
they were statistically different.
Hence, it was concluded that mass of the plunger does not affect wear rate.
5.4.7 Plunger Buildup
It seems reasonable that a coating of passive or hard material on the plunger will act as
a protective layer from corrosion (or erosion) as there is no flaking effect of scale (or
abrasion) against the crust. It is difficult to assess the effect of this variable as plunger
buildup (dags) come and go almost randomly (Chapter 6).
It was attempted to simulate the relative effect of erosion and corrosion by immersin
the plunger totally in the bath. In this way, there was no cycling through the crust. B y
comparing the wear rate to that of normal wear in the pot previously, it was expected
that one could estimate the effect of corrosion relative to total wear (erosion and
corrosion).
Two PA spec cast iron plungers were immersed continuously in the bath at a
penetration of about 7 5 m m . The plungers were withdrawn, measured, then
immediately immersed again. Initially this was done hourly, but as there was little to
no wear, the practice was increased to daily. This continued for 8 days; it was quite
surprising that the wear was so slow. The trial was conducted on another two occasions
on different pots with the same results.
The wear rate w a s about 1 0 % of the wear rate of plungers in the same pot
previously...0.4 m m / m o n t h versus 4.5mm/month. (Wear rate was calculated based on
an equivalent wet time of 2s/cycle and 160s/cycle.)
This experiment at first appears to give results that are opposite to the trend of F
11 which suggests corrosion is 7 0 % of wear rate. The plungers in this experiment m a y
be affected by no flaking from the outer boundary layer that occurs with normal
plungers that retract out of the hole each cycle. T 9 has analyzed the inner layer of the
material that falls off a cast iron plunger to be iron oxide...rust. In this Portland
experiment, this rust layer was not removed each cycle so the surface of the plunger
was protected from further attack.
This type of experiment was not effective in simulating plunger wear, but it does
illustrate that excessive times in the bath does not have any short term deleterious effect
on dissolution of a plunger. Note that long periods in the bath will cause heat
conduction up the shaft to the cylinder seals which m a y affect seal life (Section 8.4.2).
It is important during feeder changeout that the plunger is not dropped into the bath to
avoid this conduction effect. The same problem arises for n e w feeders. Pechiney and
D 3 have a bolt at the base of the plunger to secure it. Portland uses wires to secure the
plunger on n e w feeders.
5.5 TESTING EROSION THEORY
5.5.1 Plunger Speed
Portland changed the speed of its feeders in November 1991 but, at the same time,
increased dwell time due to the changes in pot pneumatics. Because of the two changes
occurring at the same time, it is difficult to separate the two effects. However, if one
makes the assumption that wet time has a linear effect on wear rate, then the frictional
effect can be calculated.
W e t time increased from about 1.4s to 3.3s; a 2.4 factor change. It would be expected
that the net effect of the wet time change would be 2.4 times the wear rate (using data
from Tables 4-1 and 4-H). This would be offset by a reduction in frictional wear
against the crust which would be expectedtohave a squared relationship between final
velocity and wear rate. The peak velocity became 0.6 times the original velocity when
stroke time increased from 0.9s to 1.6s. If friction is a linear relationship to the cube of
velocity, the friction change should be 0.2 of the original. The net effect of the 2.4
times change in wear rate from wet time and 0.2 times change due to the speed change
would be (2.4x0.2) = 0.5; a drop of 5 0 % .
Comparison of plant wear rates before and after the change is presented in Table
TABLE 5 - VIH
EFFECT OF PLUNGER SPEED AND DWELL

TIME ON PLUNGER WEAR RATE
CAST IRON STROKE OVERALL N E W POTS OLD POTS
DWELL TIME ALL DATA
SPEC SPEED MEDIAN ONLY ONLY
HR Fast 2s 5.5 5.0 6.6 5.3
HR Slow 4s 4.8 3.5 4.0 4.9
Statistically significant (90%)? Yes Yes Yes No
Difference (%) 14 30 39
Note: All values are wear rates in mm/month for plungers from production pots.
Most of the data indicate that the combination of slow speed/long dwell has improved
wear rate. The actual effect on median wear rates was a drop of 7 0 % which suggests
that the frictional effects (speed affected) m a y be rôce influential corrosion effects
A
(wet time effect).
(The exception to this comparison is for new pots which usually run at higher
temperature and higher bath levels than old pots, and the holes are generally much
larger. Hence, the main wear mechanism for plungers in n e w pots is corrosion, as
erosion is minor.)
5.5.2 Crust Hardness
Hardness of the crust is generally assessed by the fluorospar content of the bath. The
lower the fluorospar (CaF 2 ) level in a pot, the harder the crust. N o comparative testing
of crust hardness has been conducted in this study and no published articles refer to the
effect of fluorospar content on hole breakage or plunger wear. This is not expected to
be a big variable at the 5-6% fluorospar levels at which smelters usually operate.
5.5.3 Plunger Hardness
The effect of plunger hardness is best illustrated by controlled experiments of different

plungers as part of this investigation at Portland. Comparison of different materials
(which shows analysis, hardness and relative wear rates) is shown in Table 5-EX.
As discussed in Section 5.2.3, single batch trials were conducted at Portland using H
and P A spec cast iron. It w a s expected that, if erosion is a factor in the wear of
plungers, a harder material would be beneficial. The hardness of the P A spec plungers
was 3 4 6 B H N versus 2 0 8 B H N for H R spec. P A spec has a statistically much higher
hardness, yet it m a d e no difference to wear rates viz. 1.2-8.9mm/month versus 1.1-

7.8mm/month respectively. The variability was very large for both materials in these
tests as these were scattered over m a n y pots. N o n e was tested in controlled experiments
in the same pot for the same time. However, Section 5.6.3:1 does show results in
controlled experiments that there w a s a 4 0 % drop in life for the harder P A spec
compared to H R spec which supported these plant trials.
Trials of Inconel pipe showed wear rates 7 times lower than cast iron despite having
hardness of 1 3 0 B H N versus 1 7 0 B H N for H R spec and 3 5 0 B H N for PA. However,
trials of high chrome cast iron with a hardness of 4 5 0 B H N and 6 0 0 B H N showed wear
rates 2.4 times less than P A spec.
These conflicting results suggest that hardness is not a significant indication of p

wear rates.
CHAPTERS PLUNGER W E A R PAGE214
TABLE 5 - IX
PLUNGER MATERIAL COMPARISON
ALLOY
1\SSUME CAST LENGTH O.D
BRINELL COST/ COST/ WEAR RATE RELATIVE EXPECTED
Ni Cr HARDNESS PLUNGER CONVERT RELATIVE TO LIFE F O R PLUNGER LIFE
/BAR (mm) (mm) (%) (%) (BHN) ($) ($) 05mm PA SPE W A L L O.D. months) (days)
C A S T IRON
HR S P E C (COMPLETE) CAST 305 100 0 1 170 45 1.4 1.4 17 504
PA S P E C (COMPLETE) CAST 305 100 0 1 350 40 1.0 1.0 12 360
PA S P E C (COMPLETE) CAST 253 100 0 1 350 40 16 16 19 576
HIGH C H R O M E
2 7 % Cr (COMPLETE) CAST 305 100 0 27 600 100 2.4 2.4 29 864
2 5 % Cr (COMPLETE) CAST 305 100 2 25 450 100 2.4 2.4 29 864
304
COMPLETE BAR 305 100 9 19 155 255 3.3 33 40 1188
BIMETAL TIP BAR 305 100 9 19 155 110 52 3.3 3.3 40 1188
BIMETAL TIP BAR 2SS 100 9 19 155 80 52 42 4.2 50 1512
310
COMPLETE CAST 305 100 20 25 120 2.8 2.8 34 1008
COMPLETE CAST 255 100 20 25 80 8.3 8.3 100

COMPLETE ASSUME BAR 305 100 20 25 170 315 5.6 5.6 67
NIRESIST D4
COMPLETE ASSUME CAST 305 80 30 5 235 102 1.5 1.5 18 547
COMPLETE CAST 305 100 30 5 235 165 19 19 23 684
BIMETAL TIP CAST 305 100 30 5 235 145 90 1.9 1.9 23 684
BIMETAL TIP ASSUME CAST 255 100 30 5 235 105 90 2.5 2.5 30 889
INCOLOY 800HT
COMPLETE ASSUME BAR 305 100 32 21 240 435 39 3.9 47 1410
BIMETAL TIP BAR 305 100 32 21 240 270 52 3.9 3.9 47 1404
BIMETAL TIP ASSUME BAR 255 100 32 21 240 180 52 5.1 5.1 61 1825
JKS
150 9.1 91 109
COMPLETE CAST 305 100 30 26
EH
NICROFER G3
COMPLETE ASSUME BAR 305 100 40 22 240 480 11.7 11.7 140
BIMETAL TIP ASSUME BAR 305 100 40 22 240 300 52 11.7 11.7 140
BIMETAL TIP BAR 255 100 40 22 240 200 52 15.2 15.2 182
INCONEL
145 2.6 2.1 25
™ 760
COMPLETE ASSUME CAST 305 80 55 11 170
COMPLETE CAST 305 100 55 11 170 235 3,3 3.3 40 1188
BIMETAL TIP CAST 305 100 56 11 170 205 90 3.3 3.3 40 1188
BIMETAL TIP ASSUME CAST 255 100 55 11 170 140 90 4.3 4.3 51 1544
INCONEL 601
PIPE(40SCHED) PIPE 325 55 60 23 130 180 50 7.0 1.1 13 396
PIPE(80SCHED) ASSUME PIPE 325 73 60 23 130 360 50 7.0 2.2 27 806
Notes.
1. 8 0 m m castings have 14mm walls.

Z 100mm castings have 2 4 m m walls.
3. Inconel (40 shed) p p e has 4 m m wall.
4. Expected Hfe based on 1993 Portland life of 12 months.
5. Bimetal plunger has a 150mm steel top half.
6. "Relative life for wall O.D." considers the wall thickness and O.D. assuming linear wearrateviz. an 8 0 m m O.D. has 8 0 % of the life of a 100mm O.D. plunger.
7. Cost of maching tip of bimetal plunger is $40 except for Niresist which is $55
8. Assume 255mm C A S T STAINLESS materials have 2.5 times longer life than 305mm based on 310SS experience.
9. Assume 255mm B A R has 3 0 % longer life than 305mm based on 304SS experience.
10. Assume 305mm tip and C O M P L E T E have similar wear rates viz. based on 304SS and Nicrofer experience.
5.5.4 Plunger Shape
In integrated feeder plants, it is c o m m o n practice to use flat beveled shaped plungers of

similar shape to plunger C in Figure 5-2, but most independent feeder plants use flat
bottomed plungers. The reason for the tapered shape is the belief that the taper acts like
a wedge to crack the hard crust. However, there are several potential benefits of flat
bottomed plungers:
(i) less bath sticking due to smaller surface area,

(ii) larger hole due to "fracturing" rather than "drilling" a hole,
(iii) less chance of sticking in the crust,
(iv) less chance of heat retention due to smaller surface area touching the liquid
bath, so less plunger buildup,
(v) less chance of the plunger shaft tracking off centre which causes uneven and
premature seal wear, and
(vi) higher vertical component of pressure.
It is difficult to establish if these benefits have been attained as they are difficult
measure. It is difficult to find a reason to retain the tapered tip of the old design.
Comparing wear rates before and after the tip change showed no statistical difference
(Table 5-X).
TABLE 5 - X
EFFECT OF TIP SHAPE ON PLUNGER WEAR RATE

CAST NEW OLD
STROKE DWELL TIP OVERALL
IRON ALL DATA POTS POTS
SPEED TIME SHAPE MEDIAN
SPEC ONLY ONLY
HR Fast 2s Bevelled 5.1 4.5 5.9 4.9
HR Fast 4s Flat 5.6 5.0 6.8 5.4
Statistical y significant (90%)?
^Q!^|
Note: All values are wear rates in mm/month for plungers from production pots.
If there is no wear rate benefits for tapered plungers and there are several operational
benefits to flat bottomed plungers, it is concluded a flat bottom is the preferable shape
for plungers. T3 changed to flat bottomed plungers in 1988 and Portland changed in
December 1990.
5.6 PLUNGER MATERIAL
5.6.1 General Smelter Experience
There is a wide range of materials used for plungers in smelters. Table l-II shows what
materials some 24 non-Alcoa plants are using for plungers. This is in addition to the
nine Alcoa smelters which have used cast iron (until Portland changed to cast 31 OSS in
1993). There are over a dozen different materials used with cast iron being the most
common in integrated feeder plants and stainless steel (of various types) is common at
independent feeder plants. Several plants have made material changes over time as they
have experimented with different types.
Plants T4, T5, and T14 have recently changed from cast iron to 304 stainless
(304SS). T8 changed from 304SS to mild steel (which had an even worse life than cast
iron). T4 chose not to go to anything more exotic than 304SS (from cast iron) as they
felt that they could not guarantee that the other feeder components would last longer
than three years, so why waste money on a plunger that one will probably throw out
anyway after three years? This logic is very sensible. It all depends on one's confidence
that one knows the life of the other components in the feeder. Good tracking systems
and proactive maintenance teams are needed to remove the causes of other failure
mechanisms.
D3 changed from cast iron to Inconel then to alloy coating and are now looki
alternatives. Their problem was probably too frequent a break cycle (under 2 minutes)
and using alumina as anode cover (which is abrasive). They experienced erosion of the
softer alloys. They achieve 2-3 years life, but want to get over 5 years.
CHAPTER 5 PLUNGER WEAR PAGE217
In 1987, Pechiney discussed plunger materials at their "Club 180" meeting for users of
their A P I 8 (180kA) pots. At this meeting their plants had the following materials
(some are trade names).
D2 AISI309
D6 Z-12-CN
D9 309SS
D10 310SS
D14 N S 24
D15 310SS
D10 are investigating alternative materials for Lines 1 and 2, and are using 310SS
the n e w Line 3. D 6 is changing to a high nickel bar material made by a local
manufacturer (Acier Abrasion) with good results. This is the same material used for
anode setting crane grabs. The nickel contents of 309SS and 310SS are 12-15% and
19-22% respectively.
The fact that the Pechiney plants are moving away from 309SS and 310SS suggest that
these are not good enough to match the 5-6 years they are achieving for their
crustbreaking cylinders or 7 years for pots. D 9 believe they achieved at least three
times the life of 309SS when they changed to Nicrofer. The newest line uses Hastelloy
X mainly due to economics at the time. (Nicrofer has 4 0 % Ni and Hastelloy X has
4 5 % ; both have 2 2 % Cr). The results appear to be at least as good as for Nicrofer.
The most recently commissioned Pechiney lines (e.g. D4, D8, D10) have used "310S"
stainless steel instead of the standard 310SS materials. The "S" refers to a drop in
carbon from 0.2% to 0.08% which is believed to reduce the chance of carbon migrating
to grain boundaries and making the nickel less passive to corrosion attack.
5.6.2 Kaiser Plunger W e a r Experiment
In 1984, Kaiser published a paper on an experiment into plunger wear (54); this is the
only published data traced on this topic. Westerman and Harrison tested a range of cast
alloys and steel in a test pot which had three hexagonal plungers. These had strips of
the alloys welded around the plunger so a comparison could be made of alternative
materials under exactly the same pot conditions. Table 5-XI has the results of the
study.
T A B L E 5-XI
KAISER PLUNGER WEAR TESTS

ALLOY Ni
Cr RELATIVE COST PER COST PER RELATIVE RATIO
LIFE TIP TIP-DAY COST/DAY Ni/life
HASTELLOY 45 22 23.0 243 0.07 0.18 2.0
309 SS 12 26 6.0 105 0.14 0.35 2.0
304 SS 9 19 4.6 48 0.11 0.28 2.0
13-4 SS 4 13 1.9 - - _ _
CAST IRON 0 1 1.0 30 0.4 1.00 n/a
Notes: (i) All materials were cast (not bar).

(ii) "Ratio Ni/life" is calculated as follows: (Ni content)/(relative life)
e.g. Hastelloy X is calculated as 45/23.0 = 2.0
(iii) All cost values in U S $ .
(iv) Except for "Ratio Ni/life" all data from Westerman and Harrison (5
The conclusions from the paper were that chrome over 2 0 % and high nickel were
beneficial to plunger wear. However, they did not notice a more significant trend in
their data. There was a linear relationship between nickel content and relative wear
rate.
The relationship between the life (compared with cast iron) and the nickel content (%
was linear with a correlation coefficient (r2) of 0.99!; 9 9 % of the variability of plunger
wear rate was due to nickel rate alone. A 4 6 % nickel content had a life 23 times the life
of cast iron, 1 2 % nickel had 6 times the life and 9 % nickel had 4.6 times the cast iron
life. Clearly, life increases by half the nickel content rise. If one doubles the nickel
content, life is doubled. A s a corollary, this suggests that one can choose the nickel
content of the plunger to provide the desired life.
If one uses the Kaiser results as a datum and the standard deviation found at Portland
(i.e. about 3 0 % of the average (Tables 5-II and 5-IH)), one could use the following
logic to determine the optimum plunger nickel content:
Target: 95% of plungers last 5 times that of cast iron.

Average nickel content:
twice the ratio to cast iron. 2 x 5 = 1 0 %
Variability:
Assume that S.D. for the critical ingredient (nickel) is similar to
3 0 % of the average wear rate. Assuming a normal distribution, a
9 5 % confidence limit twice the S.D. below the average would be
the lower limit of nickel content. Hence, the target should be:
10+(2x0.6xl0)=22% minimum.
Chrome content:
Assume Kaiser conclusion of 2 0 % minimum.
Hence, final deduced analysis for a plunger 5 times the life of cast iron at Portlan
would be 2 2 % Ni and 2 0 % Cr (minimum). Portland trials of cast 310SS ( 2 0 % Ni 2 5 %
Cr) gave 3.3 times the life of P A spec cast iron (Table 5-Vin). Extrapolation from
2 0 % to 2 2 % Ni would give an average life of about 4 times the life of cast iron. Hence,
this prediction calculation is a reasonable estimate for plant scale operation.
Close consideration of the Kaiser report shows that their test may not be completely
similar to plant conditions. In the test, the m a x i m u m length of service was under 80
days and the dwell time was kept at 6s to accelerate the trial. This longer dwell time
tends to make the test one of the ability of the plunger to withstand corrosion only. The
effect of erosion and the erosion/corrosion interaction are not tested. Hence, the metals
with good corrosion resistance m a y be seen as the best. Also, the short length of the
test m a y question validity for plant scale performance.
Based on comments from representatives of several plants which have changed plunger
type, the following relationships were evident:
Portland cast iron to cast 31 OSS = 3 x original

T14 cast iron to 304 = 3 x original
D9 309 to Nicrofer = 3 x original
T8 steel to 304 = 2-3 x original
Thus, Portland trials and plant data tend to support the Kaiser experiment data (if not
the Kaiser conclusions). The data suggest that increasing nickel content reduces
plunger wear rate, but that a 2 0 % chrome content is also needed.
5.6.3 Portland Experience
O n e can see that the industry is in a continual state of flux with respect to plunger
materials. Often these changes are m a d e on limited trial data and the changes are
generally on a full scale plant basis. This study considered alternative materials in a
controlled manner using statistical techniques and a reliable tracking system.
Portland has conducted trials on 19 different materials to date as shown in Table 5-XH
Using data from this table, Figure 5-12 illustrates the relative wear rates of different
plunger types, and Figure 5-13 shows life based on within-pot trials; same pot, same
time, same original size.
Below is a discussion on the different materials used for the plungers tested in this
investigation. Section 9.3 examines the economics of these materials and designs.
5.6.3:1 Cast Iron
In summary, the wear rate of H R is 4 0 % better than P A despite being half the hardness.
Based on recommendations from Alcoa USA, cast iron plungers of 21%, 25% and 27%
chrome were tested at Portland. Overall, the wear rate of the high chrome cast iron
CHAPTER 5 P L U N G E R W E A R P A G E 221
TABLE 5 - X H
PLUNGER MATERIALS TESTED AT PORTLAND
CAST IRON HR SPEC

PA SPEC
BRITISH ALCAN SPEC
21% CHROME
25% CHROME
27% CHROME
ALLOY COATINGS AI2369
AM 721
A H 795
AM 236
AI2354CO
STAINLESS STEEL 310 CAST

304
INCONEL PIPE
INCONEL CAST
INCOLOY 800HT
NICROFER G 3
JKS
SILICON CARBIDE
materials was 7 0 % better than P A spec and about 5 0 % harder. Some fracturing
occurred from temperature shock or brittleness, but the main problem was plunge
buildup (dags). At Portland, it was found that there was almost 3 times more li
of having dags on high chrome plungers than any other plunger type (Section 6.3
Some 60% of multidag feeders were high chrome, despite only having 13% of these
types in operating pots.
Due to the hazards experienced with dags, high chrome plungers are not recommended
despite the better performance in wear rate. Cast irons were found to be inferi
almost all other materials with respect to wear rate and cost/day (Section 8.3)
not recommended.
5.6.3:2 Inconel
Inconel was selected for trial as it was a high nickel-based material. There was concern
that it m a y be too soft (half the hardness of cast iron), but it had a proven record at SI
on some parts of Soderberg pots. The design used for trial from February 1991 was a 2
inch pipe ( 6 0 m m O.D., 5 2 m m I.D.) with a plate crimped onto the base. Unfortunately,
the plate fell out within 3 months after going into service. This exposed both sides of
the pipe to attack and left the outer edge to peel back as it hit the crust. This led to
attack of the steel plunger shaft underneath. This is illustrated in Figure 5-2 plungers
Fl, F 2 a n d F 3 .
Despite these setbacks, the performance was very good. After seven months of
operation, the wear rate on one was only 3 m m (or 0.4mm/month). Cast iron wear rate
was about 7 times that of the Inconel pipe. W e a r rate was very low for thefirstyear
until the steel inner shaft was exposed. If one used a thicker walled pipe, the life would
be m u c h longer.
In order to try a cheaper alternative, a cast version of Inconel was tested. Cast Inc
had a relative wear rate 3.3 times better than P A spec. Although twice the wear rate of
pipe, the wear rate was still three times better than P A spec cast iron (Figure 5-11).
The cast material had a life about half that of similar analysis bar material. This b
life m a y be due to the quality of the metal grain structure relative to a cast material.
5.6.3:3 Niresist and Incoloy
Because of the cost of Inconel and the difficulty of purchasing bar in Australia, a cast
material of high nickel was tested. Cast Niresist D 4 has a 3 0 % N i content but only a
5%Cr, so was not strictly what the Kaiser tests recommended. It has considerable
carbon content (2.6%) which m a y migrate to the gain boundaries and m a y not be
attractive for corrosion efficiency. However, it was easily castable and quick to trial and
m a y indicate the effect of a compromise between Inconel and cast iron.
Ni-resist achieved 1.9 times the life of P A spec cast iron. Although this was significant,
this was not high enough to consider it further. The lower life m a y have been due to
the low chrome content.
A bar equivalent to Niresist was sourced. Incoloy 800HT had attractive temperature
and corrosion statistics, so a sample of 20 plungers was placed in service in 1992.
Relative wear rate of Niresist was 1.9 versus 3.9 for Incolov. Incoloy proved to be
twice as good as Niresist, once again confirming that bar is superior to cast for a similar
chemical analysis.
5.6.3:4 Stainless Steels
A material 4 times better than cast iron w a s found (Incoloy), so a significant

breakthrough had been achieved. However, this was still not good enough to get a
material where A L L plungers achieve over 5 years. A n average life of about 7 years
was required. The pattern so far suggested that the high nickel was attractive and that
bar gave twice the life of cast materials. Hence, study continued using high nickel
stainless steels.
Most stainless steel bars are difficult to get in Australia as most are only made ov
and costs per plunger are high. 304SS is the cheapest ($255) and easiest to obtain, with
the cost increasing with the nickel content to about $480 for Nicrofer (versus about $40
for P A spec cast iron). Based on successful experience in other smelters, 304SS was
tested at Portland with original and 5 0 m m shorter lengths from mid 1993 (Figure 5-2
plunger H ) . It was later established that the original 3 0 5 m m length gave a life 5 times
that of P A cast iron.
At the time of commencing the 304SS trial, it was decided that action should be taken
to change normal repair plungers to a high specification design in July 1993 viz. cast
310SS. It was judged that the relatively cheap cost ($120) m a d e it a good economic
risk. In addition, following testing on reduced penetration (Section 5.4.2), it was also
decided to use 5 0 m m shorter plungers from July 1,1993. This dropped the cost to $80
making the changeover even more economic.
The effect was a life 8 times that of P A cast iron which achieved not only a better life
but cost only 3 0 % that of a 304SS bar plunger. This illustrates the benefit of cheap cast
materials (with less penetration) being very cost effective compared to traditional
designs using high specification bar material. (Section 9.3 discusses a bimetal tip that
halves the cost of bar materials to make them more cost competitive.)
After discussions with D9 who had good experience with Nicrofer, a trial was
introduced in November 1993 with 5 0 m m shorter plungers. This material was an
extension in the trials to date as it had higher Nickel, a good chrome content and a
proven plant record. However, no comparison tests had been conducted at D 9 and this
was an independent feeder plant which m a y achieve better results due to lower wet time
rather than better material. It appeared to be a good risk. It was very difficult to obtain
for trial quantities desired, so about $80,000 of material had to be purchased for 400
plungers. T o offset the cost, a bimetal design was jointly developed by Hans K e m p e
and the author to get the m a x i m u m number of plungers from the expensive material.
Life 15 times that of PA spec cast iron was achieved. It is clearly the best material
tested if the target is m a x i m u m life. Even though the initial cost is high, it will be seen
that the cost/day is still economic (Section 9.3).
It is expected that an average life of 15 years can be achieved with Nicrofer. This l
m a y be a little too long, as too long a life m a y result in disposal of a significant amount
of the costly material on failure of other feeder components. That is, if other major
components fail and result in the feeder being returned to the workshop, it m a y not be
prudent to return a plunger to the plant knowing that the plunger m a y be the reason for
the next failure in under a 5 year period. Hence, the old plunger will be disposed of,
thus throwing away a considerable investment. The challenge of feeder design is to
have all parts fail at the same time. Too long a life on some components m a y in fact be
bad economics. If a plunger shaft breaks, for example, the plunger m a y fall into the pot
and cannot be dragged out as it welds itself to the bottom of the pot and does not
dissolve. Hence, the pot becomes more unstable.
T o investigate a cheaper alternative to Nicrofer and 310SS, a high grade cast material
( 3 0 % Ni, 2 6 % Cr, 5 % M o ) was invented by the author with assistance of K e n Deans
from Backwell/IXL (the suppliers); called JKS...the "Jim Kissane Special". The
analysis was based on the highest nickel content that can be easily cast with a minimum
of 2 0 % C r and some Molybdenum as used in Nicrofer and Incoloy 8 0 0 H T for grain
strength.
This was introduced in March 1993 on trial feeders. JKS wear rate was 9.1 times
better than P A spec.
This is comparable to Nicrofer but considerably cheaper to manufacture viz.: $150

versus $300 for a full length cast plunger of J K S versus bimetal tips of Nicrofer
respectively. It is also m a d e in Australia rather than fully imported as is the case with
(bar) 304SS, 309SS, 310SS and Nicrofer. It is probably the most cost effective plunger
design tested to date and is not too long a life (3,000 days) (that the investment does not
take excessive periods to achieve a return). Life is also adequate to achieve an average
life of about 9 years, which is sufficient so that few plungers will fail under 5-7 years.
5.6.3:5 Silicon Carbide
Silicon carbide has been trialed at Portland on two occasions (76). These were
rectangular with chiseled sides and a taper at the end for structural support. Ceramics
are stronger and easier to m a k e if m a d e rectangular in shape. There is no reason w h y a
plunger needs to be round.
In the trial, the plungers fractured half way down their length after a maximum of 2
weeks (Figure 5-2 plunger D ) . S o m e lasted less than a day. Analysis of the material
showed it was only 6 0 % as strong as the specification. Another trial was carried out
where the material w a s checked at independent laboratories by the manufacturer
(Carborundum) and by Portland Aluminium ( C E T E C ) to be above the specification

(Figure 5-2 plunger E).
Once again, the plungers failed within days. This material suffers from fracturing, as
silicon carbide is brittle. They break easily if hit sideways. Trials at D 1 0 also failed
even with a stroke time of 3.0s compared to the 1.6s at Portland. This material is not
recommended.
5.6.3:6 Alloy Coated Cast Iron
Four types of alloy coated cast iron were placed in service in parallel with other trial
plungers in February 1991 (Figure 5-2 plunger C). N o n e achieved m u c h better results
than cast iron (75,76). S o m e of these alloy coated materials had the coating shed off
the plunger within months, possibly due to different values of expansion coefficients.
A further trial was conducted on one cast iron plunger coated with Inconel but further
trials were not conducted due to the better performance of bar materials.
T3 are having some success with a Castolen coating. D3 have decided their alloy
coating has not been successful and are searching for a n e w material. It is the belief of
the writer that alloy coatings are a "Band-Aid" treatment and are not thick enough to
ensure long life. Hence, coated plungers are not recommended.
5.7 CONCLUSIONS ON PLUNGER WEAR MECHANISM
In the foregoing discussion in Section 5.2-5.6, it has been established that there are
many possible contributors to the large variability in plunger wear rates (Table 5-1)
which cause the large scatter in wear profiles in Figure 5-3. There were two theories on
the wear mechanism - erosion and corrosion. A few general observations m a y be useful
to focus on the critical discoveries from the m a n y tests conducted in this study at
Portland.
The results of this research are summarized in Table 5-XIQ.

ui
a. o
£
?
DC
< 2
z j"
5 g o
CM
Rz
0- Q_ Q- o
o 03
111
3 3 3 8 8 8 z
r- 8 E ro
O
Mi £
nc 2
ui o CD Q.
ro o
tr
EC |U
Q
< LU
a.
3
i
z
23 0_ 0- o Q g Ji tc
< z
ui 3 3 a
o 8 8 8° UJ
LU
3
2
<
z
z h-
«t 0- N (0
2 3 > £
X
1 ®
o
to
8
CO
_1
J? ra
ra-"
CO <
o o o += o
2 cr 2 |
o < o 3
1 15 CO CO
E s ra o to
£ to < to
I-
o
o
0
rr
T
5
Li.
CT1 J.
a 1— 2 ac 03
III
i-
1-
£ rr
LU ti
o LU o O E o c
z 2
<
m 3 z
rr co O 0. a T 1-
G
i_ a>
ro Q-
o
^
8
0-
< 0- to
o a. o
0- i- 0. Q co n LU
I i
n (fl =
(
CO/) fCO/1 _i
o
H ro 7
t/i X 5 rr
s z z o O _l1
d) ra
U u fl n i- LU OC LU ra 2 CD
in (
n rr)
I i-
II" 0C fl rr o I <r ifi
O U CO in 5> vn X o -Q Q
Ul -I CM" CM in
K ID in ° " iA p in in
SfiS
3 < ;2> ;S> d) lA 8.5 ro
£ IT
9 >- E ro E
: ! o
o > c
T- CN CO E 8 io
iO in no o »= a
U
a, Ul
LO m LO
l«s
CO *± ta <-
ta
JS ra CD
O < z LU
I o
® S ra
tn LT 0- CO
ro
CO
i- a c
UJ LU V- rr < 1-
CO D "• ro-2
n 3 co in
H Of rr 3 r- 3 ILU
z
8
CO Q ni 7 S S |
U
Ul
I-
Ul
co
h- a
III
2
o
£ LU
OC
<
5
_j
3
CQ
to
It
z
n
< *
n
IY o ro o
E
co LU
U- > < 1- LU
•>
aIT LU OC OC LU J. tE^
w 3 MC n
LU LU U. LU LU
OC
ta
<
a.
oc
y 55 5-) OC
LU
r-
H
I
3
Ii_
LU
LU
z
3
o 3
i-
co
3IT
LU
°«#
(4 5 _l 0. 3 R E o
LU 0C LU 0- n o _J "- CD 05
CO 3 0. u 2 - CD
\~ O c <£ <n
co ro E co
•e o x=
3
OOC
o
III
to CD CD ?
z o 7 7 7 7 7 7 7 7 t- Z E
o rx o o o O O a o O 7 7 z z
Ul LU 7 7 7 7 7 7 2 7 0 o o o
z 0 O O 7 2 z z
10
o
o o
to ro
o o
CO ro CO o o
ro CO ro o U
oo o
3 CO
O
I ( r>
rr) rv ()
It
O O
IY rr
O () o
rr rr OC
rr
ro r o t
() OIY o
to
o
z rr rr OC rr rr rr rr rr rr rr
DC LU LU LU UJ
OC g
CO
8 8 8 8 8 8 8 8
o
In order of priority, the most important parameters to control plunger wear were:
1. material (higher Ni/Cr the better)

2. penetration depth (less the better)
3. wet time (less the better)
4. plunger buildup (more the better)
5. plunger speed (slower the better)
Material and penetration are much more important than all the others. It was also
concluded that the wear mechanism was mainly corrosion, with a lesser effect from
erosion and that high nickel alloys with over 2 0 % C r are the best plunger materials. Bar
materials give double the life of the equivalent cast analysis plungers.
5.8 MAJOR FINDINGS FROM PLUNGER WEAR ANALYSIS

(i) Plunger wear was widely variable from pot-to-pot and within a pot.
(ii) W e a r rates of batch-to-batch cast plungers were statistically in control.
(iii) Increasing the frequency of feeding increases wear rate.
(iv) Plunger shape (flat bottomed) and mass do not affect wear rate.
(v) W e a r rate was the same wherever the plunger was located in the pot.
(vi) Reduction in speed improved wear rate.
(vii) Silicon carbide was brittle and had a very short life.
(viii) Optimum material appeared to be a nickel content of over 3 0 % and chrome
content over 2 0 % .
(ix) Bar material appears to have twice the life of the equivalent cast material.
(x) Nicrofer and J K S materials achieved the longest life.
5.9 RECOMMENDATIONS FROM PLUNGER WEAR ANALYSIS
(i) Minimize dwell time to that required to release the alumina and control the other
factors that cause blockages, rather than operate with an excessive dwell time
"just in case". Check occasionally that the dwell time has not been increased.
(ii) Investigate alternative materials for plungers only after causes of plunger wear
have been addressed.
(iii) Control liquid level so as not to get the plunger wet. This m a y require making
the plunger shorter so there is less (or no) penetration in the liquid.
(iv) Measure plungers into the workshop to monitor plunger wear.
(v) Tracking system should include pot number and location in the pot to tie in pot
conditions with feeder life.
(vi) Compare plunger materials/designs within the same pot for the same period of
time starting at similar sizes to measure relative wear rates
(vii) Install Nicrofer bar or cast J K S plungers for optimum plunger life.
CHAPTER 6 PLUNGER BUILDUP ANALYSIS PAGE 230
CHAPTER 6
PLUNGER BUILDUP ANALYSIS
6.1 SUMMARY
In October/November 1993, Portland conducted a feeder survey in order to determine

the cause of dagging (buildup on plungers) for 1,000 feeders.
It was found that high plunger temperature was the main cause, as well as high chrom
cast iron plunger material. High chrome stainless steels bar plungers had minimal
buildup problems. Dagging was promoted by high liquid levels and proximity of the
flame to the plunger. There was no relationship between frequency of dagging and
anode effects. Removal of dags had no effect on the frequency of dags returning.
Actions recommended to reduce dagging included no further use of high chrome cast
iron plungers, trials of shorter plungers, remove dags only if the dag fouls the feed
chute under the superstructure and remove the cause of dagging, not work on the effect.
6.2 INTRODUCTION
Buildup of material on plungers is called a range of names which include a dag, dough
ball, hornet's nest or ball (Figure 6-1).
FIGURE 6 -1
DAGGED PLUNGER
It is hazardous for operators to remove dags as they have to lean into the pot with a
large crow bar to knock it off the plunger. This exposes operators to dust, heat and
fumes and they are at greater risk of being splashed by molten bath. The action of
knocking off the dag can lead to back injuries due to the awkward procedure. The dags
often return after removal, making controlling dags a frustrating, as well as a hazardous,
job. Dagging can suddenly become frequent on m a n y pots then reduce again after a
few days. Operators note that dagging seldom occurs on all feeders in a pot, except for
n e w pots (less than 30 days old) when dagging often occurs across the pot and the dags
are very big ( > 4 0 0 m m diameter). Pot operators also observe that the feeders in the
middle of the pot dag more frequently.
For integrated feeders, operators must remove the dags if they affect the plunger from
fully retracting. The dosing unit on integrated feeders is activated by the same cylinder
that moves the plunger. If the plunger does not fully retract, the dosing unit m a y
remain open causing alumina to flow continuously unmetered into the pot making it
mucky. Conversely, if it does not retract to its top limit, the dosing unit m a y not refill at
all, causing the pot not to be fed and anode effects will result. Usually the dosing unit
will be affected if the dag fouls the chute under the superstructure.
Independent feeders are not affected by dags except if the dag fouls the chute in such
w a y that the alumina dose is deflected to the side of the hole in the crust. This is rarely
the case. Dosing accuracy is not affected by dags in independent feeders as the
crustbreaker is separate from the dosing unit.
O n Pechiney and most other independent feeders, there is a dag cleaner for continual
cleaning of the plunger. Refer to Figure 1-12 and 1-13 which show the dag cleaner at
the base of the feeder around the plunger. These cleaners are relatively successful if the
cylinder is 1 5 0 m m O.D. or bigger. Below this diameter, there is insufficient force to
fully retract the plunger if there is buildup. Even for the 2 0 0 m m cylinders used at D 9 ,
the plungers still get stuck if the pot is n e w or the bath level is high (which both cause
dags).
For m a n y pots with smaller cylinders, one can often see trials of dag scrapers which
have been unsuccessful. M a n y trials were conducted at Portland before the dag cleaners
were raised to above the normal high bath level. Since then, there has been no problem
with dags and feeder faults. The real solution is to stop the creation of dags rather than
use questionable designs to correct for an operating fault.
Possible causes of dagging are:
(i) the plunger is too hot to allow the liquid bath to freeze and fall off,
(ii) the bath chemical composition is low in aluminium fluoride causing material
to freeze, and
(iii) the surface of the plunger is rough causing the material to "key" into the
surface perturbations.
S o m e people believe that dags cause anode effects due to restriction of flow of alumina
into the hole in the crust. A n investigation was conducted to determine what causes
dags and develop ways of preventing them from occurring.
A survey on 200 pots was conducted during October/November, 1993 to ascertain

which feeders had dags and what relevant pot conditions existed at the time. Pot
attendants logged, every time a dag occurred, the
(i) pot and feeder location of any dag,

(ii) size of the dag, and
(iii) if the dag was removed.
Table 6-1 has a summary of the observations. As it was clear that there was a
relationship to the temperature of plungers, plunger temperatures were measured for a
range of different pot conditions in a separate trial.
This Chapter firstly discusses the dag survey results, explains the investigation into
plunger temperatures, then discusses h o w to stop dags from occurring.
TABLE 6 -1
DAGGED FEEDER SURVEY SUMMARY
(29 Oct 1993 - 12 Nov 1993)
NUMBER OF POTS SURVEYED 204 (1040 feeders)
NUMBER OF DAGS 43/day (4% of feeders)

BIG DAGS (>400mm) 26%
NUMBER OF DAGS REMOVED 10%
FEEDERS WITH MULTIPLE DAGS 3% of feeders (34% of total dags)
POTS WITH >1 DAG/DAY 4% of feeders (25% of total dags)
POTS WITH NO DAGS EVER 45%

FEEDERS WITH NO DAGS EVER 81%
DAO DISTRIBUTION BY LOCATION OF FEEDER IN POT
DISTRIBUTION OF DAQS
BY PLUNGER TYPE
DISTRIBUTION OF
PLUNGER TYPE
HIGH CHROME
11% HIGH CHROME
S0%
POTS THAT HAO ANY DAGS FEEDERS THAT HAD ANY DAGS
DURING OBSERVATION PERIOD DURING OBSERVATION PERIOD
NO DAGS EVER
«%
NO DAOS EVER
91%
6.3 SURVEY RESULTS

6.3.1 D a g Frequency
Dags occurred on 4 % of feeders per day but 8 1 % of the feeders did not have one dag at
any one time in the two week period. O f the pots that had dags, only 4 % had more than
one on a pot in any one day. Operators suggested the frequency of dagging was about
what is commonly seen. Dags tended to come and go apparently randomly irrespective
of whether they were removed or not, except for the high chrome plungers which had
higher than expected frequency of dags.
6.3.2 Plunger Type
3 0 % of all dags and 6 0 % of repeaters were on feeders with trial high chrome cast iron
plungers despite these only comprising 1 3 % of feeders in the plant. There was no trend
with any other type of plunger in the initial survey.
Following concerns of dagging on trial 31 OSS bar plungers at T9, a further survey
conducted in March, 1994. Portland was not using 31 OSS bar, but some Nicrofer trial
plungers also contained high chrome and high nickel which one would have expected to
behave similarly to 310SS. It showed there was 1 6 % less chance of dagging on
Nicrofer plungers than for normal cast iron plungers.
One observation which is common for cast materials is the rough surface on some
plungers. This m a y provide a place for bath to freeze and "key". These m a y dag when
first installed, but eventually the surface will become smooth. The bar material starts
off very smooth and is not affected by manufacture. In several plants, trials of different
cast materials have been unsuccessful despite the plungers working satisfactorily at
other plants e.g. 31 OSS at T 6 and P A spec cast iron at T9. This m a y be due to poor
manufacture quality control rather than a fault with the material per se.
6.3.3 Dags Removed
One operator removed all dags during the test period, but the number of dags was the
same in his pots over the period as the other pots; removal had no effect. Very few dags
were removed by operators (10%). Only extremely large ones (>400mm diameter)
were removed as these were fouling the feed chute.
6.3.4 Anode Effects
N o trend was seen on anode effects (Table 6-II(a)). In fact, there was more than double
the chance that no anode effects occurred if a dag was observed viz. 1 0 % versus 4 % .
This suggests that dags cause the holes to be larger, so there is less chance the hole will
block and restrict alumina flow. It m a y also suggest more alumina bypass unmetered
into the hole on these integrated feeders which results in overfeeding and potentially
more chance of m u c k forming. Hence, dags are not significantly linked to anode
effects.
6.3.5 Liquid Levels
Table 6-II(b) shows the percentage of pots (in a given total liquid level range) that have
dags. "In spec" pots are those which the total liquid level is "in specification".
The "Dag Factor" is the number of dags divided by the number of pots (in that liquid
level range) divided by a similar calculation for the "in spec" pots. It is a way of
quantifying the chance of dagging by showing the relative frequency of dagging
compared to ideal pots. For example, if the total liquid level is 4 0 m m above normal,
there is almost 3 times more chance of dags, than if the total liquid level was below
normal. Hence, there was more likelihood of dags (especially multiple dags across the
pot) for high liquid levels.
6.3.6 Location in Pot
The middle of the pot dagged the most with 3 2 % of dags occurring on the middle of
five feeder locations (Table 6-1). This trend is similar to plunger temperatures,
suggesting a relationship between dagging and plunger temperature (Figure 6-2). If
bath chemical concentration was important, one would have expected that there would
be similar frequency on all plungers and that all plungers would dag at the same time.
This was not the case, so this discounts the theory that bath chemistry is the cause of
dagging.
TABLE 6-H
LINK BETWEEN DAGS AND ANODE EFFECTS

(a) DAGS FOR DIFFERENT ANODE EFFECT HISTORY
NUMBER
OF DAGS NUMBER OF ANODE EFFECTS ON ONE DAY
ON ONE DAY 0 1 2 3 4 5
0 1020 296 56 15 8 3
1 115 36 8 1
2 33 10 1
3 18 4
4 2 3
5 3 TOTAL DAGS = 1632
NUMBER PERCENT
ANY AE OCCURRED? ANY AE OCCURRED?
YES NO YES NO
DAG? YES 63 171 4 10
NO 378 1020 23 63
(b) PERCENTAGE CHANCE OF DAGGING RELATIVE TO TOTAL LIQUID LEVEL
LIQUID
LEVEL #POTS NUMBER OF DAGS AT EACH FEEDER LOCATION (%) DAG
RANGE IN RANGE 0 1 2 3 4 5 FACTOR
NORMAL 51 65 24 10 1 1.0
>20mm 71 54 20 15 7 4 1.3
>40mm 51 52 13 9 6 6 14 2.8
Notes: (a) "Normal" liquid level is the total (bath plus metal) liquid level for normal pot operation. ">20mm" and
">40mm" are levels 2 0 m m and 4 0 m m in excess of normal operating total liquid level.
(b) The "dag factor" indicates the relative tendeny to dag. It is calculated as follows:
((# daqs/# pots) in chosen liquid level range)
((# dags/W pots) in normal liquid level range)
FIGURE 6 - 2
PLUNGER TEMPERATURES FOR
945°C POTS BY LOCATION
LOCATION 1
MEDIAN 11i"C
Ott
91-100
3
131-140 171-180
-t-H—I—I—I I I I I
211-220 261-280 291-300 331-340
LOCATION 2
MEDIAN 1S6°C
81-100 131-140 171-180 211-220 261-260 291400 331440
LOCATION 3
MEDIAN 168%
91-100 131-140 171-180 211-220 251-280 281400 331440
LOCATION 4
MEDIAN 146"C
61- 71- 91- 111- 141- 161- 171- 191- 211- 231- 261- 271- 291- 311- 331-
60 80 100 120 140 180 180 200 220 240 260 280 300 320 340
LOCATIONS
MEDIAN K6°C
61- 71- 91- 111- 131- 161- 171- 191- 211- 231- 261- 271- 291- 311- 331-
60 80 100 120 140 160 180 200 220 240 280 280 300 320 340
Motes:
(I) Plunger temperatures (°C) for 8 pots on 20.8.93, 31.8.93 and 10.9.93
(ii) Overall plunger temperature: mean 143°C, median 135°C and s.d. 42°C.
(iii) Pot temperature: mean 945°G and standard deviation 17°C
(iv) Y axis represents number of observations.
6.4 CONCLUSION ON CAUSES OF DAGS

From section 6.3, high chrome material was clearly the most important factor in
dagging. However, this material was used only on trial and is an exception. This trial
aside, it appears that dags are caused by
(i) surface roughness,
(ii) high bath level, and
(iii) high plunger temperature.
Surface roughness can be addressed by use of bar material or by quality control of cas
manufacture. For normal operation of a pot, there is little that can be done to do
anything about the plunger on operating feeders.
High liquid levels would tend to have material contact at higher levels on the plunger
which are not as smooth as the normal operating surface of the plunger by the increased
wetted surface area. This m a y give a starting point for gradual buildup as the bath
freezes. Control of bath levels is via control of anode cover bath/alumina blend,
tapping/filling pots if out of bath level target range, cover practice when setting new
carbon and cover practice over holes during normal operation.
High liquid levels not only have an affect by coverage of higher surfaces that may be
rough, but also affect the surface temperature of the plunger. There are other causes of
high plunger surface temperature such as flame from the hole, radiant heat and high
frequency of crust break cycles.
It is postulated that high plunger temperatures cause dags by the interaction of late
freezing and loose powder around the plunger. Because the plunger is hot, the bath
temperature is m u c h higher than its freezing point (usually about 920°C) and any
particles of alumina or anode cover in the area will tend to cling to the wet bath. This
causes surface roughness and further plunging into the bath allows continued buildup to
occur until eventually, if the dag gets big enough, it m a y fall off due to hitting the
feeder chute, dag scraper (if installed) or the crust.
The A 2 feeder showed quite clearly the link between buildup and this procedure. The
A 2 plunger raises and then the alumina is delivered. Because the plunger is often still
wet, some of the alumina will stick on the surface and the dag is created. Once the dag
starts to form, the rough surface allows buildup to continue by "keying" onto the
surface.
The next section discusses studies on the causes of high plunger temperatures so tha
better understanding of one of the main causes of dagging can lead to methods of dag
prevention.
6.5 FACTORS AFFECTING PLUNGER TEMPERATURES

For a range of pot and feeder conditions, 120 measurements were taken of gas
temperature next to plungers when fully retracted. The results are discussed in the
following sections.
6.5.1 Pot Temperature
The spread of plunger temperatures was very wide and was independent of pot
temperature (Table 5-VH). The higher the pot temperature, the higher the plunger
temperature... 10-20°C rise in plunger temperature for similar rise in pot temperature.
However, the variability was extremely large. The peak plunger gas temperatures were
independent of pot temperature.
6.5.2 Plunger Temperature Over Time
The scatter of temperatures of a 3 week period also showed temperatures varied over
time by a large margin (Table 6-ID). It should also be noted that pot temperatures
averaged 945°C for this survey which is very low for pot operation. Normal pot
temperatures are generally hotter than this - about 955-960°C Despite this, the peak
plunger temperature reached over 300°C (versus a mean temperature of 143 C ) .
There was a random variation in temperature for the 40 plungers, with large swings i
plunger temperature independent of pot temperature e.g. 1007#3 dropped 180 C for a
rise of 35°C in pot temperature and 2074#3 increased 85°C for a drop of 2°C in pot
temperature.
fS
o
<
OH
CD 0 CD
CO CU cu
E E E CD
m CO
LL
CO
LL
CO
LL
CL a.
O O
CL
O
o
111 —.
o o o o O o o a o o O o O
m o i m o i CO o • a o CO 8 • in in o O o O
in
n
a: r>
* - CM T- CM T- CM CO T- CN
UJ y CO CO "^ CM
sg o> o o
0 • - .
Z CL.
CO
CM CM r^ r~ in CN o in
CM "3"
3 5 x- CM °> <D CO O CM
CO m in o o a> CN CM
CD o CO
_l LU
OL H
CO c cu cu CU CZ 0
a> E E E CD E
m a. CO CO CO a.
ra
O LL LL LL O LL
o
ui —«
-• E o a O O o o oo o o a O O o o o O
o • <N O i o o • • o in Q O
CM in
T- x- CN o
CO o o
CM CO »- o
•q- *T
"3- 9 E
X ^-
UJ y
v r ' ^ • CO CM •* r- co ^- N - in CO o OO CO CO N
CO T -«- co r* CM "3- CO in CO CN CO o ,— -^ CO a in •^r o CO CO r^ in
z Q- CM CO co rr in vr in •a-
O S ^'o CM
-1 Ul
Q- 1-
CO CD a> cu c cu cu c cu c 0
E E E CD E E S E CD E in co _. o
Q. (0 CO CL
m roLLCO CO
O
J9- co
ra co -^ Q! CM
z O
LL LL LL LL
on o LL
o
1-
UJ ? o o o o o o o o
5 CM o
•>- T
O
i CM
O
O i f- o
m ' CO
O
o
CM
O
o 8 o
i o
o
a o
T— o o
1 o
ZU9 CO
o CO
•<- C M T-
*~t- T CM CO CN o
BC
_l
UJ
Q
ftpin co K§
LU o o> m O) T O I O S CO o CM CO TT
UJ U Q." CM s •a- ai co IO T- T-
o L?gi
LL Z
_5 5
CM
T-
r> <N -"a- a < D CN
CM f » * - T - r— T —
in Tf CO
T — C M •>- s
CM T-
•"T
x—
in
*~
m n
T- t- o
CM TT CM
CO CO CM r^ in r^
T~
^~ rt'-**—
T~ CD
^N CO
-J UJ O)
Q- 1- DC
UJ "5 O)
ta ta CD CZ CD CD CD c a> < Ul w
CO
2 CD
E
CO
E
CO
E <u E
CO
LL
a. CO
O LL
E E
CO CO
CU
Q.
O
E
ro
LL
D -1
Z Z D-
cn
o
LL LL LL LL
S CO CO <
CN
UJ ^~. c= J5
o o o o o a o o
CM IO
x- CM
i CM O , 88,
T"- X -
o O
o
O
a
CN To—
o
i o
m
o
in
t- o o CN
O
o
CO UJ UJ UJ Ul o o
CN i —- CM^
>>>>
or r?
UJ j-*
oooo CO o
CM O CO CM O O I D 9 C N in in CD 00 r~ CO CD (N — CD CO CO in
Q 1
O IT CM <o Tin B T.
§1
O CM CO T - CM T - p» co in CN <o CO CO M •=tC M •q- ^r
00
CO in in
^^ 'tf- CO^- CO
J LU U) JD
0- 1-
^ J» ?
w 0 cu cu c: CU v
CU CD •S (1) ^
E E E CD E E E E O 0 _
CO CO CO Q- CO CO CO
m
O ra gP 8>
LL LL LL
o LL LL LL LL
o o o o o o o O O O o o o o o o O o
s i "g
E CM O O CM O • o o O o o • o in in o O o ct 0 <0
9 *- CM to *- ^r T — X— <2 °
in *- CM T - CM .. O U
i —-
S 5%
|:5
cc (7 s-s O
ui
CD y" CM Cl O ta CM CM f 05 CM o CO T — CO *- CO ,o CN CO n in CO a o ioS °
Z Q-
<e 8°
£28
O CO r- r^ a <- o •>»• ^-
s T o CD T CO
in
0 S o
_l UJ
3 CL h- 5 cn £
w
C •"
CD
CO
CL
s ^ (? JO
(JJ-~.en LO O m o ^ t <N CO CM in co CO in o CO CO •<- CO CO CO CM in CN •sr r^
O--CO CO IO co r^ CM CN <D CO CO CD CN CO CO r t •* -3- CM CO ^r $y CD CO -<T £
H— S O) 0> O) Ol Ol Ol Ol Ol Ol Ol 0) 0) O) O) «
1
o
CL
o> oi ai ot Ol Ol Ol O) Ol U) Ol
I,! i
E 5 -o
z CN CO r- CM CO x- CM CO CM CO CM CO CN CO CN CO E J5 0
D T— T- CM CO T— T- • * —
*~
ce E LL or
M2
Figure 6-2 shows the same data sorted by location in the pot. There was an overall
trend of #3 being the hottest, followed by 2 & 4 (10-20°C lower), then 1 & 5 (a further
20-3 0°C lower). The middle of the pot had the highest temperature. This was similar
to the pattern for dags discussed in Section 6.3.6 and illustrated in Table 6-1.
The reason for the highest temperature in the centre of the pot is probably due to th
pots being side-by-side (similar to the layout in Figure l-4(b)) where there is less air
circulation in the space between pots. This the normal layout for prebake pots. End-to-
end pots (such as thefirstthree pots on the left hand side of Figure l-4(a)) m a y not be
affected as m u c h by this problem as there is better air circulation along the complete
length of the pot.
6.5.3 Hole Size and Flame
Table 6-III also shows the hole dimensions. If the hole had a strong flame that is near or
touching the plunger, the temperature reached up to 300°C If the hole is closed,
plunger temperatures were about 100 C This variability was far greater than variation
from location in the pot.
The size of the hole interacts with the amount of gas flow at the plunger. The strong
the flame under the plunger, the higher the plunger temperature. If the hole is large, the
gas can escape more easily, so the plunger is generally cooler.
6.5.4 Number of Holes Open
The number of open holes affects gas velocity at the plunger. If all holes are open or if
all holes are closed, plunger temperatures are cool. If only one or two holes are open,
then the high gas velocity tends to impact the exposed plunger more, raising plunger
temperature.
A survey of 44 random pots (220 feeders) at Portland showed almost half the holes
were closed (Table 6-IV). In comparison, T 1 7 rarely has holes open, and Til has
under 3 0 % open. Both plants have 3 feeders per pot and little to no plunger
penetration. D 9 generally has all of their 4 feeder holes open, as they have deep
plunger penetration. Also, operators check there are no blocked feeder holes every 32
hours as part of their pot checks.
TABLE 6-TV
SURVEY OF FEEDER HOLES OPEN

ON OPERATING POTS
#HOLES OPEN POTS WITH HOLES OPEN
NUMBER PERCENT
0 21 48
1 12 27
2 9 21
3 1 2
4 1 2
5 0 0
Note: Survey of 44 pots (220 feeders) at Portland.
It is best to either have all holes open or all holes closed, then gas velocity is minimal at
the plunger. Venting the gas elsewhere from the feeder or low penetration will help
prevent dags (but will have a negative effect on heat loss).
6.5.5 Plunger Length and Penetration
In the survey of 44 pots at Portland, a number of feeders were test pots where the
plunger length had been shortened by 5 0 m m as part of the trials on optimizing
penetration depth (Section 5.4.2). Short plungers had a 4 0 % greater chance of having
the hole closed compared to the historical length. This is due to less bath penetration
with short plungers and more chance for anode cover to fall into the hole to protect the
plunger from heat. This suggests that plunger temperatures are less and there will be
less dagging for the shorter plungers.
Operating pots at a target total liquid level (rather than just bath level) optimizes
life and dagging. The lower the penetration, the lower the plunger temperature and the
better the wear rate.
6.5.6 Proximity of Plunger to Hole
Table 6-V shows temperature at different distances from the chute. This indicates the
effect of proximity of the plunger to the hole. The closer the plunger was to the anode
cover the hotter the plunger. Every 1 0 0 m m closer raises plunger temperatures by about
50°C This has been confirmed by similar tests by Patrick at Boyne Island smelter (83).
The higher the anode cover or the closer the anodes are to the plunger the more the ho
jet of gas impacts the plunger and the hotter the plunger becomes. Deep anode cover
does not help dags. The deeper the metal, the higher the anodes and the hotter the
plunger. Once again, liquid level affects dags.
It is useful to explain the relationship between crust level and liquid level. The pot
voltage is controlled by moving the bridge, and hence the anodes, up and down to
maintain a target resistance between the bottom of the anodes and the top of the cathode
(Figure 1-3 and 1-5). There is generally a gap of 4 0 - 6 0 m m between these two surfaces.
A s the metal rises due to metal production, the anodes are raised and the crust attached
to the anodes rises as well. Thus, the hole gets closer to the plunger until the metal is
tapped and the metal level drops again. Smelters tap metal usually between 24 and 48
hours apart. Clearly, the plants with longer times between tapping expose plungers to
more period of high temperature and have relatively greater chance of dagging.
If a pot becomes unstable, the resistance often rises to move the anodes away from the
unstable metal level. Once again, this causes the cruot to approach the plunger and
exposes the plunger to higher temperatures. Note also that bath level rise does not cause
the crust to rise, as the anode height is purely affected by the anode-cathode distance.
However, the higher the bath level, the deeper the plunger will be wetted, so dags can
form on the surface.
The ideal is to have a large feeder stroke length to keep the plunger as far away from
the hole as is possible. It is interesting that D 7 actually raise/lower their crustbreakers
in the superstructure to achieve m i n i m u m bath penetration. In this way, they effectively
CHAPTER 6 PLUNGER BUILDUP ANALYSIS PAGE245
TABLE 6-V
PLUNGER TEMPERATURES RELATIVE TO CHUTE TEMPERATURE

ACTUAL TEMPERATURES (°C) TEMPERATURE RISE (*C)
DISTANCE BELOW CHUTE (mm L_ DISTANCE BELOW CHUTE (mm)
TIP T E M P
(*C) ZERO 100 200 300 400 100 200 300 400
<100 98 107 107 107 100 9 9 9 2
90 106 - 16 -
75 87 95 12 20
92 122 167 240 - 30 75 138 -
79 150 262 •QUI - 71 183 430 -
96 106 107 10 11
100-150 126 211 - 85 184 -
105 120 231 - 14 126 -
136 220 tfbfm - 84 267 -
145 212 - 67 -
148 250 K H - 102 242 -
110 105 105 100 117 5 5 7
>150 350 360 442 532 +10 92 182
233 280 443 578 780 47 210 345 547
217 279 344 430 476 | 62 127 213 259
Note: Red numbers are over 300 U C.
RELATIVE TEMPERATURE INCREASE BELOW CHUTE (°C)
DISTANCE (mm) 100 200 300

MEDIAN 14 85 140
MEAN 35 92 192
STD DEV 33 80 140
NUMBER 9 15 8
MEDIAN AIR TEMPERATURE RISE

VERSUS DISTANCE BELOW PLUNGER TIP
CO
S
ui
K
Ul
CL
s
150 200
DISTANCE BELOW PLUNGER TIP (mm)
have a different stroke length, but still have a standard feeder. Hence, dagging and
feeder life are optimized. However, this feature is a very costly design and cannot be
easily incorporated in most plants. It is preferable to control the cause (liquid level)
rather than control the effect (feeder heights).
Another factor worth mentioning is that this temperature rise from proximity of the
crust can have a very significant effect on cylinder seal life. If the seals are operating
just under their m a x i m u m rated temperature (as is the case at Portland), a change in
distance from the hole could push the seals over their limits and cause failure (Section
7.3.3:3 and 8.4.2).
Integrated feeders have a crustbreaker stroke distance of 356mm (except for T8 at

4 7 0 m m , T 8 at 4 7 0 m m , and Portland and T 2 at 5 0 8 m m ) . Independent feeders have
stroke distances of 500-600mm (Table l-H) so are farther from the crust and hence
plunger temperatures are lower. This will assist prevention of dags forming.
6.5.7 Wet Time
Although the time that the plunger was wet was not varied in the dag survey pots, trials
of a group of 17 pots at different dwell times showed that changing from 4.5s to 2.5s
dropped dags by 7 0 % .
Wet time is governed by:
(a) whether the plunger immerses in the bath,

(b) time in the bath per crustbreak, and
(c) frequency of crustbreak cycles.
Increasing dwell time may have no effect on wet time if the plunger does not touch the
bath. Hence, addressing theoretical penetration depth can allow higher dwell times
without affecting dagging or feeder life. With respect to penetration depth, Section
5.4.3 discusses plunger wear and Section 8.4.2 discuss the heat conduction up the shaft
to the seals.
Generally, dwell time is seldom changed in a plant. Hence, feeder speed, air leaks and
valve quality need to be controlled to minimize dags. T o test this, thermocouples were
placed on a feeder at the top of the plunger shaft and halfway d o w n its length (Section
8.4.2). The temperature trace of the two thermocouples is shown in Figure 6-3. Figure
6-3(a) and 6-2(b) shows the situation where the plunger temperature was initially cool
(70°C); (a) has thefirst60 minutes and (b) has thefirst200 minutes.
Figure 6-3(c) was at ambient temperature when first placed in the pot. Within 16 hours
the plunger and shaft hadrisento 250°C despite the fact that the cylinder had not cycled
into the bath at any time. This test showed that the temperatureriseis slow (5 minutes
to rise 50 C), but within hours the plunger temperature reaches a steady state
temperature irrespective of whether it is cycling in the pot or not. Thus, penetration
m a y not be a significant factor in heating plungers, only the effect of gas flow and/or
radiant heating.
If the plunger is hot, heat will conduct up the shaft to the feeder components. Hence,
the whole feeder reaches the plunger temperature within 16 hours. If a dag occurs and
if the primary cause is high temperature, the dag will probably continue to reform if
knocked off. This theory is substantiated by the observation in the plant dag survey that
removal had no effect on dags returning (Section 6.3.3).
Whatever temperature the plunger is subjected to, the plunger temperature will quickly
report as the same temperature at the cylinder seals. A s seals and stainless steel springs
cannot take temperatures m u c h over 200 C, the situations where there is a flame at the
plunger will rapidly affect cylinder life (Section 8.4.2). This normally results in air
leaks, ore leaks, and spring failure. A good rule of thumb is that a feeder with a dag is
probably also cooking the seals.
Plunger wear may in fact reduce with dags. The buildup reduces the frequency that the
outer microlayer of corrosion products falls off the plunger at each immersion (Section
5.4.7).
FIGURE 6 - 3
PLUNGER T E M P E R A T U R E PROFILES
(a) INITIALLY COOL PLUNGER

(3 HOUR WARM-UP)
500
O 400 •—•"
•— *
0
UJ PISTON ROD,^—•""""""^
jjj 300 .-.Af".
r
- -
-j—j_ZJ^j^-z '\>-^
H
a 200
0. ^ S ^ ^ PIUNGER SHAFT
Sf 100
1 1 1 1 1 1
10 20 30 40 50 60 70
MINUTES OF PENETRATION IN BATH
(b) INITIALLY COOL PLUNGER

(3 HOUR WARM-UP)
500 -p.—
PISTON ROD t 4 f- K PLUNGER WITHDRAWN
*^>. FROM BATH
O 400
o
• q
UJ
§ 300 —
i- T — ^ P L U N G E R SHAFT
£ 200 •
Q.
BJ 100
1 1 1 1
50 100 150 200 250
(c) INITIALLY HOT PLUNGER

(12 HOUR WARM-UP)
400
O 300
a
Ul
BC
J- 200 •
111
a.
100 -
10 20 30 40 50 60
6.6 DAG PREVENTION AND CONTROL

From the previous discussion, it appears that excessive plunger temperature is the
principal cause of dagging. Methods to minimize dagging include both prevention and
control.
6.6.1 Plunger Coolers
O n e solution to dags is to cool the plunger with air. In 1993, trials were conducted on 6
startup pots in which airlines were installed to cool the plungers. This proved very
successful in stopping dags and proved quite clearly that cool plungers do not dag.
However, at about $1,000 per pot (plus the cost of air) the solution was too expensive
for hundreds of pots. A cheaper mobile cooler was developed which can be moved
from pot to pot so only the feeders with problems are treated. Once again, air usage
would be high if many units were in use and it would be difficult to control the number
in operation as operators would tend to install them wherever a dag occurred rather than
control the cause. Coolers would be used to control the effects without addressing the
principal cause.
This method of preventing dags is not recommended because of capital and operating
costs. However, it is a useful tool in the event of a major dagging problem on a very
hot pot.
6.6.2 Liquid Level Control
As noted in Section 6.5.6, the crust and hole moves farther from the fixed retracted
plunger if the metal level falls. The bath level has no effect on penetration through the
crust, only on the depth the plunger gets wet.
A feeder design must be chosen so that the plunger penetrates every day for all pots
irrespective of normal variability in liquid level.
It is commonly thought that total liquid level (bath plus metal level) can be control
by good control of bath levels (as metal level varies little from day-to-day in most pots;
about 1 5 m m to 3 0 m m ) . O n e must also recall that there are different metal levels in
each pot. Pots m a y differ in metal level by up to 1 0 0 m m across all pots of a plant. A s
in almost all plants the feeder position isfixed,the penetration depth will vary from
pot-to-pot (Table 5-VI). Thus, metal level in each individual pot is important.
In addition to pot-to-pot variability, accuracy of tapping will affect total liquid lev
Variability in anode cover quality causes a changing addition rate of bath material into
the pot so bath levels vary as a result. G o o d control of bath level is very difficult to
achieve in most operating pots. In a pot with a large carbon area (such as Portland) this
is even more difficult as a small addition of material m a y result in a large change in
liquid level. Pechiney have invented a bath dosing unit for their A P I 8 pot which is
similar to a pot feeder (53). This an expensive but an effective w a y to ensure a fixed
liquid level in the pot.
A total liquid level target for each pot would address the metal level variation. These
targets should be based on a laser measurement of "tide mark" on the plungers in the
pot, such that there is a standard penetration depth for target liquid levels. If the metal
reference is changed, then the target is changed. In this w a y feeders will be standard,
but pots will be different. A s there would not be one target value for all bath levels,
"variation from target" would be used to report on performance of those w h o remove
metal from the pot (called tappers). This would help both dagging and feeder life.
The main problems with this method is the variation in bath volume from pot-to-pot.
This affects dissolution rates of alumina and aluminium fluoride additions, so could
affect fluoride concentration in the pot. High bath levels usually result in the pot being
more unstable and less efficient. For pots with high metal levels, the bath level could
be quite low, which could affect alumina dissolution and setting. In addition, having
different targets for each pot is more confusing to operators, and could lead to less
control of liquid levels.
As pot control is more important than feeder life, targetting total liquid level is not
recommended.
6.6.3 Plunger Penetration Depth
Reducing plunger length (without otherwise changing the feeder) will raise the bottom
edge of the plunger farther from the crust. This will wet the plunger less, and reduce the
gas heating effect.
The main problem with this type of action is that some pots may have low metal level
(and hence low total liquid level), so m a y suffer from insufficient penetration.
Unfortunately, to cater for the exception, all feeders will suffer from either higher
failures and dags (if metal level is high), or will be more prone to blocked feeder holes
(if metal level is low).
A method has been designed by the author to modify feeders on pots with high metal
levels, so that the net penetration will be less. These special feeders will have shorter
plungers. A series of pins on the mounting flange prevents these feeders being installed
on incorrect pots, and vice versa. It is therefore possible to change (say) 9 0 % of feeders
to very low penetration, yet cater for the remaining 1 0 % with special feeders. In this
way, it is not necessary to accept poor plunger wear, low feeder life and poor
penetration for all feeders in order to protect a few. All that is required is to keep some
special feeders as spares for a few pots. Cost is only about $50/feeder.
Section 5.4.2 discusses plunger penetration in detail.
6.6.4 Pot Gas Venting
It is desirable that the gas from the pot does not vent near a feeder. If the feeder hole is
buried (by use of anode cover), the plunger cools and the pot tends to find another place
to vent. A s long as the real cause of dagging is removed (e.g. anode effects, high bath
level, slow feeders) this procedure would be successful. Section 3.3(b) noted that high
gas flow can cause blockages in the crust. Thus, high gas flow affects feeder dagging
and dose delivery.
Breaking a vent hole away from the feeders reduces the gas velocity at the plungers.
This is not possible at many plants as there are no means to break a hole. Breaking the
hole at the tap hole (where the metal is removed from the pot) is very difficult to keep
consistently open as this is commonly covered to reduce airbura of the end anodes in
the pot. Hence, if possible, make gas holes to vent gas away from the feeders.
6.6.5 Reduced Anode Cover Depth
Having the anode cover far away from the plunger causes heat loss from the hole and
probably burning of the carbon due to exposure to air. Covering anodes is more
important to the economics of operating a potline than feeder life and dags. HiaV>
carbon consumption and poor pot efficiency are many times more expensive than feeder
life. There are other things that can be done to minimize dagging that do not affect pot
performance. It is reasonable to comment, however, that increasing the cover depth (as
is desirable for minimum airburn and minimum carbon consumption) will deteriorate
feeder life and increase dags. It is wise to design and operate the feeder to maximize
feeder life without dictating h o w m u c h anode cover is placed on the anodes.
6.7 MAJOR FINDINGS FROM PLUNGER BUILDUP ANALYSIS

(i) Dags form on hot plungers.
(ii) High chrome cast iron plungers dag more than H R or P A spec cast iron or
stainless steel plungers.
(iii) Dags appear and disappear randomly over time.
(iv) Removing normal sized dags has no effect on the frequency of a dag reforming.
(v) Anode effects and dags are not related, except for very big dags that m a y foul
the chute sufficiently to freeze the feeder in the retracted position.
(vi) Operators often blame dags for anode effects because they cannot think of any
other reason. Dags are not a major contributor to anode effects.
(vii) High total liquid level promotes dagging.
(viii) Proximity of the plunger to flame or the anode cover promotes dagging.
(ix) W e t time affects dagging.
(x) Rough surface castings can encourage dagging.
6.8 RECOMMENDATIONS FROM PLUNGER BUILDUP ANALYSIS

(i) D o not use high chrome cast iron plungers.
(ii) Ensure that cast plungers adhere to a surface roughness specification on delivery
from the supplier.
(iii) Conduct trials on shorter plungers.
(iv) Only remove dags that foul chutes butfixthe real cause of dagging.
(v) Use longer plungers for sustained low level liquid pots.
(vi) D o not target total liquid level for each pot as this m a y result in lower
efficiencey on some pots which m a y be less cost effective than the savings in
feeder life.
(vii) If possible, vent gas away from feeder plungers.
CHAPTER 7 PNEUMATIC CYLINDER INVESTIGATION PAGE 254
CHAPTER 7
PNEUMATIC CYLINDER INVESTIGATION
7.1 SUMMARY
Joint high frequency trials by Portland and Parker Hannifin (Australia) on 12 cylinders
found that the life of seals was affected mostly by (i) temperature, (ii) speed and (iii)
eccentricity...in order of priority with the most important first.
A cylinder lubricant investigation identified that unless the correct type and correct
amount is used, seal life will deteriorate due to lubricant breakdown. Teflon has a lower
coefficient of friction and higher temperature limit than Viton. Improved seals and
alignment are expected to extend rod and piston seal life and reduce air consumption.
7.2 INTRODUCTION
A s can be seen from the previous discussion, cylinder design is critical to the life of
feeders. Smelters have tried m a n y different cylinder designs over the last 50 years but
one is still faced with cylinders that fail within a few hours to over 7 years; even in the
one plant. The causes are often difficult to trace due to a poor or non-existent tracking
system and lack of data on plant conditions relevant to cylinder life.
Clearly cylinder performance has a significant effect on shot size accuracy (Section 2),
crust breakage (Section 6) and frequency of failures. It is also clear from Figure 6-3 in
Section 6.5.7 that cylinders are exposed to similar temperatures to plungers due to
conduction of heat up the shaft, so the pot conditions have a most significant effect on
cylinder life. O n e needs to ensure any comparison of cylinder type or cylinder
component type are conducted under cell-like conditions.
No papers have been able to be traced that have trialed different cylinder types under
controlled conditions comparable to aluminium cell conditions. Cylinder companies
quote vague data of h o w well their products operate on various plants but it has been
found that these data are not objective. The author has not as yet heard a voluntary
statement from a supplier that their product was no good. In fact, there have been
several instances w h e n the author (and other plant representatives) have been
deliberately mislead by cylinder suppliers on the assumption that plants would not be
contacting each other due to competition. This was quashed when m a n y plants started
to converse as part of this research. A number of suppliers found that their products
lost favour w h e n the facts emerged.
In order to understand the features that both improve cylinder life and the effects of
plant conditions on cylinders, two investigations were conducted. T h efirstwas a rapid
cycling of cylinders of different component under controlled conditions. T o date it has
not been possible to do trials of different designs, but equipment is available as a result
of this research to do this if required. The second investigation w a s concerned with
lubrication of cylinders as this is a major contributor to seal life which is in turn the
most frequent part that fails in cylinders. These investigations are discussed in Sections
7.3 and 7.4 respectively.
C H A P T E R 7 P N E U M A T I C CYLINDER INVESTIGATION P A G E 256
7.3 CYLINDER TRIALS
NOTE: D A T A F R O M T H E P O R T L A N D ALUMINIUM A N D PARKER HANNIFIN (AUSTRALIA)

CYLINDER TRIALS A R E C O N T R O L L E D B Y A CONFIDENTIALrrY A G R E E M E N T A N D
C A N N O T B E PUBLISHED O R R E P R O D U C E D OUTSIDE A L C O A O R PARKER HANNIFIN
W I T H O U T EXPRESS WRITTEN PERMISSION F R O M P O R T L A N D ALUMINIUM A N D PARKER
HANNIFIN (AUSTRALIA).
7.3.1 Introduction
From March 30 to September 3,1992 a joint investigation was carried out by Portland
Aluminium and Parker Hannifin (Australia) on twelve (12) 1 2 5 m m crustbreak air
cylinders of similar type to those used at Portland.
The purpose of the trials was to determine which factors had the greatest effect on
and piston seal life and to test alternative Parker Teflon seal units. The opportunity also
existed to test other components such as piston rod material, rod/shaft connections,
springs, stroke speed, cushioning and grease types.
The accelerated trial was conducted at 40 times the normal plant frequency rate.
Parker supplied facilities, designed equipment, and provided personnel to monitor

performance during the time. The author specified operating conditions to which the
cylinders were exposed, specified duplicate monitoring equipment, supervised
independent validity checks, conducted gauge capability tests, selected parameter
settings for each run, collated/analysed the data and wrote thefinalreport (85).
7.3.2 Procedure
Six cylinders at a time were mounted in the basement of Parker Hannifin's workshop in
Sydney (Figure 7-1).
Confidential: Written permission required by both parties to publish or reproduce outside Alcoa and
Parker Hannifin. Alcoa permissionfromPortland Aluniinium Technical Manager.
FIGURE 7 -1
CYLINDER TRIAL EQUIPMENT
TEST RIGS
Figure 7-2 shows the features of a cylinder trial rig. Each cylinder was heated by an
electric trace in a frame which had its o w n individual temperature controller. The
sensing points were thermocouples on the barrel and in a hole in the bottom block
drilled immediately next to the rod seal unit so as to correctly monitor the actual
temperature to which rod seals were exposed. The two temperatures followed similar
trends. This duplication assisted in a few cases of thermocouple failure.
Confidential: Written permission required by both parties to publish or reproduce outside Alcoa an
Parker Hannifin. Alcoa permission from Portland Aluminium Technical Manager.
FIGURE 7 - 2
CYLINDER TRIAL RIG
CYLINDER
HEATING
IACKET
[THERMOCOUPLE
CHAMBER
F O R ALUMINA
AERATION
FLUIDISATION
INSPECTION
PORTS
AERATION
CHAMBER
MECHANICAL
STROKE
COUNTER
The rigs were assembled and monitored 5 days a w e e k by Greg Mansfield of Parker
(under the direction of Gary Nauer). Special frames were constructed with the
following features:
(i) alumina aerated around the seal/mass,

(ii) identical weight to that suspended under operating cylinders in the pots,
(iii) eccentric masses to simulate misalignment seen on the feeders,
(iv) temperature cycling to simulate hot/cold operation in the pots and anode
effects,
(v) on-line accelerometer on each cylinder to see the effect of cushioning
over time, and

(vi) pistons stationary for a short time prior to change of direction.
Parameters tested included:
(i) 2 different piston rod materials (10L45 and 4140),

(ii) 2 different piston rod/shaft connections ( M 2 4 x 2 and stud),
(iii) 2 different top cushion lengths,
(iv) 2 different springs (stainless and K e m p e high temperature),
(v) 2 different piston seals (Viton and Teflon),
(vi) 2 different rod seals (Viton and Teflon),
(vii) 2 different stroke speeds (0.7s down/1. Is up and 2.0s/2.2s),
(viii) 3 different suspended mass eccentricities ( 0 m m , 5 m m and 1 0 m m
off-centre) and
(ix) 4 different greases (Tranz 414, A 2 M , B 2 and B1).
Although the list of variables is long, the prime test was for rod and piston seals. Sea
life w a s expected to be affected by stroke time, eccentricity, grease and temperature but
not by piston rod material, the rod/shaft connection, cushion length or spring type.
Parker Hannifin. Alcoa permissionfromPortland Aluniimum Technical Manager.
C H A P T E R 7 P N E U M A T I C C Y L I N D E R INVESTIGATION P A G E 260
The equipment comprised a large compressor, on-line air dryer (for -40°C d e w point),
16 pen chart recorder (for temperatures), alumina circulation system, seal leak detection
equipment, C R O (for accelerometer monitoring of cushioning) and high purity air
filters. About $80,000 worth of equipment was used in the trial.
A seal failure was deemed to have occurred if the leak rate was greater than 6 1/min fo
three consecutive days. The reason for the "consecutive days" was that the leaks tended
to change as the cylinder cycled, possibly due to side movement or uneven seating of
seals from time to time. (This is not unusual in the plant either, as operators see ore
leaks c o m e and go on a feeder.) O f the 12 runs, 10 failed due to rod seals and 2 due to
piston seals.
To ensure that there was complete candour, Russell Overhaul (with assistance from Dr
H u g h Stark) from Unisearch (University of N e w South Wales) was engaged to
randomly monitor the equipment unannounced. In addition, Russell Overhaul
independently checked instruments with high precision calibrated instruments to ensure
that results were valid. All Unisearch information cross-checked correctly with Parker
monitoring throughout the trial.
After each run, the cylinders were jointly stripped and inspected by Greg Mansfield and
the author. The actual eccentricity of the mass was measured with a plumb bob, and all
relevant dimensions of seals and components were measured in a detailed post mortem.
The trial took 6 months to conduct, and analysis of data took about 200 m a n hours.
Parker Hannifin. Alcoa permissionfromPortland Aluminium Technical Manager.
7.3.3 Results
Table 7-1 and Figure 7-3 show the results of the trials. R u n 5 failed due to
thermosetting of the "O" ring of the rod seal housing, and R u n 6 from incorrect
installation of the seal unit.
7.3.3:1 Sensitivity Analysis
Comparisons are m a d e where most factors are held constant except one parameter viz.
stroke time, seal material, eccentricity.
(i) Fast Speed - Viton
The data for Runs 11 A , 11B, 22A, 2 2 B and 6 are shown in Figure 7-3. It was
concluded that the longer the time that the seals were exposed to high temperature, the
shorter their life. The reason for the large differences in temperature exposure was due
to random periods that the cylinders were exposed to simulated anode effects or
simulated hot pots during the trial.
(ii) Eccentricity
Runs 1 1 A and 11B had no eccentricity and 2 2 A and 2 2 B had eccentricity of 1 0 m m ;
equivalent to Portland plant conditions. At higher temperature exposures (40 hours at
180°C) the eccentricity had no effect, but at low exposure times (less than 10 hours)
there was a large increase in life for less eccentricity viz. 5 0 % .
(iii) Temperature versus eccentricity

Temperature exposure had the greatest effect on seal life viz. changing exposure from
zero hours to 55 hours changed life by the equivalent of 4.5 years. At low temperature
exposure, eccentricity was the next most important factor viz. changing eccentricity
from 1 0 m m to zero increased life 1.5 years at low temperature (Runs 1 I A and 22A).
Tf K <- Z
30-Mar 30-Ma
01-Jun 05-Jui
PISTON PISTO
hard
CO 317 CO CO (N CN
CN J»sa 1.6 CN T CD CO CO
CN -t CD
<0
-1
350
OK
s
1.8
ro 5- in x- IO to a> i- in
CN 5 5 85 CN CN3
s
z
o
T- CN CN TT r- CN CN
o
(0 o j _
03-Jui
30-Ma
ROD
hard
_]
306
CO 0* in *- <-
1.5
u. *- S CN S CN TT
Ul
S <° CN m
CN
1-
CN
CN 2T in CN p fe CO _, Q
CN CO in co TT CO Tf
cB "> CN CD fe S S S JJ O in *-
o CO r-
3
25
0.8/1.1
14-Jut
23-Jul
ROD
A2M
128
<
OK
o O) O
0.6
CO r*- o r>
IO Tf ^ IO
a B
75
10
2.0/2.3
14-Jul
29-Jul
ROD
A2M
OK
159
CN CM Jl
0.8
UD S <N CO io 3<N Tf
o
H
25
TRANZ
3
2.0/2.:2
I
414
OS TT
13
* r~ co *- cr
CO T-
CN CN 5 8 N CN
CO t-
CN CN §
w
25
5
TRAN Z
8S%
2.0/2.1
Q
414
T? = 8" r- ~*
-I co
™ 2 "? <» S °
LD O s> CO
TT
00
5 ° io ™ O CO
az ^ ^ c4 «> ^ K gfe S
SB «- o m N
2
0.8/1.
14-Jul
TRAN
*"* o
1
24-Ju
ROD
CO CD 2 O M
o CN
228
OK
s
414
CO CO CM
1.1
1- CN
§™ IO IO S«M Tf J3 S |
> •* ts
3s n»
0
Z
0
Rg
y
r
-rt IS CQ — LJ
3
CN iE <* if) to T- CO '•3- IO IO
? 8 _E " E
5 p * feCO 5T- S ? ^ O ro £ "O <n CO
T- o in N co" g g I g g
2
RAN Z TRAN
0.7/1
i 24-Ju
ROD
r 14-Ju
m
OK
CN -o c
163
414
CO
0.8
5- *~ 5 TP "" Tf
•" « a ^
£.2 •"= § £ >
25
0
I.7/1.I3
< * 8 | "8
414
fe 5 7 S 2 O CO CO in CO CO CO CO lO rr rr
^— " H
•= 6 it) » ' K
co *- ^. r^ 5
s- ?c
•> o rr
200- 220C
180- 200C
@ 8H ?S N O
2204•C
R ay
cond'n
X1000
years
hours
hours
220+
180-
200-
6 ci 4
date
date
I
seal
JJ E E 8.
c coEE.fr
o
? R ar-
1 1
3 t- r-
z Ul
'3 O UJ rr
3
UJ
CE 10 rr
tj UJ 3 O
tn
5 f 1 § O fe O
EASE @ COI
R0KETIME(
SHION LENC-
USE OF FAIL
CENTRICITY
TJ r: UJ a.
UJ _l ul
UIV. LIFE
a a Q
EASE
CLES
Ul
»iRT
z CO s
D
10
D h-onrr
cc » I U O O
£r p z >- a < o
a
1(3 «| ID O III O
z
(0 in
3
rr a.
o <
CO
•a S
s 13
u
T3
o
•a
*ao •fl
o
U H
0 m CO CN o
• «
*-
' 1 1
1 ' 1 1 ' 1 '
CO tr>
CM CCuj
_
* an*
O o
o
z 00
UJ _l .J
O or o •«-
ID „
DC < <
zO >0
UJ (0
_I (A LU 5
0 z
O CO ft CO
.Q
ra
to
< aa Q 1- K UJn L.
OO (0 co il
u.
a: ac
Q. § CClil •
CO CD
*- II DCUJ
• /
SL CCJ
O
OC LU in VC CJ
• 3
a
UJ
i-3 i- § +01
CN *-.#% M
CO
O
09
ro wo (/> O n CL H
a.
CU
LU <-> < _l
i CL LL W LL (0 2 a/u X
LU
T3
c _t
T3
(Si
z z 7• UJ tCOa> £ a
I UJ
0.
Z Z
P-P- 33
> > u.
(-u.
1-
LU LU
T- II
or UJ
- 0 *
1-
2
0.
UJ
O
0 E E
a. +- c
c
o
§ 8
E$ S. E
3 -a
o
cu
CJ
o s re & | s
BEST FIT LINE ^V.
• • « + LU
1-
a T-
•a S-
a
o _ >
9* B *
E CO I c
o
VITON, FAST^x.
(9 O
I £
o o °
c Q
® £E E
o
O O
5
•a
«s
C^
-O u£
I
- CN 2
m p.
n I ?
/<o
h # CM*""
CCUl
• /a*
' a
UJ
u.
1 , 1 • 1 1
<u
CO
CJ
CJ
UJ
H
(iv) Slow Speed - Viton
All the slow speed runs had temperature exposure in the region 40-50 hours over
180°C. A s eccentricity reduced, life increased viz. Runs 5, 3 and 4 had 10, 5 and 3 m m
eccentricity respectively and resulted in 0.8, 2.8 and 3.0 years life.
(v) Slow V's fast speed - Viton
For similar high temperature exposure, comparison of Runs 11B/22B and 5/3/4 showed
slow speed gave longer life; about 2 years at 5 m m (or less) eccentricity.
(vi) Fast speed - Teflon
The lower eccentricity of R u n 12 gave longer life (6 months) compared with R u n 21 for
similar temperature exposure. The wear rate on R u n 12 was exceptionally good for the
period it ran prior to thermosetting of the seals. The reason was unknown.
(vii) Slow speed - Teflon

The eccentricity for both Runs 13 and 23 was identical, but R u n 23 had m u c h higher
temperature exposure time (40%) 1 2 % more rod seal wear (as measured by seal
thickness) probably due to the greater temperature exposure. This illustrates that
greater temperature exposure weakens the seals.
(viii) Slow V's fast speed - Teflon
It is unclear of trends for Teflon with respect to fast and slow speeds from the test data
due to the compounding effect of significant differences in temperature exposure across
Runs 12/21 versus 13/23 respectively.
(ix) Viton Versus Teflon

Teflon seals were superior to Viton. Extrapolation of the Viton line for Runs 11 A,
22A, 11B, and 2 2 B suggested a 2 year longer life for Teflon w h e n over 50 hours above
180°C.
7.3.3:2 Teflon Versus Viton Seals

Initial examination of total life indicates that the n e w Teflon rod and piston seals
proved unsuccessful relative to existing designs. The teflon seals had a silicon " O " ring
to press the seal against the piston rod or barrel. T h e silicon " O " ring hardened and
formed flat faces so that there was no pressure on the seal. This provided a route for air
to escape. A t the time of failure of the energiser, the seals were expected to have
significant operating life still available.
Note that silicon has a lower recommended operating temperature than Teflon viz.
260°C versus 300°C peak (87,88,89). O n e must consider " O " rings as well as the seal
material w h e n designing a cylinder. Stainless energisers are a safer design for high
temperature applications (Section 8.4.6).
From Figure 7-3, one can see that the Teflon seals were exposed to generally a much
longer period of high temperature. This will have affected their life. Prediction of life
at lower temperature w a s not possible due to the restricted number of runs with the
budget and time available. Despite this, one can assume that the Teflon would have
been superior to Viton if one compares the life of Viton at similar periods at high
temperature. Note that independent feeder plants use Viton, so there is a chance that
cylinders m a y not give good life in some potlines that have high background
temperatures e.g. Portland or T8.
7.3.3:3 Temperature
There is a clear result that exposure to high temperatures (for similar other conditions)
drops seal life. These results were not unexpected. In the plant, several feeders from
the same pot have been seen to fail within shifts/days after a pot becomes very hot (over
970°C).
In order to optimise seal life, the periods of high temperature exposure need to be
minimised and the design of seal/bush/scraper need to be addressed. Refer to Section
8.4.2 for discussion on methods to reduce the effect of high temperature exposure on
seal life.
7.3.3:4 Speed
O n e would have expected failure of Runs 3 and 4 after about 3 months based on the
results at the fast speed at similar temperature exposure. However, they failed at over 2
years. A n 8 factor improvement w a s seen for this condition. This suggests a cubic
relationship between speed and wear rate of seals which agrees with theory of frictional
wear.
7.3.3:5 Eccentricity
The effect of designing out a 1 0 m m eccentricity relates to about a 5 0 % (or 18 months)
improvement in rod seal life at low heat exposure and fast speed, but no improvement
at high temperature exposure.
At slow speed, the trends for eccentricity were reversed at high exposure. However
there was only minor difference in eccentricity for this comparison (viz. 3 and 5 m m for
Runs 3 and 4), so this m a y be due to sample variability or other factors. The ideal w a y
to test this factor is to repeat the trials at 2 exposures and 2 eccentricities. This was not
possible due to lack of time and money.
Run 4 failed earlier than Run 3 due to thermosetting of the 'O' ring i.e. the energisers
behind the seal hardened and lost their spring effect to push the seal against the rod or
barrel. A s there was effectively 4 0 % of the seal still available, it is expected the rod
seal should have lasted at least 4 0 % longer (or 3.2 years). Thus it would be better than
run 3 at slightly worse eccentricity (Figure 7-2 has R u n 4 located at its assumed life).
Hence, it was concluded that there is better life for less eccentricity with the differe
being more noticeable at lower temperatures. Benefits of up to 2 years are possible.
Improving alignment by increasing bushing length at the rod seal, maximising bush
length of the feeder/crustbreaker and using spigotted joints are all useful actions to
minimise the effect of this factor (Section 8.4.4).
Independent feeders have less eccentricity than integrated feeders, so will have longer
seal life (Table 8-III).
7.3.3:6 Grease
Tranz 414 was used in all runs of Viton except for 2 runs with A 2 M . B 2 was the only
non-insulating grease.
Seven of the 12 runs showed grease breakdown. There was no pattern between
temperature profile and lubricity at failure. However, A 2 M did not suffer from
breakdown despite a long exposure to high temperatures. Technical data from the
supplier suggests this grease should be superior to the others due to 60°C higher
temperature rating and smaller particle size. This grease was later tested to be better
than Tranz 414 in the lubrication investigation discussed in Section 7.4.
7.3.3:7 Piston rod/shaft connection

There was no evidence of chrome lift-off on any piston rod, and no fatigue failure of
connections. This was probably due to generally lower than normal acceleration rates
on these cylinders (compared to plant experience) and low cycles encountered - all
except one run were under 3 years of service.
7.3.3:8 Top cushion length

The effect of top cushion length was not conclusive due to the small sample size and no
piston rod failures to illustrate the differences between the two designs.
7.3.3:9 Spring design

One stainless spring fractured a couple of turns from the top (Run 11 A ) . This type of
failure has not been seen before, so it was assumed it m a y be related to having the
spring compressed length too small in the retracted position on this cylinder. O n e
cylinder (Run 4) started to shrink which is c o m m o n in service at Portland. There were
no faults with the K e m p e high temperature springs. It m a y be concluded that the
K e m p e springs were superior to stainless.
Confidential: Written permission required by both parties to publish or reproduce outside j\lcoa and
Parker Hannifin. Alcoa permission from Portland j^luminium Technical Manager.
7.3.3:10 Deceleration rate
Deceleration rates were monitored daily on each test viz. 5 days per w e e k plus spot
checks by Unisearch. There w a s a trend that deceleration rates tended to cycle up and
d o w n for several days at a time. There w a s no trend that they increased over time.
Deceleration rates increased several fold when the cylinders were exposed to high
temperatures. This is probably due to the very large drop in air density with increasing
temperature. For example, at a pressure of l,000kpa, a change in temperature from
100°C to 200°C drops air density by a factor of 40^. This drop reduces dramatically the
cushioning as there is no "body" in the air to stop the piston and metal-to-metal contact
occurs. This will have the biggest effect w h e n the cylinder is retracting as the dosing
cup is hit with no cushioning on an A E D D feeder (Figure 1-9).
This may explain an observation by operators that feeders get "stuck" on hot or new
pots. W h a t they often hear is a large "bang" and assume it was due to the plunger
getting stuck in the crust. It is suggested this noise is mainly due to the higher
deceleration force at the higher temperature. Once again, the effect of pot conditions on
feeder operation is evidenced. There appears to be little that can be done to avoid this
happening other than the obvious action of reducing the temperature of the pot.
7.3.3:11 Air leak rates

A small leak varied from cycle to cycle. This is w h y a run was only terminated if the
leak was excessive for readings on three consecutive days during the trial. If plant air
checks are conducted on a plant to determine if air leaks are excessive, one must be
careful that decisions are not affected by this variability. It is preferable to conduct a
gauge capability study to determine at what levels of air leak that feeder changeout is
required. A leak rate of over 301/min per feeder w a s found to be critical after these tests
were completed at Portland.
Air leak rates at room temperature were up to 7 times higher than at operating
temperature, but this ratio varied considerably. Portland is n o w monitoring air leak tests
into and out of the workshop and overhauling only those that have leaks over 30 1/min.
Previous to this cylinder trial, if there was any leak at all, the cylinder was completely
overhauled. D u e to the effect of temperature, this old procedure led to overservicing of
cylinders and excessive repair costs. T h e old procedure is c o m m o n at most plants.
(Refer to Section 8.4.4 for discussion on non-routine cylinder overhaul cost savings).
The 30 1/min plant acceptance limit at Portland is much higher than the acceptance
criteria used in the trial viz. 6 1/min. T h e plant acceptance condition is based on
economic grounds, but the trial value was based on Parker's standards of a satisfactory
seal. In plant conditions, it is possible to accept higher leaks if the leak does not cost
too m u c h air/money or cause other process problems such as ore leaks.
The results of the trial on Run 22A (Figure 7-4) suggests that leak rate may not deca
very fast and that the leak rate is not a linear relationship to seal life. Hence, the best
changeout procedure appears to be to replace only on a needs basis rather than on a time
basis (Section 9.4.3).
FIGURE 7 - 4
CYLINDER 22A R O D SEAL LEAK RATE
LEAK RATE
(1/min)
80 80
70 70
60
50 -
40
h 60
- 50
40
30
20
10 —
- ULJ a D
n OO
D
-J 30
20
•
- 10
- nn c -
0 nHff0.l5 UUUn1 • I 1 u J 0
1.5 2 2.5 3.5
EQUIVALENT Y E A R S IN SERVICE
Note: 200,000 cycles is equivalent to one year service in Portland pots
7.3.3:12 Insulation
Insulation dropped to zero over the life of every cylinder trailed. This has also been
confirmed in the plant (Section 8.7.2). Checks on all types of greases used in the trial
(except B 2 ) showed that they were insulating initially but all broke d o w n during the
trials. It is assumed that the small amount of rust and material that gets through the
filters in the airlines gradually causes cylinders to short out electrically. Note that a
very fine filter was used in these trials (10 micron versus 50 micron at Portland) to
minimise any chance of this happening, yet the insulation still dropped to zero. Section
8.7 discusses insulation further.
7.3.4 Cylinder Trial Summary
Table 7-II summarises the data as discussed above. This shows that the importance of
parameters were temperature, speed and eccentricity (in the order most important to
least important). Table 7-in lists recommendations for design and use of cylinders.
Portland and Parker have actioned m a n y of the recommendations.
Almost all trial results have been substantiated in plant conditions at Portland. These
plant scale results are discussed further in Sections 8.4 and 8.5. This verifies that the
trial conditions were comparable to plant conditions and is testimony to the effort put
into planning and implementing the trial.
Parker Hannifin. j\lcoa permissionfromPortland Aluminium Technical Manager.
CHAPTER 7 P N E U M A T I C C Y L I N D E R INVESTIGATION P A G E 271
TABLE 7- II
PORTLAND/PJ\RKER CYLINDER TRIAL DATA SUMMARY
ROD SEAL HIGH TEMPERATURE CYCLES

ECCENTRICITY SPEED LIFE
MATERIAL EXPOSURE (X1000)
Viton 5 hrs Zero Fast 5 years 1,000

Normal Fast 3 years 600
40-50 hrs Zero Fast 1 year 200
Normal Fast 1 year 200
40-50 hrs Zero Slow 3 years 600
Normal Slow 1 year 200
Teflon 50-60 hrs Zero Fast 2 years + 400 +
Normal Fast 2 years + 400 +
50-60 hrs Zero Slow 2 years + 400 +
Normal Slow 2 years + 400 +
IMPORTANCE PRIORITY
PARAMETER CHANGE D R O P IN LIFE
Temperature 50 hours exposure over 180°C 4 years
Speed 50 hours exposure over 180°C 3 years
Eccentricity zero to normal 1.5 years
Teflon was better than Viton at high temperature
Parker Hannifin. .Alcoa permissionfromPortland Aluminium Technical Manager.
TABLE 7-ID
RECOMMENDATIONS FROM CYLINDER TRIAL
* Temperature, speed, eccentricity in priority of importance.
CYLINDER SUPPLIER
* Reverse bush and rod seal to have seals near air cavity.
* Soften viton seals to accept some thermoset.
* A2M is best grease.
* Temperature stability of energisers is important as these were the first

cause of failure for 6 of 12 runs.
* Rod seal/bush installation can be installed incorrectly easily.

- need keying to prevent errors at installation.
* Teflon seals better than viton at high temperature.
CYLINDER USER
* Temperature - use old feeders for new pots.

- vent air leaks and delay feeder changeout on hot pots.
* Speed - slow the feeder down as much as possible.
* Eccentricity - install better guides and bushes or universal joints.
* Use A2M grease not tranz 414.
* Use more grease on overhaul.
* Check air leak rate in workshop on overhaul and accept a small air leak.
* High temperatures cause high G forces - plungers do not get "stuck".
Cushion settings and speeds do not change with time, so workshop

settings can be used for records and decision making.
* Insulation not necessary for new cylinders.

7.5 LUBRICATION INVESTIGATION
7.5.1 Introduction
A key factor in the life of piston and rod seals is the type, quality and nature of
lubrication. In m a n y plants no lubrication is used at all - even during overhaul of the
cylinder e.g. most Alcoa smelters. S o m e plants have continuous oil dispersant systems
in the air line e.g. T17. Most smelters use a lubricant only during overhaul e.g. plants
using Atlas Copco or C P O A C cylinders (Table l-II).
At Portland, analysis of rod seal life of identical units in Terry and Parker cylinders
over several years shows a mean life of 550 months and 1080 months respectively using
Weibull analyses. O n e possible cause of this difference is grease. Terry cylinders have
used Magnalube and Parker have used Tranz 414 until this grease investigation.
Workshop experience is that Magnalube breaks d o w n more than Tranz 414.
If the lubricant and/or the wiping system are not adequate, the lubricant can break dow
and form a grinding paste that can seriously affect seal life. Lubricant breakdown
products can also be vented from the cylinder and block the exhaust muffler. This in
turn causes back pressure on the cylinder which affects the net pressure drop across the
piston. For integrated feeders this results in a speed change, change in shot size
variability and reduced pressure at the crust. Blocked feeder holes can develop (refer
Sections 3.5.3 and 4.5.3).
Discussions with several seal and cylinder manufacturers showed varying attitudes to
lubricants. Almost all suggested that use of a good grease could enhance significantly
the life of seals, but, except for Atlas Copco, none had trialed alternatives or had advice
on the best type to use.
Based on this lack of information and difficulties in finding any independent expertise
C E T E C Chemical Technologies (Dr. Vyt Garnys) was engaged to analyse a range of
greases and to co-ordinate analyses by Shell and C E T E C under the author's direction.
Technical and laboratory assistance was given by John Quinlivan, Rainer Schade and
Mauro Del Frate from Shell.
An investigation of alternative greases (plus no grease) was conducted by the author

using a Kepner Tregoe Decision Analysis. This method lists the "musts" and "wants"
of good lubricants and rates the alternatives based on weighting of importance of each
"want" and the relative merit of each alternative.
Eleven (11) alternative lubricants were investigated after initial screening of produ
offered by Shell, Mobil, Optimol, Tranz Performance Lubricants, Skega and
Magnalube. Except for Mobil, at least one grease was used from each company plus
use of no grease at all (which is the situation in several Alcoa plants). N o suitable
Mobil lubricants were available in Australia so they could not be included in the trials.
Except for Skega G P S O (which is a suspension) all alternative lubricants in the trial
were greases.
Detailed results are available in Kissane (85), but for simplicity only the relative
used in the Kepner Tregoe Decision Analysis are listed in the summary of results shown
in Table 7-IV.
Ill
to o ,_ o ,_ ^_ (B S n a
o < > > > I TJ TJ "g TJ
o o o
g UJ
c c c
^ OC
o
UJ
> > > eo •» °

ra
E TJ
3 £ 1o o - CO
a: O TJ .C TJ CM IO
IA
c O 3 O co
UJ
c
<
> >- >

is TJ
3 £ sa CO IO in
F O TJ
O 3 O «
» TJ
c c
ra ra 1£ CN
>- > >- ra TJ CO CN
o TJ
TJ O CN CD
!A
c
8£ c co
4= ro to .-£ ro (-, ,—.

> > >- s « o° C TJ U r TJ ° C 4 °
> >- > 2 in CD 'E I -Si 1 co g o 8°

o o S o CN *^
>• >• > TJ TJ £ TJ ^ Ol 8^

o o 3 o CN ^
0 5 co JJs ro
> > >- T- T-
TJ TJ !c TJ
2 O) -
o o s o
E E C
>- >- >

1 lljflo $
"3- E TJ TJ r tl ° <- r-
hi • 8 S * 8
UJ
S3
3 V >- > PJ CT> ° T f O l TtSO ° r- o
_i E CO
O O S O
< (fl c c c
Z
o
f H -* 1
> >- > 'E ra co o ro ,—, CO
00 O TJ TJ j TJ o N °
O O J2 O
E C E
< IN >- > > 00 0> CNOIO) c TJ TJ .E TJ ^£ CO

< s 9 2 5 o
a> oo ^C CD CO CN T- io m CO
e
a. S'
E
A
o O Is 8, »
i? £ S
O) CO „
| o I 8|TJ
11!
Iiff SJ ^ ° % c t
&»•
Ol v. 3 o o o e
w
• 3 « ° TJ C C C p
u llfli
i— i— <D o — c
o o * >=
£ £ *- (A o £ a e§
i| J5
N e t "
I 5 S e
w « » g » I- — =*>e J
IS g.E-55 3•<—1 *-
E ° «= y
= c s:o o s S o
h- fl H . E Z 3 z
o Q; ti? •"
< CO E 5 < 3 O I -: -i ^ -i
«» 5 .a IB Es OQui
i
2 3 3 O r
S to to Q o
7.5.2 Procedure
D u e to the complex nature of the duty of a pneumatic cylinder, a range of tests were
used to simulate long term medium temperatures, short term peak temperatures,
oxidation, side load, and grinding nature of grease residues. It is useful to briefly
explain the methods of analysis and the equipment used.
(i) Differential Scanning Calorimetry (DSC)

A 5 m g sample was heated in an aluminium pan at the rate of 10°C/min in air at 70kPa
to 400 C. The heat flow into and out of the sample (using an empty pan as a reference)
was measured as a function of temperature. Dupont D S C analytical equipment was
used.
(ii) Thermogravimetric Analysis (TGA)

A lOmg sample was heated in a platinum boat at the rate of 20°C/min in air to 1,000°C.
The amount and rate of change in mass of sample was measured as a function of
temperature. Dupont T G A analytical equipment was used.
(iii) Penetration (ASTM D217)

The consistency of grease was measured by the penetration of a standard cone after
working the grease 60 strokes. The penetration was determined at 25°C by releasing
the cone assembly from the penetrometer and allowing the cone to drop freely into the
grease for 5s. A coarse penetrometer was used. Note that greases are classified by a
" N L G I (grease)" consistency number according to the penetration measured.
(iv) Dropping Point (IP396/93)

The dropping point of a grease is the temperature at which the first drop of material
falls from a cup. It is the temperature at which the grease passes from a semisolid to a
liquid state under the conditions of the test. Samples were heated at both the rate of
0.2°C and 1°C per minute. A Mettler Drop Point Tester was used.
(v) Sen wing Reibung Verschleiss device ( S R V ) (ISO 9001/9003, D I N draft 51/834)
The Schwing Reibung Verschleiss (Vibration Friction Wear) device uses a ball and
plate, or roller and plate mechanism. The test specimens were mechanically loaded at a
given frequency, load and stroke. Tests was conducted at set temperatures up to 180°C.
Friction force was measured continuously and friction coefficient was calculated and
plotted on a chart recorder. This device was used with Teflon as the plate material.
The tests using a ball simulated "extreme pressure" (which is applicable to wear of a
bush) and tests using a roller simulated seal wear. This device was considered to be the
best instrument that exists to measure seal friction coefficient under a range of
operating conditions. It was used to compare the wear scar of different types of Teflon
(Section 8.4.5).
7.5.3 Kepner Tregoe Decision Analysis Procedure
The Kepner Tregoe decision analysis considers M U S T S and W A N T S (Table 7-IV).

M U S T S are "not negotiable" requirements; if the alternatives do not meet these
conditions they were scrapped. In this study, all 12 alternatives passed the M U S T S .
WANTS are desirable outcomes but of different levels of importance. The WANTS
were listed and weighted by importance from 10 (best) to 1 (worst). Each alternative
was tested against the W A N T S . "Weights" were placed against each W A N T to compare
the importance of each W A N T relative to the other W A N T S . The "weightings" were
listed prior to any comparison of alternative to avoid any influence of "favoured"
alternatives.
Each WANT was then examined to establish the best parameter to quantify the
'WANT*. For each alternative, the actual analytical result is listed and a "rating" across
alternatives established their relative benefits from 10 (best) to 1 (worst). Some
' W A N T S ' had no data because there was (a) no data available, (b) no reasonable method
to analyse for it, or (c) no equipment was available to conduct such a test. It was
considered at this stage (and later assessment at the end of the decision analysis) that the
CHAPTER 7 P N E U M A T I C C Y L I N D E R INVESTIGATION P A G E 278
results of the W A N T S which had no data would not have changed the outcome due to
their low "weighting".
Once WANT "weightings" and "ratings" were established, these were multiplied
together and added to give a total for each alternative. This reduced the list from 12 t
4 alternatives (Table 7-V) For the final four, "adverse consequences" were examined
for each to reassess if there was a critical element that was very weak thus making it
sensitive to a particular extremes of operation.
TABLE 7 - V
FINAL LUBRICANT RATINGS
LUBRICANT POINTS ADVERSE C O N S E Q U E N C E S RISK
1. Darina 359 unstable at high temperatures HIGH

2. TA2R2 357 untried, but high rating in all areas LOW
3. Titan A2M 345 untried except PA/Parker trials LOW
4. Skega 338 good record at Atlas Copco plants LOW
5. Aeroshell 323
6. TA2R1 322
7. Magnalube 310
8. Tranz 414 296
9. No grease 267
10. Stamina 262
11. Paste M P 260
12. Innertox 229
Confirmation tests for top four:
SRV roller and Teflon plate, 1 hour at 180°C after 8 hours at 300°C.
Final rating:
1. T A 2 R 2
2. Titan A 2 M
3. Skega G P S O
Darina... major failure - deleted
O f the short list alternatives, the first alternative (Darina) was particularly sensitive to
spikes of high temperature so was a "high risk/high probability" combination. The
other three alternatives (TA2R2, Titan A 2 M , and Skega G P S O ) had no serious adverse
consequences.
Further comparative testing was conducted of heating each of the top four to 300°C fo
16 hours then doing a 180°C roller and plate (Teflon plate) test on the residue. It was
concluded that none dropped out from the list except Darina. This confirmed the
suspected sensitivity of Darina and it was decided to reject it.
This left TA2R2, Titan A2M and Skega GPSO for future trials on plant scale routine
cylinder overhaul. Tranz 414 would act as a "control" grease, as this was the best of the
greases already in use.
7.5.4 Discussion
7.5.4:1 Benchmarks
Tranz 414 and Magnalube were used as benchmarks throughout the investigation as
these have historically been used at Portland.
Historically, Tranz 414 has been superior to Magnalube at Portland. It would be

expected that Tranz 414 should have a higher rating to Magnalube in the laboratory
tests and that other greases should be superior than both Tranz 414 and Magnalube. At
the end of the decision analysis, Magnalube rated better than Tranz 414, but the final
scores were not significantly different. Both were well behind the best lubricants
(despite being considerably cheaper).
Skega GPSO has been used by Atlas Copco in all their crustbreaking cylinders for
many years. Atlas Copco have done considerable large scale testing of m a n y greases at
elevated temperatures. Skega G P S O was one of the three top options from the decision
analysis. It showed good all-round ratings with few low-rating qualities.
Thus, lubricants found to be successful in operating plants also fared well in the
laboratory tests and decision analysis. It was therefore concluded from the benchmark
comparisons of Tranz 414, Magnalube and Skega G P S O , that the laboratory tests and
decision analysis should indicate a satisfactory rating of performance that would be
expected to give comparable relative results in operating plants.
7.5.4:2 No Lubrication
The advice received from Atlas Copco, Parker, Terry, C E T E C and Shell was that
grease must be used with Viton. Teflon has some lubrication ability, but Viton has
none. This has been verified at Portland T9, T16 and T17. T w o cylinder types have
been used at T 9 and T15, and neither had any lubrication on either Teflon or Viton.
There was a major difference in life despite having identical assemblies connected to
the cylinders.
The option of "no lubrication" had interesting results. This had a better friction
coefficient at elevated temperatures but was far worse than the others at temperatures
under 180°C; almost three times worse than Tranz 414. It appears that "no lubrication"
is not as good for normal operation, but is not affected by the residue grinding action
when the seals are "cooked".
It is believed that insufficient lubrication has been a major problem at Portland,

especially from October 1991 to February 1993 when the amount was reduced still
further because it was considered that it was a waste of money to put too m u c h grease
on. Seal life increased from the time that the amount of lubricant was increased
(irrespective of the grease type).
Figure 7-5 illustrates the effect of the amount of grease. Initially, grease was wiped on
with a small spatula on the barrel, piston seals and rod seals. Then, in October 1991,
this was reduced to using a small paint brush. In February 1993, the amount was
increased to the m a x i m u m that would physically fit into any part of the rod seals/bush
area and onto the piston seal/bush area plus the barrel was liberally covered. The rod
seal life dropped by a minor amount on the first reduction, but increased by 4 5 % when
the amount was a maximum.
FIGURE 7 - 5
EFFECT OF AMOUNT OF GREASE ON
ROD SEAL LIFE
(a) NORMAL AMOUNT (PRIOR TO OCT,91) AVERAGE LIFE:
450 days
100 -y.
y'
j-
80
50
20
10
8 = 1 .,
9
y*
FaTk J re 3 = -
C e n i (or 3d 2E
M. 1B0. XAL MO. 1000L XKICl WOO.
A V E R A G E LIFE:
(b) REDUCED AMOUNT (OCT,91^JAN93) 440 days
100 /
80 ,r
y
s
y~ -
50 *+
/ r
20 .^
10 > " r
_« , < " ^
0 = 1 -{ ^
TJ —
*
xJJ>"
-'FoTU i =f 53 2
C a m por ad = 3 1 -k
is. aa i DO aoo. 600D. 1 <HKN.
AVERAGE LIFE:
(c) I N C R E A S E D A M O U N T (MAR93-JUN93) 660 days
100
80
50
_*:^ -
•
20
10
\/ '
f
5
ff = 1 .1>
A
•n = 73 4 ''
* •
y
FoTh ire 1 a 1 3
/
Certi tor 3d ; 1 E l> . '
io. aa. ai. 100. an. soa. tace.
Total O p w o t f n g T i m * (©aire)
Notes: (i) Basis: Magnalube grease in Terry cylinders using FEC part #423.
(ii) Vertical axis is "cumulative percent occurred''.
(iii) "Censored" refers to replacements that were not due to failures e.g. preventative maintenance.
Note that this comparison was for the same seal and grease, and there had been no
changes to bath penetration. Stroke speeds dropped between (a) and (b), so this m a y
explain w h y there was not an even greater drop in life w h e n the amount of grease was
initially dropped. The slower speed and friction drop compensated the potentially large
drop in life from less grease. Conversely, the expected increase in life of seals from the
lower speeds did not eventuate probably due to the reduction in the amount of grease at
the same time.
This example illustrates the compounding effect of making two changes at the same
time. It could be suggested (after the fact) that doing two things at one time was
unwise, but one must recall that the decision to reduce the amount of grease was
considered by Parker and Terry to have insignificant effect on seal life at the time.
Only after the cylinder trials and initiation of investigation into lubricants was the
possibility that this m a y be significant was contemplated. Thus, controlled experiments
can be affected by one's perception of what is significant.
7.5.4:3 Lubricant Options

If a lubricant is to be used, it is important to select one that not only operates well at
low temperature, but does not form a grinding paste at high temperatures. This requires
it to withstand oxidation, "bleed" (flow away from its original location by gravity) and
high pressures. W h e n these considerations were examined, the lubricants tested
performed quite differently. S o m e were better at high temperature, others take a long
time before breakdown, then suddenly totally disintegrated.
If insufficient lubricant is used, the higher friction forces cause rapid breakdown (of
even good greases) due to shear. It is necessary to have a reservoir always available so
the lubricant in the wear area can be replenished easily. For a vertical cylinder this
requires a lubricant that is thin enough to replenish the work area yet not too thin to fall
to the bottom of the cylinder over 5-7 year period even under extreme temperatures.
To select a lubricant that handles all the desirable requirements is quite a challenge.
Shell's opinion, selecting a lubricant for smelting pot alumina feeders is one of the most
difficult duties they have ever undertaken due to these complex issues. The solution to
the problem is state-of-the-art lubrication. However, not to use grease will give poor
seal life (especially for Viton).
Shell advised that copper and iron are catalysts to the breakdown of grease. Hence u
of bronze wipers (as used by Portland and T8) will affect grease life. Teflon wipers are
used on Atlas Copco, C P O A C , Lindberg, and Scheffer cylinders. Portland changed to
Teflon late in 1993.
7.5.5 Conclusions and Plant Scale Results
The recommended lubricants were
1.TA2R2
2. Titan A 2 M
3. Skega G P S O
The top two are made by Tranz Performance Lubricants (Sydney), and GPSO is
marketed by Atlas Copco (Sweden). These were trailed in the field since July 1993
using liberal quantities for each cylinder overhaul.
Compared to Tranz 414, cylinders using the above three lubricants have doubled their
rod seal life. Thus, the conclusions from this study have proved to be correct in
operating pots.
As a result of this lubricant study, Tranz Performance have initiated enhancements

T A 2 R 2 to develop T A 2 R 7 . Tranz T A 2 R 7 has been used at Portland since early 1994
with very good results. It is considered to be the best lubricant used to date, but it is
fairly expensive at about $10/overhaul or $300/kg. However, the cost is justifiable if it
increases cylinder life even by such a small amount as the average cost of a cylinder
overhaul is about $400-500. M o r e time is required to do a cosl/benefit analysis
comparison between T A 2 R 7 and T A 2 R 2 (which is considered to be the best of the
other three).
7.11 MAJOR FINDINGS OF PNEUMATIC CYLINDER INVESTIGATION
(i) Main factors affecting rod and piston seals (in order of priority from most
important) are
(a) temperature,
(b) stroke speed, and
(c) eccentricity.
(ii) Teflon is superior to Viton with respect to wear and temperature sensitivity.
(iii) Good lubricants for pot feeder pneumatic cylinders are T A 2 R 2 , Titan and
Skega G P S O .
(iv) Viton seals need lubrication.
(v) Lubrication can drop friction coefficient by up to three times that of no
lubrication (when under 180°C).
(vi) Grease rapidly breaks d o w n if insufficient quantity is used.
(vii) Copper and iron catalytically break d o w n grease.
7.12 RECOMMENDATIONS FROM PNEUMATIC CYLINDER INVESTIGATIO
(i) Use adequate quantities of high temperature lubricants for feeder/crustbreaker

cylinders.
(ii) Use Teflon seals.
CHAPTER 8 EXTENDING FEEDER UNIT LIFE PAGE 285
CHAPTER 8
EXTENDING FEEDER UNIT LIFE
8.1 SUMMARY
From the frequency of each type of failure of feeder/crustbreaker units, an analysis was
carried out of ways of extending their life.
Ore leaks on integrated feeders can be addressed by a minor alteration to a feeder to

allow leaking rod seal air to escape from the pot so as not to affect the pot operation. A
system to measure air leaks in the potrooms and in the workshop monitors excess air
usage and poor feeder quality so that these do not get out of control.
Opportunities exist for plants to extend life and reduce costs by accepting limited ai
leaks at overhaul, not overhauling feeders from off-line pots and not requiring
insulation for newly purchased cylinders.
8.2 INTRODUCTION
Following research into the dosing system (Chapter 2), plunger wear (Chapter 5) and
cylinders (Chapter 7), a clearer understanding of the components of feeding and
crustbreaking units has been established. This Chapter will consider in detail the
causes of failure of cylinders and feeding assemblies that lead to the feeder being
removed and h o w to minimise the frequency of these failures. Firstly consider the
interaction between operators in the potline and the repair tradespersons.
It is useful at this point to highlight an important difference between "reason for

removal" and "mechanical fault". "Reason for removal" is w h y the observer believed
the feeder was not performing (such as not feeding or feeding too much). The observer
cannot see the internals of the feeder so cannot state the reason was due to, say, a spring
failure or to the spool jamming. There m a y be several mechanical faults that can
contribute to a feeder being removed. For example, alumina leaking unmetered into the
pot can be caused by either cylinder failure or by dosing unit failure. It is important
that one is aware of the nomenclature on removal reasons as one m a y spend a lot of
time working on the wrong things.
As little testing is conducted on shot size accuracy of overhauled or operating feeders

for both integrated and independent designs, the usual reasons for removal are:
(i) a significant overfeeding or underfeeding situation which is visually obvious

to an observer, or
(ii) if the pot is not operating properly and the feeder is diagnosed to be
potentially at fault.
This diagnosis is often fairly subjective so feeders are frequently removed

inappropriately. G o o d troubleshooting procedures should be used before feeder
removal as feeder repairs are expensive and expose people torisksof falls and exposure
to heat, dust and fumes. Removal should only be used w h e n one is sure the feeder is
the cause.
Section 8.3 links the reason for removal and the mechanical faults. Subsequently,
Sections 8.4 to 8.8 discuss the mechanical faults in more detail.
8.3 CAUSES OF FEEDER FAILURES

are.
The causes of feeder removal of 11 integrated feeder plants in 1991 represented in
Table 1-V. These are due to ore (alumina) leaks (44%), pointed or short plungers
(20%), arced (13%), pot offline (13%), broken piston rods and dosing spools (8%) and
not feeding (2%). In recent years, removal of feeders because of excessive cylinder
bypass (piston seal leaks) has been added to this list. The reasons and extent of failures
is quite different in each plant despite having very similar designs. Clearly, there are
different problems at each plant, however, this presents an opportunity to optimise the
designs, maintenance and operations of feeders by exploring the strengths and
weaknesses between different smelters.
Independent feeders do not generally suffer from ore leaks but the other causes do occ
(though less freoêrAV, than for integrated feeders). T h e major causes of failure of
are
independent feeders leaking piston seals (cylinder bypass) and worn plungers. R o d
seal leaks are important for air usage reduction, but do not pose a problem to feeder
operation as the leaks are usually vented well and do not have m a n y downstream effects
(such as ore leaks). R o d seal leaks mainly create a cost penalty in independent feeder
plants rather than a cause of feeder removal.
As there are no springs, the dosing unit failures for independent feeders are associat
with jamming of the spool and failure of the spool shaft (or bush that supports it). T h e
author has limited data o n spool failures in independent feeders so there is minor
discussion on this subject. However, the discussion on spool dosing accuracy (already
discussed in Section 2.5.7 and 2.5.8) and integrated feeder spool failures (Section 8.8)
m a y be relevant to those interested in this subject.
CHAPTERS EXTENDING FEEDER UNIT LIFE PAGE 288
Cylinder life and plunger wear remain the main causes of feeder unit failure in most
plants - integrated or independent. Consider n o w the "reasons for removal" and trace
them to their "mechanical faults".
8.3.1 Ore Leaks
A n ore leak is where alumina (ore) pours into the pot unmetered and potentially causes
the pot to become mucky. This is usually seen as a dribble of alumina flowing from
inside the feeder chute or, if very bad, at the kidney plate (Figure 1-11). It can be
caused by several situations:
(i) Rod seal air leaks

The rod seals are located where the piston rod goes through the end of the
cylinder (Figure 1-20). These leaks aerate the alumina in the dosing unit so the
material does not seal where the spool seats on the fixed face of the assembly
(Figure 1-10). If the air volume is too large, the dosing unit pressurises, causing
alumina to come out of the kidney plate area like a curtain around the feeder
chute.
The reason the air goes downwards instead of upwards is that most integrated
feeders have a solid joint at the top mounting flange where there is usually an
insulating gasket or there is a metal-to-metal mating of flanges. This prevents
air from escaping from the top of the cavity where the feeder is mounted. Most
plants have a vent pipe from this cavity but this pipe can block if alumina is
carried out with the vented air (which is usually the case as fluidised alumina
has a 7° angle of repose and the level of alumina in the superstructure is above
the level of the vent pipe). Section 8.4 discusses rod seal leaks in detail.
(ii) Spool J a m m e d
The overlap between the dosing cup and the spool closing face is only 2-5mm
on an A E D D feeder (Figure 1-11) and is less than 2 m m for independent feeders.
Either partial or complete jamming can occur if the spool is not m a d e correctly
(refer Section 2.5.8 and Figure 2-16), if there is mechanical damage at the
dosing cup, if mechanical failure causes the spool to be bent or if there are
foreign objects fouling the sealing face. The c o m m o n cause of damage is where
feeder mounting bolts become loose and fall into the alumina and foreign
objects subsequently try to pass through the tight clearances of the dosing unit.
Depending on the alumina supply system, there are other potential causes of
foreign objects getting into the alumina. Plants often have mesh installed
upstream or just prior to the inlet to the feeder (T7), but these are commonly not
inspected or cannot be inspected (for example, T7). Pechiney developed a
sophisticated filtration system to prevent foreign objects getting into their hyper-
dense phase alumina supply system as part of their A P 3 0 pot design (51).
Section 8.8 discusses dosing failures in more detail.
(iii) Spring Failure

Spring failures are not relevant for T 8 or independent feeders. If a spring fails,
it does not force the spool flat against the horizontal sealing plate (Figure 1-11).
A s the spring only extends for a feed every 3 minutes or so, it is compressed for
about 9 8 % of the time. This is a most difficult task for any spring let alone a
302SS or 304SS spring commonly used in Alcoa A E D D feeders viz. Portland,
T2-T7 and T9-T17. The spring will creep over time, but is seriously affected by
excursions in temperatures. Section 2.5.3 explains the effect of spring quality
on dosing accuracy and Section 8.8.3 discusses what can be done about
preventing spring failure.
For independent feeders, ore leaks do not usually occur because:
(i) there is no spring in the dosing unit to fail,

(ii) dags do not affect dosing cylinder stroke as the dosing unit is separate from
the crustbreaker, and
(iii) generally the cylinders vent into the pot cavity in large vent passages so
there is little chance of back pressure.
When one considers that in 1991 44% of integrated feeders were removed for ore
leaks, this attraction of independent feeders is a very significant advantage and a clear
contributor to better life for independent feeders.
8.3.2 Pointed and Short Plungers
A s seen in Table 1-V, after ore leaks, the next most frequent failure (20%) of integrated
feeders failures is the combination of pointed and short plungers (which is due to
excessive penetration in the bath). N o data is available for independent feeders except
D 1 2 which had 8 5 % of failures for this reason. Most independent feeders use stainless
plungers to address this problem. Plunger wear is discussed in Chapter 5 where not
only the material of construction is discussed but a range of other things that can be
done to improve life viz. less penetration, less time in the bath, do not feed on anode
effect.
8.3.3 Arcing
The feeder and superstructure should be electrically insulated to prevent an electrical

circuit being m a d e between anode and cathode if the plunger penetrates the bath.
Although for normal operation there is only about a 4 V D C voltage drop across a pot, it
is possible that this m a y reach 100V D C in anode effect. At the higher voltages of
anode effect, arcing can occur between metal parts that are close together. This is
usually where the dosing unit part of the assembly touches therimof an A E D D feeder
(Figure 1-11) or similar location on other feeder designs. The other locations for arcing
are at the rod seal area of an air cylinder or, less frequently, at the piston of the cylinder.
About 13% of feeders changed in the integrated feeder plants surveyed in 1991 were
from arcing. It is rare that independent feeders suffer from arcing due to good mounting
and lower end insulation. Arcing is discussed further in Section 8.7.
8.3.4 Fatigue of Piston Rods and Spools
Broken spools and broken piston rod comprised 8 % of integrated feeder failures (Table
1-V). Both can be related to excessive piston speed, but both can be caused by inferior
design. Side loading of the plunger w h e n it is fully extended puts considerable strain
on piston rods where they join the piston (Figure 1-20). Piston rods generally fatigue at
this location. Note that only Portland and T 6 of the 11 plants surveyed in 1991 had
broken piston rods. Since this time, D 1 2 , T8, T15 and T 1 7 have made enquires to the
author about problems in their plants from broken piston rods. This failure mechanism
m a y be greater than first indicated and is probably plant specific. Section 8.6 discusses
broken piston rods.
AEDD feeder spools can fatigue at the horizontal face below the spring (Figure 1-11) or
they m a y wear at the top pipe section closest to the cylinder. Independent feeders can
be affected by bent shafts or broken bushes that support the shaft above the spool, but
generally do not fatigue due to slower speeds and lower driven mass than integrated
feeders. Section 8.8.1 examines spool failures.
8.3.5 Cylinder Bypass
Cylinder bypass is w h e n the piston seals leak causing air to bypass through the cylinder
(Figure 1-20). This a major cause of failure of independent feeders due mainly to
proactive attention to reduce plant air usage rather than a factor that prevents the feeder
from doing its job. For integrated feeders this historically has not been a major issue
(as illustrated by the fact that in 1991 not one of the 11 plants surveyed listed it as a
cause of removal. In the last few years following the authors feeder recommendations
(Table 1-VH), Alcoa plants have looked closely at cylinder bypass and have detected
very significant air usage concerns.
The cause of cylinder bypass is generally piston seal failure (refer Section 8.5). The
control strategy to minimise air usage is discussed in Section 8.4 as the choice of
whether to change the feeder is often based on economic grounds provided the plant has
sufficient air available.
C H A P T E R 8 E X T E N D I N G F E E D E R U N I T LIFE P A G E 292
8.3.6 Pot Off Line
This is the one cause of failure that most people want; 1 2 % of feeders changes out in
the surveyed plants in 1991. However, this research has found that these repairs are not
cost efficient, they are unnecessary and they have a negative impact on pot operation, so
there are considerable effort and savings by not doing feeder changeouts when pot
superstructures are overhauled. This will be discussed in Section 8.5 as this is purely a
cost based decision and not related to failure as such.
8.4 ROD SEALS
8.4.1 Causes of R o d Seal Failure
Rod seal failure is one of the major causes of ore leaks in Alcoa pots and hence one of
the major causes of feeder removal (Table 1-V). Examination of the cause of failure of
46 rod seals at Portland is seen in Table 8-1.
TABLE 8 -1
CAUSES OF FAILURE OF ROD SEALS AT PORTLAND
Worn 34%
Cooked 24%
Preventive maintenance 15%
Alumina in cylinder 15%
Broken piston rod 4%
Other 8%
100% Sample size: 46
The priority order may be different at other plants, but the factors are often similar.
Wear is due to temperature, speed and eccentricity as established in the cylinder trials
(Section 7.3.4). The other causes relate to high temperature exposure (cooked), ingress
of alumina past the wiper (alumina in cylinder), fatigue (broken piston rod) and other
factors.
Figure 8-1 illustrates schematics of some rod seal and bush combinations commonly
installed.
FIGURE 8 -1
CYLINDER ROD SEAL AND BUSH DESIGNS
(a) CONVENTIONAL - SINGLE SEAL
CYLINDER
CAVITY
WIPERl
ROD SEAU
SPACER BUSH
'O' RING
(b) SPLIT BUSH
^^^1
1 Mr^.
(c) REVERSED
(d) TEFLON DOUBLE SEAL
TEFLOl
WIPER
\EI\S,
T W O SEAL CARTRIDGE
The causes of rod seal failure are discussed in turn below (with the exception of broken
piston rod which is discussed in Section 8.6). This section only discusses rod seals, but
one needs to also note comments in Section 8.5 on piston seals, as they are interrelated.
8.4.2 High Temperature Exposure Control
The Portland/Parker cylinder trials (Section 7.3) confirmed that high temperature
reduces seal life. There are several areas where operators or designers can contribute to
cylinder life.
8.4.2:1 Pot and Feeder Design

There is little that can be done about increasing the distance between feeder and the
hole except on the design of n e w pots where it is best to use long stroke cylinders to
take the plunger as far as possible from the crust.
Integrated feeders have a crustbreaker stroke distance of 356mm (except for T8 at

4 7 0 m m , T 8 at 4 7 0 m m , and Portland and T 2 at 5 0 8 m m ) . Independent feeders have
stroke distances of 500-600mm (Table l-II) so are farther from the crust and hence
plunger temperatures are lower. This will assist in keeping seals cool.
As discovered in the cylinder trial, eccentricity is a major cause of rod seal failure
(Section 7.3.3:5). Failed seals often look worn on one side in m a n y of the plants the
author visited. This is probably from insufficient side load support of the feeder and/or
cylinder rod bush. The greater the angle of the rod the more the rod tears at the seals.
Independent feeders can achieve more support and have less misalignment of the
crustbreaker as the spool in not affected. This is a particular fault with the A E D D
design. Portland development has been unsuccessful in providing a reliable bush low
down in the dosing cup area as the bush designs tended to slow the spool and create
jamming. Section 8.4.4 discusses eccentricity in more detail.
Another factor on the temperature exposure of seals is the background temperature of

the feeder in the superstructure. All plants surveyed (except Portland) have the ducting
(that removes gases from the pots) at the base of the superstructure. Portland has the
ducting at the top. This causes the heat to be drawn up around the cylinder and
contributes to the seal cooking. Portland temperatures at the rod seals are about 155°C,
with peaks of 330 C being attained for hot pots (Table 5-VII). Measurements at T6, T 8
and T16 plants showed rod seal temperatures of 60-80°C; Atlas Copco, Parker, and
Terry believe this is typical. Thus, Portland conditions are probably worse than most
plants for temperature exposure on seals.
8.4.2:2 Normal Operation

Because the plunger dips into the 950 C bath every 3 minutes or so, it was considered
that reducing dwell time was an effective method of reducing temperature exposure at
the seals. T o find out h o w quickly the heat conducts up the plunger shaft, theoretical
heat transfer calculations were m a d e on infinite slabs, semi-infinite slabs, and irregular
solids using formulae and tables from Holman (78). These showed that the temperature
rise up the shaft was slow; minutes to hours were required before there was significant
effect, rather than seconds as commonly thought.
As this was hard to accept, a device was manufactured to test the temperature rise up a
shaft. This was basically a Portland feeder minus the assembly. Thermocouples were
placed at the top and halfway up the plunger shaft. It was the "worst case" design as
there was no heat protection from alumina between the cylinder and the pot cavity.
S o m e of the results have been noted in Sections 6.3 and 6.6 and shown in Figure 6-2.
The figure shows that it takes over 50 minutes to raise the temperature at the seal area
by only 50°C. This verifies the theoretical calculations. Clearly, plunger wet time
(which is no more than 5s) has no significant effect on seal temperature.
However, the seals reached the same temperature as the plunger in less than 16 hours.
Therefore, plunger temperature is the most important factor in seal temperature
exposure and, hence, seal life. It is interesting that almost identical results were
achieved in a similar trial in 1993 by Patrick at Boyne Smelters Limited (83).
This conduction up the plunger shaft will cause both rod and piston seals to reach
temperatures up to 300 C (Table 6-DI) which exceeds the m a x i m u m temperature rating
of any of the normal materials used for seals (Table 8-II). This means that dagging is a
probable sign that seal temperatures are excessive and seal life is dropping every second
the dag is there. The dag is an indication of seals cooking. Control actions are the
same as those discussed in Section 6.4.
TABLE 8 -H
MAXIMUM TEMPERATURE LIMITS

FOR SEAL MATERIALS
Material Operating Temperature Peak Temperature
Viton 210°C 240°C

Silicone 240°C 260°C
Teflon 260°C 300°C
(References 87, 88, 89.)
8.4.2:3 Hot Pots

Plunger temperature surveys at Portland show a direct link between pot temperature and
plunger temperature (Section 6.3.6:7). There is also a direct link between plunger
temperature and seal temperature (Section 7.6.2:2). Hence, hot pot temperatures will
affect seal life. (A "hot pot" is one with bath temperatures over about 970 C).
Once a feeder is changed out in a hot pot, it is likely that the new one will also cook
leading to more changeouts. It is a good idea to avoid replacement until the pot is back
to normal operating temperature. This procedure has been commonly used at Portland
since mid 1993 with improved overall feeder life and less repeating feeder changeouts.
Only w h e n several feeders are leaking in the same pot are they changed.
8.4.2:4 N e w Pots
Protection of feeders from the extreme conditions encountered on start-up will benefit
feeder life. N e w pot conditions m a y include:
(a) pot temperatures over 970°C for extended periods,

(b) high radiant heat from holes, and
(c) high bath levels.
Some plants put in old feeders with smaller plungers from off-line pots to assist the
crust breaking and sticking problems associated with the larger anodes eg. Til and
T17. This in effect protects the next overhauled feeder from heat exposure, and hence
extends its life. If they worked prior to the scheduled pot outage, w h y overhaul them?
Reusing feeders from old pots was trialed at Portland in April 1993, then reinstituted i
February 1994, (after installation of spool inserts on all feeders) with savings of
$12,000/month and no increase in failure rate.
8.4.2:5 Rod Seal/Bush Orientation

To minimise the effect of high temperature on seals, it is best to have the seals mounted
as close as possible to the cooler air of the internals of the cylinder (Figure 8-1). This
idea also has the advantage that it gives greater protection of the seals from foreign
objects/dust, as the bush is n o w between the wiper and the seals. Also the bush is closer
to the driven mass, so there is better side support for uneven loading of the shaft when it
is fully extended. This design cycles air about every 3 minutes, so the seal is
continually being cooled. This is a very good design to help keep the seal cool.
This design feature is used in Atlas Copco cylinders (Figure 1-20) and recent Terry
designs for Portland, T6, T 8 and D l 1 at the suggestion of the author.
Independent feeders are not affected by alumina ingress into cylinders because the
alumina is quite some distance from the rod seal area (Figure 1-12).
8.4.3 Speed
Based on best engineering judgement, slow speeds would seem to be better for seal life
due to less frictional wear. W e a r generally follows a cube relationship with velocity;
for example, half the velocity relates to 8 times the life. The cylinder trials suggested a
cubic relationship between speed and seal life (Section 7.3.3:4).
Plant experience at Portland did not show this magnitude of improvement when the
speed was changed in November 1991. This was due to a simultaneous drop in the
amount of lubrication on seals (Section 7.5.4:2). This illustrates the importance of h o w
one does an overhaul.
Independent feeder plants and those with long stroke times have little problem with rod
seal life; even for Viton seals which are of 50°C lower temperature rating to Teflon
(87,88,89). This is partly due to lower frictional wear from the slower speeds
(compared to integrated feeders).
8.4.4 Eccentricity
If the piston rod does not m o v e along its centre line, the eccentricity can cause more
wear of the seal and leaks will develop. Useful actions to minimise the effect of
eccentricity include
(i) improving alignment by increasing bush length at the rod seal,

(ii) maximising bush length of the shaft closer to the dosing unit and plunger,
and
(iii) using spigotted fittings at joints between assembly components.
The integrated plant T2 had only a 2 year seal life versus 5-6 years for independent
feeders with almost identical Atlas Copco cylinders (Table 1-H). If the life of cylinders
are so different despite identical cylinder design, it is suggested that the different seal
life is due to the feeder design and not to the cylinder. Better support for side loads
would improve seal life.
Support for side load can be improved at the cylinder itself by design of the bush at the
rod seal (Figure 8-1). The Atlas Copco seal/bush length is about 5 0 m m . O f the two
types of cylinder used at Portland, there is little uneven wear and better life on Terry
cylinders that have an 8 0 m m seal/bush length compared to Parker which has a 5 0 m m
seal/bush length. Lindberg seal/bush length is about 5 0 m m . A long seal/bush length
helps alignment and hence seal life; the longer the bush the better the seal life.
It may appear that there is little flexibility for plants that are restricted by cylin
length and are unwilling to spend capital on replacing all existing cylinders. O n e option
is to split the bush in half and place the halves at each end of the front end cartridge.
This effectively increases the bush lengthtothe full dimensional span of the front end
cartridge. Thus the bush can be "lengthened" without changing the cylinder. Especially
in a large cylinder with a long stroke, misalignment and side load can cause a very large
force on the seals and rod bush, resulting in seal wear.
Table 8-IH shows deflection for several plants. Note that independent crustbreaking
units (D3, D l l ) have only 5 m m side deflection versus 20-30 for integrated feeders.
Pechiney A P 1 8 pots have a 5 0 0 m m long cast iron bush housing (Figure 1-12) with (in
the latest potlines) a 1 0 0 m m long bronze bush at the lowest point. There is only a 1 m m
difference between shaft O.D. and bush I D . Alcoa feeders have no bush at the dosing
unit except for the 6 m m hardened insert invented for the Portland feeder (Figure 1-11).
In addition, in the most recent Pechiney designs (eg. D 1 0 Line 3), universal joints have
also been installed to prevent misalignment. This is expected to result in longer feeder
life by reduced eccentricity at the rod seal but at a very high cost. It is debatable if the
cost (well over $100) is cost effective.
It is desirable to have a bush as close as possible to the crust. However, for integra
feeders, the equipment at the lowest location is the dosing unit. The tighter the
clearances between bush and shaft, the more the bush affects the dosing unit (spool)
free movement which in turn affects shot size (refer Section 2.5.3). Thus, better
cylinder life is constrained by the need to meet a primary duty of the feeder. Once
again, the independent feeder is superior.
TABLE 8-m
PLUNGER SIDE DEFLECTION
% ROD SEAL COLLAR TO EXTENDED

AVERAGE
CYLINDER TOTAL BOTTOM PLUNGER TO
PLANT LEAKS AT FEEDER
O.D. DEFLECTION CYLINDER CYLINDER
WORKSHOP LIFE
FLANGE BOTTOM FLANGE
(mm) (mm) (months) (mm) (mm)
T11 100 18 90 12 1041 1397
T9 100 19 70 24 - -
T7 100 20 60 6 1043 1399
T17 100 20 75 36 1043 1399
T15 100 22 70 24 1092 1448
T16 100 22 10 24 1020 1376
T6 100 30 80 12 1022 1378
T13 100 33 90 9 1025 1381
T3 125 15 N/A 36 1260 1768

T10 125 15 80 N/A 1740 2096
T2 125 25 95 24 2555 3063
Portland 125 30 36 10 958 1466
D10 200 5 N/A 48 1000 1508

D3 150 12 N/A 48 1010 1518
Notes:
(i) D10 have 7 0 m m shaft in a 70.75mm ID cast iron bearing of 5 7 8 m m long. This wears at the end,
so another 1 2 0 m m bronze bearing of similar ID has been installed. Deflection measured on 23.6.92.
(ii) D10 line 3 incorporates a universal joint with its Atlas Copco 2 0 0 m m cylinder.
(iii) D 3 have two 3 0 m m x 60.5mm ID steel bearings to support the 6 0 m m O D of four years such that the
gap is shaft. Due to wear over a period now up to 4 m m , they will be lengthened to 6 0 m m and probably
will be changed to bronze (based on the author's recommendation).
(iv) T10 have same shape feeder as Portland but have a longer and smaller ID spool to support the shaft
viz 1457mm not 8 3 2 m m long and 52.5mm not 53.1mm respectively. The shaft is 5 0 m m diameter.
They were changing all feeders every 2 years whether they have failed or not until 1992 when they
changed to failure based changeout after a visit by the author.
(v) Atlas Copco recommend larger piston rod O D for integrated feeders to take side load.
(vi) T13 rod seals mostly fail duetoelectrical arcing, whereas other plants suffer from rod seal wear.
At Portland, spigotted flanges were installed between the cylinder and the feeder
assembly on 800 cylinders to improve alignment of the cylinder to the feeding assembly
in 1992. N o significant improvement resulted due to the fact that this was not the main
cause of misalignment. The more important place to address is the bush at the feeder
dosing cup which is closertothe crust to prevent twisting of the whole plunger shaft.
A design was developed at Portland that effectively achieved maximum deflections of

6 m m by use of bushes on or against the spool (76), but plant trials proved these seized
in service due to clearances tightening under heat and alumina fouling the spool. It was
concluded, therefore, that it is not practicaltoinstall a bush against the spool on A E D D
feeders without seriously affecting shot size.
Another factor that can affect seal side load and piston rod breakages in AEDD feeders
is the spool eccentricity and face. If the spool outer pipe is not exactly square, the spool
will pull the shaft to one side and cause side load, stripping of the chrome coating from
the piston rod, cushion bush failure and uneven wear of rod seal. Checking squareness
in a lathe before installation is good practice. This is discussed in more detail in
Section 2.5.8. Figure 2-17 shows the uneven wear of the spool prior to commencement
of this procedure.
8.4.5 Rod Seal Material
The recommended m a x i m u m operating temperatures of the three most c o m m o n seal

materials are shown in Table 8-II (87,88,89).
Clearly, Teflon has a better temperature rating than Viton and has the advantage of
having reduced friction. Comparison of the friction coefficient of different types of
Teflon using the S R V from the lubrication trial is illustrated in Figure 8-2. It shows
that carbon filled P T F E has half the friction coefficient of glass filled Teflon (used by
the Alcoa U S A smelters).
FIGURE 8 - 2
SEAL MATERIAL FRICTION COEFFICIENT

250 250
200 200
o
o
o
z 150 150
O
DC
LL
% 100 100
LL
LL
LU
8
50 50
CARBON BRONZE CERAMIC GLASS

COMPONENT IMPREGNATED IN PTFE
Note: SRV Analysis @50°C, 1hour, Steel roller - PTFE plate
Anderson (Atlas Copco) states that Viton seals have a m e m o r y (90). They "remember"
the hours of exposure at elevated temperature and w h e n the hours exceed a certain
critical signature time, the seals fail. Thus, a seal m a y be heated excessively over a
period of time and not fail, yet for what appears no real reason, the seal fails after only
a short period of high temperature. Anderson maintains that this explains w h y some
seals survive days at high temperature and others fail in hours. W h a t the observer is not
aware of, is that the seals have gone through a period of high temperature on a previous
period and are just about to fail w h e n the short period of high temperature occurs. If
this theory is true, then it is not just the temperature that is important, but the period
too. Failure m a y occur in days at 150°C, hours at 200°C and seconds at 250 C.
Portland, T 2 , T6, T8, TIO, D 3 and smelters using Atlas Copco and C P O A C cylinders
(Table l-II) used (or still use) Viton seals. If Viton seals are soft they can take more
heat before they lose their flexibility. Parker seals possibly fail quicker than Atlas
Copco due to Parker using harder seals. Thus, the method of manufacture, even for the
same material, can have a significant effect on seal life.
Most integrated feeders use Teflon. Based on the research of the author, since 1991,
T6, T9, TIO and Portland have changed to Teflon, with T 8 currently in the process of
changing. Improved seal life has resulted in all plants.
8.4.6 Rod Seal Design
The seal material is not the only requirement of a good seal. The shape of the seal,
lubrication type (and quantity), preload pressure at the rubbing face at room
temperature and the ability of the seals to cope with elevated temperatures are all
important factors. A seal that does not leak at room temperature m a y be too tight at
elevated temperatures and m a y erode at rapid rate due to the higher friction effect. This
was found in development of the A 3 dosing cylinder w h e n seals that did not leak in the
workshop seized the dosing unit at operating temperatures. Once the clearances were
increased between the seals and the barrel/rod, the seizing disappeared. Seal design is a
very complex skill.
Trials have been conducted at Portland on many seal designs. Parker has trio led 2 type
Terry 3 types, Norton 1 type and Dover 1 type. In addition, all Parker cylinders have
been converted to Terry front end cartridge units due to better relative life from Terry
seal units.
Figure 8-3 shows Weibull analyses for the three different types of seals used in Parker
cylinders at Portland. O n e can see a 10 fold increase in life. The reversed seals (c) are
benefited by better grease and less plunger penetration, so the effect is not all related to
the seal design. However, the original Viton polypack and Terry front end cartridge are
directly comparable as the same amount of grease was used.
FIGURE 8-3
WEIBULL ANALYSES FOR PARKER ROD SEALS

AVERAGE LIFE:
(a) ORIGINAL PARKER POLYPACK (PRIOR TO NOV.91)
360 days
100
80 ^
7
,.
50
20
^ '.'ir
B = 1 .. ( • ^
•n =»
ypjr
F o l l l ire a 5 2 1 35 i
lor 3d 1 1 .U3
IDD. »&. GM, 1MQ. 9D0B. HMD. IO
Total Opflrottng Tim* <Dayc)
AVERAGE LIFE:
(b) ADAPTED TERRY FEC (NOV,91-OCT,93) 1,800 days
100
^L ^
50 ::_^
_^r: I
20
5 B — 1 .'
P irj£
2S 5^
ire S *7
"C«r»i tor 3 d 1 e :;s
o ,7 HI. 10Q. 20B-
Tatol Operating
I MM.
(Oays)
2aai>-
AVERAGE LIFE:
(c) TERRY REVERSED FEC (NOV,93-SEP,95) 3,600 days
100
50
20
soa. IODO-
Notes: (i) Vertical axis is "cumulative percent occurred".

(ii) "Censored" referstoreplacements that were not due to failures e.g. preventative maintenance.
O n e observation in this research project and in the cylinder trials (Section 7.3.2) is the
apparent fickle nature of seals. For example, Terry F E C s (front end cartridge) work
better in Parker than Terry cylinders despite beingAalmost identical rod seal cartridge.
In fact, the Terry F E C in Parker has a shorter bush length which would result in less
support on side load support (Section 8.4.4). Figure 8-3 can be compared to Figure 7-
5(a and b) to see the effect of the same seals in two different cylinders operating at the
same speed. This shows that Terry seals had only 450-650 days life in Terry cylinders
versus 1,800 days w h e n the Terry seals were placed in Parker cylinders.
The better results may be related to use of Teflon (rather than Viton) and better greas
The cylinder trials established that Teflon was superior (Section 7.3.3:2). Tranz 414
was used for Parker cylinders instead of Magnalube. The lubrication investigation
established that Tranz 414 is better than Magnalube (Section 7.3).
Norton and Dover have achieved 4-5 years life in plants that use Alcoa designed
integrated feeders, yet w h e n placed in Portland feeders, they were worse than the
existing Terry design which achieves a one year life. This suggests that Portland
conditions are worse than other Alcoa locations. It is important that in any trial of
alternative seals in an air cylinder, all practical conditions are identical before
comparing units.
Til, T13, T16 and D12 have double rod seals in their cylinders. These are made by
Norton or Dover. At Portland, Terry have double seals and Parker have single seals.
Atlas Copco and C P O A C have one seal. T 1 7 use a single Dover seal with a four year
average life.
If the cause of failure is exposure to high temperatures, it is most likely that all s
will fail with a similar life as they are all in contact with the same hot surfaces. In the
case of the plants listed above (viz. T6, T9, T10, Portland, Til, T13, T16, D 1 2 and
T17), the best seal life is at T 1 7 which has single seals. All the plants have identical
bore and stroke cylinders using identical assemblies. Hence, it is concluded that
multiple seals are probably not worth the extra expense.
There are two c o m m o n ways to keep a pressure behind the seals to ensure they press
against the piston rod. Most commonly an 'O' ring of silicone or rubber is used, but
some manufacturers use stainless steel spring type energisers. A s noted in Table 8-II,
silicone has a higher m a x i m u m operating temperature between Viton and Teflon. If
temperatures are high enough to affect Teflon seals, it is likely that silicone energisers
will fail before the Teflon does. " O " rings m a y flatten due to thermoset, thus reducing
seal pressure or allowing air leakage between the seals and the cylinder casing. This
occurred for several runs in the cylinder trials (Table 7-1). If 'O' rings have to be used,
rectangular energisers appeartobe better than circular ones as there is more even wear
at the seal surface due to the pressure being applied over a greater area.
Use of stainless steel is becoming more common for Teflon seals, as the weakest point
in the design of most seals has been the 'O' ring type of energiser. These are used by
Norton at T17, and recent Parker designs at T 8 and trial units at Portland. This design
is recommended as it totally prevents energiser failure and the pressure of the seal
against the rod is not affected by thermal expansion which is a problem with Teflon
seals.
8.4.7 Rod Wiper Design
The wiper in the front end cartridge is designed to keep alumina out of the cylinder in
the seal area (Figure 1-20). Portland and T 8 use bronze wipers on the rod seal cartridge
and both suffer from grease breakdown.
Shell advised in the lubricant investigation that copper and iron are catalysts to the
breakdown of grease (Section 7.4.4:3). In October 1993, Portland wipers were changed
to Teflon to avoid both the pitting and grease breakdown. Since this time, seal life has
deteriorated, although there m a y be insufficient data to date to confirm a distinct
relationship (Figure 8-2).
8.4.8 Venting R o d Seal Air Leaks
A s noted in Section 8.3.1, rod seal leaks can aerate the alumina in the dosing unit or
overpressure the cavity where the feeder is mounted due to inadequate venting of the
leaking air. It is preferable to design a large vent pipe (that is angled downwards) to
prevent alumina from collecting and blocking the vent. This pipe is best vented into the
pot rather than out of the superstructure to avoid alumina spillage on a bad air leak.
Routine air leak checks on each pot will ensure that bad air leaks are not overlooked.
Section 9.4.2 explains h o w this is done.
One plant (T9) which has little problem with ore leaks has a snorkel type vent pipe so
that the air outlet is above the alumina level in the superstructure (76). This is a very
good idea and simple and cheap to fit on pots that use overhead cranes to fill
superstructures.
One plant with good ore leak history (T13) suffers from loose mounting bolts.
Although T15 shows a high rate of failure for ore leaks (80%) this is mainly from no
lubrication on Viton seals, but they also have loose flanges that not only cause ore leaks
but also arcing as the feeder moves sideways and touches the oppositely charged
superstructure. Most other integrated feeder plants that have better success with ore
leaks have an adequate vent pipe and/or better seal quality. The worst plant seen was
at T 9 where Hanna rod seal leaks were so bad that they could not even hold the plunger
retracted. These have n o w been changed to Lindberg.
A design developed as part of this research is the fixed installation of insulation wa

on the top flange with a gap for air to vent from the feeder cavity (Figure 8-4 and 8-5).
The boomerang bracket design not only solves the air vent problem but also has the
advantages of less parts (6 parts instead of 25 on the old design), is quicker to install
(less time people are exposed to heat, dust, and fumes) and prevents arcing (refer
Section 8.7.1). The air vent hole is equivalenttoa 7 5 m m O.D. vent, so is far superior
to any of the vents commonly installed in pot superstructures. The design costs only
about $40toinstall and is easy to retrofit if plants currently use insulated flanges.
FIGURE 8 - 4
B O O M E R A N G B R A C K E T S A N D F L A N G E INSULATION
INSULATION AT BOOMERANG
BOLT HOLES BRACKETS
FIGURE 8 - 5
MOUNTING BOLT AND BRACKET
AIR GAP
SIDE VIEW DETAIL X
MOUNTING
BOLT
BOOMERANG
BRACKET
BOOMERANG
BRACKET
VIEW A
STEPPED WASHER MOUNTING
(ATTACHED TO BOLT) BOLT
FEEDER NSULATION
MOUNTING NEMA GRADE G1 1
PLATE
SUPERSTRUCTURE
FLANGE
BOOMERANG
BRACKET
DETAIL X
8.5 PISTON SEALS
Until recent years, most integrated feeder plants did not realise piston seals were a
problem area. Following tabling of the author's "Feeder Recommendations" in 1992,
all Alcoa plants investigated piston seal leaks (Table l-VQ). In almost all plants, piston
seal leaks proved to be a major cause of air consumption and, in many cases, it also
severely affected feeder operation. D l l reduced air usage by one compressor by
addressing air leaks. T 6 found air leaks were a major contributor to anode effects.
The comments in Section 8.4.5 on rod seal material and temperature exposure also
apply to piston seals as the conditions are similar. Pistons usually have one or two seals
and one or two bush rings to keep the piston running parallel to the centre line of the
cylinder. It is possible to have a seal leak in one direction only, so it is important that
two air leak tests are conducted with the cylinders energised in both directions,
otherwise leaks m a y not be detected.
In the A3 trials, seizing of the piston seals occurred due to clearances being too tight
(Section 10.6.3). A small air leak at room temperature is probably satisfactory for
piston seals to ensure they are not too tight at elevated temperatures (Section 8.4).
No significant trials on alternative piston seals were conducted as part of this researc
due to its lower priority in integrated feeder failures. It is expected that further research
will be needed in the future as the life of other components increase.
8.6 BROKEN PISTON RODS
8.6.1 Introduction
Most plants surveyed indicated fatigue failures such as broken piston rods and broken
spools in (Table 1-V). Usually less than 5 % of overhauls were done for these reasons.
T17 had 2 5 % piston rod failures with Portland and T 6 both having 1 0 % piston rod
failures. Portland also had 3 0 % of spools fatiguing in 1990 probably due to excessive
speeds of the plunger. N o sign of piston rod failures has been reported for independent
feeders presumably due to the lower piston speeds.
The problem of fatigue was tackled by (i) finding the stroke time to break a hole and
(ii) finding the m i n i m u m stroke time to avoid mechanical failure. Once these were
established, either the stroke time had to be reduced to minimise fatigue, or continue
with existing failure rates.
Ultimately the hole has to be broken and, if this resulted in failure of the rod, the
would have to be accepted. Trials of timetobreak a hole at Portland showed that one
could do so at zero velocity o n a continual basis (Table 3-HI). Thus, there was
considerable scope to reduce the speed of the plunger. Hence, it was decided to
investigate slowing the piston speed to achieve acceptable fatigue levels.
8.6.2 Fatigue Analysis
Dr H u g h Stark (Unisearch, University of N e w South Wales), an expert in fatigue, was

engaged to establish what speeds resulted in fatigue (86,91,92). H e conducted a series
of tests of 14 piston rods from the two Portland cylinder types using an Instrom 8504
load tester at various stress loads from 13-60 k N at 10-20 H z (86).
Once the limits of the piston rod were established, one needed to find out what stroke
speed was necessary to fall below the critical stress level. This was then cross checked
against the stroke speed necessary to prevent fatigue of the spools. Similar tests also
found that the stroke time to prevent fatigue of spools was similar to that for piston rods
(Section 8.8.1:1). Thus a final setting was established to minimise all fatigue.
The fatigue analysis showed that Parker 10L45 piston rods were not as good as the
Terry piston rods m a d e by K e m p e Engineering from 4140 induction hardened steel.
Failure was partly related to the design of the run-out of the piston rod thread as well as
the type of rod material.
There were basically three different types of failure as illustrated in Figure 8-6. The
Terry breakage occurred at higher stress levels and with a less crescent shaped profile
on the break surface. The latter feature suggested a less brittle rod material than the
Parker rods. These types of failure are similar to that found at T 6 and T17.
FIGURE 8 - 6
SKETCH OF BROKEN PISTON RODS
BREAK
TYPE1 TERRY POINT
ACROSS FIRST
THREAD
BREAK
TYPE 2 TERRY POINT
1-3 THREADS
INTO MASS
1£*A3!
BREAK
TYPE 3 PARKER POINT
ACROSS FIRST
THREAD
A random sample of 5 Parker and 5 Terry cylinders were tested at the Portland
workshop at various stroke times to find their shaft deceleration rates under varying
speeds. Using the test data from Unisearch, it would be expected that about 4 0 % of
Terry and 6 0 % of Parker rods would fail. This closely followed plant experience with
Terry piston rods.
The surprise from the tests was that Parker rods had worse results despite a lower
failure rate in the plant at that time. Weibull analyses also showed similar results to
Unisearch. About 6 months later there was a sudden increase in failures of Parker rods.
The reason w h y the failures were not high at the time of the Unisearch tests was that the
average life of rods was low and they had not reached their critical failure life. O n e
could say it was the "honeymoon period". This experience substantiates predictive
methods as a worthwhile tool.
There was little change in average deceleration rates once a stroke time of 1.5s was
achieved as at this point the cushioning on the cylinder became the dominant factor in
speed reduction. The drop in stress was by about a factor of 3 for a change from 0.8s
(which was the existing stroke time) to 1.5s. These data suggested that a stroke time of
about 1.5s would prevent fatigue.
Note also that testing at Portland showed a higher deceleration rate going up. than do
viz. 800ms' 2 versus 500ms"2. This is partly due to the collar hitting the cup before the
spear on the top of the piston engaging the top cushion viz. 5 0 m m versus 2 5 m m from
the top respectively (Figure 1-11 and 1-20). The peak upward velocity from plant data
was l.Om/s (Table 3-1). At this speed the spool will close a 2 m m gap between the
stationary assembly housing and the moving spool cup in 0.002s. The ricochet causes
the high deceleration rate on the shaft. In a number of plants visited (T9, Til, T13)
there was evidence that there was in fact no gap at all between the cup and the
stationary feeder assembly. This metal-to-metal contact can lead to fatigue of the
feeder components and explains their history of broken rods, spool failures and failure
of piston/piston rod connections. This can be a problem in independent feeder dosing
unit where the spools almost (or definitely do) touch the housing.
8.6.3 Results and Conclusions of Fatigue Tests
Based on the above discussion, the stroke time of Portland feeders was increased in
November 1991 to 1.6s (downstroke) and 1.9s (upstroke) from original stroke times of
0.9 and 0.8s respectively to reduce the speed and hence kinetic energy of the moving
components of the feeder. The target of 1.5s downstroke was not quite achieved and a
smaller upstroke was chosen as this was not a critical factor for shot size control.
The results of the change in piston speed is seen in Figure 4-5 (for plant changeout
rates) and Figure 8-7 using Weibull predictions over time. A reduction of about 7 0 % in
piston rod failures occurred in the plant and the Weibull analysis showed an increase in
average life from 1,500 days to 2,600 days. In addition, the thread run out was
minimised and all n e w piston rods were changedtosteel grade 4140 for better fracture
toughness. The latter change had little effect.
FIGURE 8 - 7
WEIBULL ANALYSIS BY QUARTER

OF PISTON RODS
WEIBULL BY QTR REPORT
from 1 January 19H9 tD 3D September 1 395
Repair Part 421 (PISTON &0D). 521 (PISTON ROC)
1 2 100 N
5O0O
go u
L
1 4O00
\/.V \ so?
f
1
3000 Ill
so V
n 40 T
2000
0 30 I
a
1OO0- 10 u
r
IO «
0{T-i r p*"' I ' ' I ' M ' ' T' T " l " I ' ' 1'» ' I ' ' I ' M " ' • i \ | i i i1 i 'I | ~
J A J 0 J A J O J A J 0 J A J 0 J A J 0 J A J O J A J
A P U C A F l l C A P U C A P U C A P U C A P U C A P U
u T
M R L T M R L T M R L T M R L T N R L T M R tj 5 k
B 8 a 8 9 9 9 9 9 9 9 & 9 9 9 9 9 9 9 9 9 S 9 9 9 f t
9 9 3 9 0 0 0 0 1 1 1 1 2 2 2 2 3 3 3 3 4 4 4 - 4 - 5 5 5
Month Re paired
,m.= 20K * * 6 = 5 0 K * * # = 80« - = Number Failures

Months wlffi leas rhan 5 failures Ignored
Date Created: 2 October 1995
Values aver 500D days truncated
1 SLOW SPEEDS 2 DUAL DWELL TIME

Following advice from Portland on piston rod design, T 1 7 changed the piston design
and have reduced failures by 7 5 % . In addition, trials of smaller D C V s indicated
A
that slower rates have not affected hole breakage and longer piston rod life is expected.
The trial of smaller valves was expanded in 1994 at T 1 7 to more pots.
Hence, plant results verified the investigation on fatigue of piston rods. The results o
this study also questioned accepted theories, some 30 years old, that kinetic energy is
requiredtobreak feeder holes (Chapter 3).
8.7 ARCING
j\rcing occurs if there is an electrical path between the crustbreaker and superstructure
when a pot is on anode effect. Insulation is generally achieved by installation of high
impedance washers and rings between the top feeder flange and the superstructure and
similar material at the bottom of the feeder assembly. S o m e plants also specify their
crustbreaker cylinders to be insulated.
For the 11 integrated feeder plants surveyed, arcing occurred in half of them. (Figure 1
V ) . There have been no reports of arcing occurring on independent feeders due to
better insulation and less time in the bath, although the A P 3 0 has better insulation than
the A P I 8 to prevent electrolysis.
8.7.1 Insulation of the Assembly
Plants with a history of arcing usually have problems with loose mounting bolts which
allows the feeder to m o v e sideways and reduces the clearance between feeder and
superstructure.
This study identified the cause of loose mounting bolts was mechanical failure of the
insulation at the feeder-to-superstructure mounting flange. Use of N E M A grade G l 1
(at least) insulation is recommended. In addition, it is best to use large steel washers to
avoid excessive point compression that will break d o w n the insulation. Failure of
feeders due to loose mounting reduced from 30/month to under 7/month after changing
to G l 1 insulation (from G 9 ) and using bigger washers (Figure 8-8).
VO
Hi
ra
z
<
•st
H J* A3 I
w a O 1U s
i-s
PQ i E
o o TJ5 w
90 HI
p KUJO
00 m
w 5^ - 2
CL <
§ S U
> Si
o Ql "3 -< v
\* o
u x SS UJ
z in
w Z
•4
p
^£ ro
a 2
£ <"S ra
(a
xS tH
CL
a"- m
w LU Z
.-1
H
z |Utl»3
ai
w
a
w
w
o
XS
w
oao
This further dropped to under 2/month w h e n the mounting flange was changed as
discussed in Section 8.4.8 and shown in Figures 8-4 and 8-5. This design could be at
many smelters. It is cheap, simple to install and provides adequate surface to protect
insulation from mechanical damage as well as other advantages to feeder failures and
exposure of people to heat, dust and fumes.
At Portland, the bottom feeder assembly insulation cannot withstand the temperature -
peaks of over 400°C are evidenced. The $10 silicone rubber insulation bands fail within
days of installation, but, as long as the feeders are not moved, the bands continue to
provide adequate protection even when mechanically broken. M a n y alternative
materials were trialed but these were either too expensive (some over $100 each), could
not be retrofitted easily, or were mechanically weak. Those trialed included Teflons,
ceramics and ceramic sprays. It was finally concluded that -fc*s was the cheapest
solution. The insulation used by Pechiney, T 2 and D 3 cost several hundred dollars but
are almost never changed. The Portland design is a most cost effective solution,
especially if the feeder has a lower average life.
8.7.2 Cylinder Insulation
Alcoa, Alcan and several other companies specify that the piston rod and piston should
be electrically insulated from the stationary barrel and end blocks of the crustbreak
cylinders. Insulation costs about 3 0 % of the cost of a n e w cylinder. Most cylinders from
Parker, Atlas Copco, and C P O A C are not insulated.
Some plants that claim to have insulated cylinders (and specify this requirement in the
purchase of n e w cylinders) generally have stainless or bronze cartridges (T17, T8),
graphite impregnated seals (T8), or have an electrical path between block, guide, spool
and shaft (all A E D D feeders). At most plants using A E D D feeders (T6, T9, Tl 1, T13,
T15, T16, T17), the feeder is virtually touching the superstructure at the bottom of the
feeder. If the feeder is not placed exactly central on the mounting frame of the
superstructure, there is an electrical path. In more recent A E D D designs (Portland,
TIO, T7, T4), there is an insulating ring at the rim area of the feeder (Figure 1-11).
In the plants that have the worst arcing problems (T13, T9, T15), the problem is
aggravated by poor mounting bolt securing. Loosening of the feeder bolts allows the
feeder to m o v e sideways so the feeder touches the assembly at the feeder base.
Mounting of the feeder and the design of the feeder itself are the main causes of arcing,
not the cylinder insulation. So insulating the cylinder is pointless; a waste of money.
Tests were conducted at Portland on 34 cylinders when they arrived in the workshop for
overhaul. Nearly every one monitored showed no insulation capacity. After routine
overhaul with n e w seals, bushes and wipers, some 7 5 % had more than 10 mega O h m
resistance at 1,000V D C . O n e can assume that this sample is typical of the population,
and that insulation breaks d o w n in service. This was verified in the Portland/Parker
cylinder trials where all cylinders tested had lost insulation after rod seal failures despite
using insulating grease (Section 7.3.3:12).
In most cases, plants that already have insulated cylinders would find it expensive to
take the insulation out. Mostly the insulation does not fail mechanically so there is no
maintenance cost in continuing to use it. However, it is recommended that, if there is a
need to purchase n e w cylinders, it is unnecessary to specify that the cylinder must be
insulated; thus saving about 3 0 % on the cost of the n e w cylinders.
Note too that not specifying insulation in cylinders opens up opportunities to use seals
that m a y be based on graphite or other conducting materials. The friction coefficient of
graphitefilledTeflon is lower than other Teflon based material so m a y provide longer
life (Figure 8-2).
8.7.3 Other Methods to Control Arcing
Feeding on anode effect increases the time plungers are in the liquid which in turn
increases the magnitude and frequency of high voltages between feeder and
superstructure. Whatever the quality of the insulation, the least the plunger enters the
liquid and the lower the voltage of the pot w h e n this occurs, the lower the probability of
arcing. A s mentioned in Section 5.4.5, there are advantages to plunger life if the
plunger does not cycle into the bath w h e n the pot is on anode effect.
There are distinct advantagestoreduction of the chance of arcing it the plunger does
not touch the bath at all. Clearly, if no circuit is made there can be no arcing.
Choosing low plunger penetration m a y not always mean that the plunger does not touch
the bath, but it at least reduces the frequency. A s discussed in Section 5.4.2, plunger life
is far superior if it does not touch the bath or at least has minimal penetration as
corrosion is one of the dominant causes of plunger wear.
One other option to prevent arcing is to insulate the plunger from the plunger shaft (
done at T13). This breaks the circuit. This is a costly solution, restricts the use of
standard plunger materials and cannot be used with electrical end stroke sensing. For
older plants with very tight clearances between feeder assembly and superstructure this
option m a y be solution, but one should consider a more permanent solution such as
better mounting flange insulation washers or reconsider the strategy of feeding on
anode effects.
8.8 DOSING UNIT FAILURES
8.8.1 Broken and W o r n Spools
8.8.1:1 Integrated Feeders

Spools on Alcoa A E D D feeders fail by two mechanisms; wear and fatigue. Wear is
considered in Section 8.4.4, where the effect of eccentricity on rod seal wear was
discussed. This problem can be solved by use of a bush on the inside bottom of the
spool and/or by checking clearances in the assembly w h e n the spool is manufactured.
Portland has had little to no wear since these checks were initiated in 1992.
Spools of Alcoa feeders suffer wear from misalignment and inaccuracy of manufacture.
Checks on 60 spools at Portland in 1991 showed that 9 0 % of spools had wear (Figure
2-17 and Section 2.5.8). A similar trend was seen by the author when visiting T2, T6,
TIO, Tl 1, T13 and T15 in 1992. It is good practice to turn the spool in a lathe prior to
use to ensure all surfaces and edges are symmetric. It takes a few minutes but it m a y
save the cost of another spool and/or bad shot size.
Portland installed a hardened insert from 1990 to prevent spool wear (Figure 1-11).
This was later changed to bronze in 1992 as the plunger shafts were wearing. There has
been minimal wear on the spool since "hardened" inserts were installed. This design
can be installed in almost all A E D D feeders.
Fatigue of spools on integrated feeders generally occurs by a horizontal fracture at t

top of the dosing unit at Portland. Consideration of the spool fatigue failures was
carried out by use of computer stress analysis and laboratory tests by Dr. H u g h Stark
(91,92). T w o actions were taken to reduce fatigue; slower speeds and vertical webbed
sections to reinforce the horizontal plate of the spool. Stark found that both these factors
extended spool life. Testing of deceleration rates of the spool showed that there was a
marked drop of deceleration rate for 1.5-2.0s stroke time (as was found for piston rods
in Section 8.6).
Webbing was introduced in July 1991 and feeder speeds were reduced on all pots by
November 1991. Although the target was a stroke time of 2.0s, the downstroke was a
little quicker at 1.5s. The result was a median life of about 10,000 days. This relates to
virtual infinite life. Hence, the changes achieved the objectives.
8.8.1:2 Independent Feeders

Because the independent feeder spool cylinder is so small there is less speed, however,
the cushioning can be inferior. D 3 has had problems in this area. Because the
independent feeder spools generally have very small clearances between spool and fixed
seat, any error in stroke length on assembly can result in metal-to-metal contact which
can cause a considerable shock to the spool shaft fittings. It is very important that there
is a clearance between these faces but not large enough to cause an alumina leak. O n e
must also consider expansion differences between room temperature assembling and pot
operation. A tight clearance at room temperature m a y be m u c h tighter at elevated
temperatures which could lead to fatigue
8.8.2 Spool J a m m i n g
One cause of failure that occurs on integrated feeders is "not feeding". This is often
caused by complete jamming of the spool. Partial jamming of the spool can result in
inconsistent dosing of alumina that can often go undetected. Ore leaks can occur if the
spool does not seat fully after feeding the shot. Jamming can occur because of foreign
objects fouling the spool or by incorrect spool manufacture.
This problem was addressed at Portland by:
(i) better installation techniques (explained further in Section 9.6.1:2),

(ii) increasing insulation rating from N E M A Grade G 9 to G l 1 (Section 8.7.1),
(iii) drilling holes in the assembly (Figure 1-11), and
(iii) changing the mounting flange (Section 8.4.8).
Prior to mid 1990, 25% of Portland feeders failed from jammed objects (which includes
spool jamming). This has n o w droppedtoalmost zero.
Jammed objects at most plants are generally nuts, bolts and washers from the mounting
flange falling into the hopper. The reasons the bolts come undone in service is related
to insulation quality, steel washer size being too small (causing the insulation to
squash), poor cushioning or too small an insulation washer. It is also important that no
foreign objects are let fall into the alumina hopper.
For most integrated feeders, the spool is the dosing mechanism in the integrated
mechanism (Figure 1-8). For most independent feeders, the dosing unit is a smaller
device which is driven by a small (40-50mm O D ) pneumatic cylinder (Figure 1-2).
Both can suffer from similar problems.
Big improvements have been made at Portland on its AEDD feeder when holes were
drilled in the superstructure into the spring chamber (Figure 1-11). The reason for
drilling holes in the assembly was to prevent a partial vacuum occurring in the spring
chamber. This vacuum tended to drag alumina into the gap between the spool and the
assembly and lead to jamming. Holes and cutting the spool guide solved this problem
by letting the chamber breathe (Section 2.5.8).
This problem with alumina being trapped between the spool and the assembly is
probably present at all A E D D plants but is probably not detected as most plants do not
do any shot size testing on operating pots and the fact that feeders are cleaned before
they get to the repairers. There are few repair technicians w h o will not relate stories of
feeders with no faults being returned to the workshop. These are probably due to
alumina jamming the spool. This modification is recommended in all A E D D plants.
Jamming can occur on the 150mm outer pipe of the AEDD spool (Figure 1-11) or the
cone of the independent feeder because the spool:
(i) is oversize - too tight a clearance to the assembly allowing alumina to

pack in to the gap and seize.
(ii) has a bottom face not horizontal - this causes the spool/cone to rotate
vertically and foul the shaft and the outer assembly casing.
(iii) is not central - if the outer pipe/cone is not exactly on the centre line it
will foul one side.
(iv) has an outer face not parallel to centre line - this causes the spool/cone to
rotate vertically and foul the shaft and outer assembly.
Specific data on dosing unit failure of independent feeder plants was not available in
this study. However, it is likely that the above factors could be a problem with any type
of dosing unit unless the quality control of manufacture is very good. A s independent
feeders have smaller dosing spools, the faults are expected to be less than for integrated
feeders. M a n y of the problems of quality control of manufacture m a y be minimised in
independent feeders due to the use of pneumatic cylinders rather than springs. A s
discussed in Section 8.8.3, there is a m u c h higher pressure for cylinders, so many errors
can be overcome by more force. Once again, this shows the robustness of the
independent feeder designs.
Note also that most plants do not check the accuracy of their feeders. Hence, they m a y
well have a problem with their feeder unit accuracy and do not k n o w it.
8.8.3 Spring Failure
This type of failure applies to A E D D integrated feeder designs only, as independent

feeders do not have springs. Note that the Alcoa specification for A E D D feeder springs
specifies 302SS not 304SS which is more readily available. The former is a better
spring material, but only marginally so. It is unclear w h y A E D D feeders have stainless
springs as stainless is a very poor material for springs as creep is possible. It could have
been due to prevent magnetic effects, but one usually finds that stainless steels
gradually lose their non-magnetic properties after a period of time in the potrooms, so
this is not relevant. Whatever the reasoning, it has been found that steel springs work
fine in Portland and T16. Other plants are n o w changing over to steel springs from
stainless springs. They are cheaper and work better.
T8 (an integrated feeder but not an AEDD) uses the stroke of the plunger shaft to
control the dose and does not have a spring to control the dosing spool. The shaft itself
is the spool, thus reducing the feeding components and potentially increasing life.
However, the seal between the shaft and the fixed assembly has a vertical rather than a
horizontal sealing face w h e n extended, so there is more opportunity for alumina bypass.
Spring failures are reported at several plants using AEDD feeders. Springs fail by
reduction in length generally with a colour change. The shorter the spring the darker the
colour. This problem is due to creep of stainless steel at elevated temperatures; over
300 C. The higher the temperature, the darker the colour. The effects of spring length
change include:
(i) slower spool movement causing alumina bypass and affecting

shot size variability (refer Section 2.5.3);
(ii) less closing pressure on the spool causing leakage w h e n in the d o w n
position and hence larger dose size (or leakage during the complete
dwell time);
(iii) less cushioning of the spool w h e n the shaft reaches its top position,
contributing to fatigue (not normally a major problem in most plants).
Hans Kempe (Kempe Engineering) has designed a high temperature die steel spring that
provides a larger force which, in tension, is not affected up to 500°C. Figure 2-12
compares spring tension of 304SS and K e m p e springs. This shows h o w the stainless
spring used at nearly all A E D D plants (until this investigation) shrinks with
temperature. Portland, T9, T15 and T16 are in the process of changing over to high
temperature steel springs on a needs basis on overhaul.
If one compares the force of springs used on Alcoa feeders with the cylinders used on
independent feeders, the independent design is far superior (Table 8-IV).
TABLE 8 - TV
AEDD SPRING FORCES
Stainless steel spring 230kN

Kempe high temperature spring 420kN
Air cylinder 1,350kN
Basis: springs at 200°C fully extended (Figure 2-12) versus 5 0 m m O.D. cylinder at 700kPa.
Figure 8-9 shows the Weibull analysis of stainless (part number 312) and Kempe high
temperature springs (part number 321). A Weibull analysis is a log-log Y axis of
percentage failed, versus a log X axis of life. The analysis clearly shows that K e m p e
springs are superior to stainless steel springs with a 5 0 % failure rate of 3,500 days
versus 1,800 days.
T16 started to install steel springs of another design in 1992 and similarly found t
they performed superior to stainless steel.

FIGURE 8 - 9
WEIBULL ANALYSIS OF SPRINGS
AVERAGE LIFE:
1800 days
Repair Part 512 (SPRING (STAINLESS))
100
~y.
y' -
80 ,s
y
y
-y
S
50 y
y
*'
20
10
ft — i -i
22 1
J'f
^ •i
FOTIL s1 1 3C
3 =
tor 3d s 97 :r
10- 60. 10D. 200. fiOCL 1000L soon. 1000a.

Total Op«rorTr»g T i m e ( C a y s )
AVERAGE LIFE:
3500 days
R^pflirP-art 521 (.HIGH TEMP SPRING J
100
80 ..£
_-""--
50 h-->
.^
^y"
20
3"
10 4
4*
,ylf? V
ft = 1 .1> f.+
T} =S 5 2 - ^
,> •r*-
* re'i • 1 54
tor ad •• 8 £ •
10. 20. BO. 1DD. J»B. SML 1WWL aUOBL 6WB. I*HKM.
Tatol Op«rorTr»g TIFT» (Dajre)
Notes: (i) Vertical axis is "cumulative percent occurred".

00 "Censored" refers to replacements that were not due to failures e.g. preventative maintenance.
8.9 MAJOR FINDINGS OF EXTENDING FEEDER UNIT LIFE
(i) Attention to piston thread runout and piston speed reduces fatigue of piston
rods significantly.
(iii) Lengthening cylinder rod bush extends seal life by protection from side
loads.
(iii) Reusing feeders from off-line pots extends feeder life.
(iv) Having the rod seal closest to the cylinder cavity reduces temperature
exposure of the seals.
(v) It is just as important to optimise energisers behind the seals as the seals
themselves. Stainless steel is better than silicone or Viton.
(vi) In theory, Teflon rod wipers are superior to bronze as bronze wipers catalyse
grease and do not wipe the rod as well as Teflon.
(vii) It is not necessary to insulate cylinders if the crustbreaker/feeder unit is
insulated from the superstructure.
(viii) Cylinder insulation breaks d o w n with time.
(ix) Spool fatigue on integrated feeders can be reduced by webbing and slower
speeds.
(x) Attention to spool manufacture quality control will reduce wear and poor shot
size accuracy.
(xi) Attention to mounting bolts, insulation and foreign objects reduces fouling of
the dosing system of feeders.
(xii) High temperature springs are superior to stainless steel springs, but not as
good as air cylinders for dosing spool movement.
8.10 RECOMMENDATIONS OF EXTENDING FEEDER UNIT LIFE
(i) Minimise crustbreaker speed to reduce fatigue.

(ii) Reuse feeders from pots that are scheduled to be taken off-line
for startup pots.
(iii) Locate rod seals next to the cylinder air cavity.
(iv) D o not insulate cylinders.
(v) Feeder spoolstobe made to drawings.

(vi) Ensure mounting bolts are secure and insulation is evenly loaded.
(vii) For plants that have space, install boomerang brackets on the feeder mountin:
flange.
(viii) Use high temperature springs on A E D D integrated feeders.
CHAPTER 9 COST REDUCTION PAGE 328
CHAPTER 9
COST REDUCTION
9.1 SUMMARY
Statistical techniques have been used to analyze data on failure rates of parts and
components on feeder/crustbreaker units. These showed opportunities for maintenance
cost savings.
Non-traditional methods of plunger design using bimetal, high nickel alloys and short
length can reduce costs by up to 9 0 % with corresponding increase in life by up to 15
times that of cast iron.
Complete changeout of all parts every time a feeder gets to the workshop leads to
excessive costs with no improvement in life. A repair strategy based on air leak tests
can reduce costs by up to 3 0 % .
Tracking systems are not only useful for day to day control, but are critical in assessi
the effective life of any design changes and improving design in a cost effective
manner.
9.2 INTRODUCTION
Historically, on failure of crustbreaker or feeder units, smelters have routinely changed

out the plunger, all cylinder seals and any other items that were detected as being faulty.
Costs averaged about $500-700 (Table 1-IV). This Chapter questions this practice as
current maintenance strategies for m a x i m u m reliability suggest a more statistical
approach is more cost effective. Use of Weibull analyses, control charts and Pareto
charts along with an effective tracking system can identify what items need to be
changed based on their current life and the failure history of each part.
It is also important if one wants to reduce failures, to link the reason for removal with
what the tradesperson finds in the workshop. This has often not been researched, such
that modifications and design activities have concentrated on areas that are not
necessarily the highest risk or greatest failure items. A tracking system that links:
(i) where the feeder was installed,

(ii) reasons for removal ("operator observation") and
(iii) what the repairer sees ("mechanical faults")
is critical to improving feeder life.
This Chapter draws on different attitudes to the traditional method of feeder

maintenance. Main areas examined are. the use of exotic materials for plungers, use
of plungers m a d e from two different metals ("bimetal" plungers), acceptance of some
cylinder air leaks at maintenance, acceptance of air leaks on the pot, not overhauling
feeders from pots that are taken off-line, uses of an effective tracking system and
quality control both into and out of the workshop.
Irrespective of the repair philosophy used, the best method to reduce feeder costs is to
reduce the number of feeders that reach the workshop; install parts that will last the
longest time almost irrespective of cost. About 7 5 % of repair costs are for parts, so one
must maximize the performance of the critical parts, rather than concentrate too much
on labour costs.
9.3 PLUNGER SAVINGS
In Chapter 5, the causes of plunger wear were identified and it was explained w h y
different materials were trialed and their relative success stories (Table 5-IX and
Figures 5-12 and 5-13). The relative wear rates in comparison trials indicated the
materials and designs that achieved the longest life. However, the longest life plunger
m a y not necessarily be the most economic. Depending on the plant concerned,
economics m a y outweigh the life prediction, especially if one tells accountants or
managers that the cost of n e w plungers m a y change from $40 to $400. Financial
considerations have historically restricted usage of exotic materials as most engineers
had little to no evidence to justify the more costly materials.
To consider economics of different plunger designs, one needs to consider the plunger
length (or penetration depth) and whether the complete plunger is made of the same
material.
From discussion of the effect of penetration depth, Portland has tried two different
plunger lengths ( 2 5 5 m m and 3 0 5 m m ) which result in 5 0 m m different plunger depths;
3 0 5 m m being the deepest.
To minimize the cost of plunger tips, a "bimetal" design was developed where the
lower half (which gets wet) is made of expensive passive material, and the top half
(which does not get wet) made of inexpensive steel (Figure 9-1). This design was
developed with significant assistance from Hans K e m p e w h o has considerable
knowledge on metallurgy and thread designs for arduous environments. The bimetal
design almost halves the cost of stainless plungers. For example, Nicrofer price drops
from $480 to $300 for 3 0 5 m m plungers (Table 5-TX). The bimetal design is less
PAGE 331
FIGURE 9 - 1
BIMETAL PLUNGER
PLUNGER BODY
MILD STEEL
ANTI SEIZE
THREAD
PLUNGER TIP:HIGH TEMPERATURE

CORROSION RESISTANT ALLOY
economic for cast tips because of the expensive drilling costs. For example, Niresist
complete plunger costs $165 against a bimetal cost of $145 (Table 5-TX) with a steel
adapter of $90.
Table 9-1 shows detailed calculations on the cost per tip/day for 14 plunger designs of
the best 11 of the 19 plunger materials tested at Portland for different plunger length
and complete/bimetal designs. Details are given using "within-pot" trials (where
comparisons were conducted in the same pot for the same time at similar starting
diameter) and plant data directly from workshop measurements at overhaul. "Within-
pot" data is far more representative of relative wear rates than plant data as the
variability is m u c h lower due to no pot-to-pot differences. However, plant data is a
useful cross-check on limited trials. In general, the trends are similar for "within-pot"
and "workshop" data, but the plant data is more conservative with respect to life.
Three bases have been considered for the calculations of cost/tip day.
(i) Material cost only:

The cost of manufacture of a plunger is divided by the life. This price is
conservative as there are handling costs in overhauling a feeder, even if the only
problem is the plunger. Usually there are m a n y other problems that exist when
a pointed feeder is removed, as usually the feeder has been in operation for a
very long time. Most plants tend to do a complete overhaul for pointed feeders
to avoid a quick return into the shop for some other reason.
(ii) Labour cost only:

This assumes that for every plunger failure, $75 is incurred due to transport,
overhaul checks and replacement of some basic minor parts. For plants which
use contract labour for feeder removal and replacement (e.g. T6) the cost will be
m u c h higher. This compares with an average overhaul cost of $635 for Portland
and $718 for Alcoa in 1991 (Table 1- TV). At Portland, in 1994/95, about 3 0 %
of feeders pass through the workshop without major work at a handling cost of
$75 each.
TABLE 9-1
PORTLAND PLUNGER COMPARISON

WITHIN-POT DATA
(Q 305mm LENGTH
WEAR MEDIAN MEDIAN COST/TIP SAMPLE

COST PER TIPDAY
MATERIAL TYPE RATIO LIFE LIFE (A$) REPAIR COST SIZE
(mths) (days) NIL $75 $600
PA SPEC CAST 1.0 12 360 0.11 0.32 1.68 333
INCONEL 601 PIPE 1.1 13 396 180 | 0.45 0.64 1.53 20
HR SPEC CAST 1.4 17 504 45 KjTjH 0.23 1.21 139
NIRESIST D4 CAST 1.9 23 684 165 0.24 0.35 Q.90 90
HI C H R O M E (25-27%) CAST 2.4 29 864 100 0.12 0.20 0.72 49
31 OSS CAST 2.8 34 1008 120 0.12 0.19 0.62 38
INCONEL CAST 3.3 40 1188 235 0.20 0.26 0.53 30
304SS BAR 3.3 40 1188 110 0.53 22
INCOLOY 800HT BAR 270 0.19 | 0.24
BKH 28
(ii) 255mm LENGTH
WEAR MEDIAN MEDIAN COST/TIP SAMPLE

C O S T P E R TIPDAY
MATERIAL TYPE RATIO LIFE LIFE (A$) R EPAIR C O ST SIZE
(mths) (days) NIL $75 $600
PA SPEC CAST 1.6 19 576 0.07 0.20 1.06 51
304SS BAR 4.2 50 1512 80 0.05 0.10 0.42 69
31 OSS CAST 8.3 100 2988 80 0.23 100
JKS CAST 9.1 109 3276 150 0.05 0.07 0.21 26
NICROFER G3 BAR
BE9I ••j^j^ 200 0.04 H*2JT^H K X E S 126
WORKSHOP DATA
WEAR MEDIAN MEDIAN MEAN MEDIAN STD C O S T P E R TIPDAY SAMPLE
MATERIAL TYPE RATIO LIFE LIFE (mm/m) (mm/m) (mm/m) REPAIR C O S T SIZE
(mths) (days) $75 $600
ORIGINAL LENGTH
PA SPEC CAST 1.0 11 326 5.2 4.6 4.6 0.35 1.94 2225
HR SPEC CAST 1.4 15 441 4.3 3.4 5.4 0.27 1.44 2616
HR SPEC (HEAVY) CAST 1.4 15 455 4.2 3.3 2.3 0.27 1.41 19
NIRESIST D4 CAST 1.3 14 429 3.2 3.5 1.6 0.55 1.76 16
HIGH C H R O M E (27%) CAST 1.5 17 500 4.0 3.0 3.9 0.35 1.38 18
HIGH CHROME(25%) CAST 2.0 22 652 6.1 2.3 9,8 0.26 1.06 11
310SS CAST 2.1 23 682 2.7 2.2 2.1 0.28 1.04 29
INCONEL CAST 1.9 21 625 3.1 2.4 3.1 0.40 1.23 19
304SS BAR 3.8 42 1250 1.3 BEIMM 1.2 0.15 0.56 20
INCOLOY 800HT BAR 2.2 24 714 2.8 2.1 2.3 0.48 1.20 17
50mm SHORT
304SS BAR 6.6 71 2143 2.2 0.7 3.5 0.07 0.31 8
310SS CAST 5.8 63 1875 1.1 0.8 1.5 0.08 0.36 317
JKS CAST 3.8 42 1250 1.1 1.2 0.5 0.18 0.59 4
NICROFER G3 BAR 15.3 167 5000 0.4 0.3 0.7 0.16 57
Notes:
(0 In pot measurements were conducted every 3 months from September, 1992 to September, 1994
(ii) Workshop data is from January, 1990 to June, 1995.
(iii) "Wear ratio" = (median wear rate/median wearrateof 3 0 5 m m Cast iron).
(iv) Inconel pipe "metaT wear ratio was 7.0, but, due !o the Itiin pipe waits, itte "effective"fifewas a wear ratio of 1A
(v) "Cost per tip/day" = (cost per tip+tabourj/trfe in days.
(vi) Bar costs assume Portland bimetal plunger design.
(vii) "HR spec (heavy)" is a 2 0 % heavier plunger with the same penetration and external dimensions as a standard
H R cast plunger. The space between the top of the plunger and the cup was filled with cast iron. Cost was $50.
(viii)"High Chrome" is a cast iron of 25-27% chrome.
(iii) $600 repairs.
A s a sensitivity measure, one can consider higher costs of overhaul. In 1991,

only 1 2 % of failures for 11 integrated feeder smelters was due to pointed
feeders (Table 1-V). A s plants identify and design out cylinder and assembly
faults (e.g. Portland), the cause of failure will become increasingly limited by
plunger wear. This is typical for independent feeder systems such as D 1 2 which
had 8 5 % of changeouts due to pointed p\unae/b. The situation is different for
each plant. A figure of $600 per overhaul is used as a conservative estimate of
plants which have failures mainly due to pointed feeders.
Table 9-1 shows comparison costs for metal only, $75 and $600 per overhaul for 14
plunger designs trailed at Portland. Figures 9-2 and 9-3 illustrate the cost per tip/day
for $75 and $600 per overhaul. Compare this to the wear ratios and life of these
plungers shown in Figures 5-11 and 5-12.
It is clearly evident that the plunger designs with long life become more and more cost
effective as the cost of overhaul is increased. Clearly the bar materials are very
attractive both in life and in costs. This is not so evident if only material costs are
involved. For example, H R spec and 304SS are the same cost per tip/day at $0.09 if
one considers material only, yet they are $1.21 and $0.53 respectively at $600/overhaul.
Note that some materials such as Nicrofer and JKS were only tested at 255mm length.
This gives a biased improved cost over 3 0 5 m m plungers due to lower wear rate. Using
304SS cast as a guide, the cost/tip-day of 2 5 5 m m length m a y be about 3 7 % of 3 0 5 m m
length viz. $0.23 at 2 5 5 m m length versus $0.62 at 3 0 5 m m for $600/repair. However,
304SS bar was not so cost effective at $0.42 versus $0.53. This m a y be due to the more
passive material where depth is not as significant to wear rate.
Clearly, stainless plungers are more economic for plants where failures are mainly due
to plunger wear. Nicrofer, Incolloy and 304SS appear to have both long life and good
economic returns. Nicrofer is the preferred material (even if one adjusts for the benefit
FIGURE 9 - 2
CALCULATED PLUNGER COST
PER TIP-DAY AT A$75/REPAIR
WITHIN-POT DATA
fa) STANDARD LENGTH
(305mm)
PA SPEC CAST
INCONEL PIPE
HRSPEC CAST
NIRESIST CAST
HI CHROME (25-27%) CAST
31 OSS CAST
INCONEL CAST
304SS BAR
INCOLOY BAR
fb) SHORT LENGTH

(255mm)
PA SPEC CAST
304SS BAR
310SS CAST
JKS CAST
NICROFER BAR
WORKSHOP DATA
(a) STANDARD LENGTH
(305mm)
PA SPEC CAST
HRSPEC CAST
HRSPEC (HEAVY) CAST
NIRESIST CAST
HI CHROME (27%) CAST
HI CMROME(25%) CAST
310SS CAST
INCONEL CAST
304SS BAR
INCOLOY BAR
(b) SHORT LENGTH

(255mm)
304SS BAR
31 OSS CAST
JKS CAST
NICROFER BAR
0.00 0.20 0.40 0.60 0.80

$ PER TIP-DAY (@A$75 repair cost)
FIGURE 9 - 3
CALCULATED PLUNGER COST

PER TIP-DAY AT A$600/REPAIR
WITHIN-POT DATA -
-I:
(a) STANDARD LENGTH
(305mm)
PA SPEC
INCONEL
HR SPEC
NIRESIST
HI CHROME (25-27%)
310SS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH
(255mm)
PA SPEC CAST
_:
304SS BAR
310SS CAST
JKS CAST
NICROFER BAR
WORKSHOP DATA
(a) STANDARD LENGTH
(305 mm)
PASPEC
HR SPEC
HR SPEC (HEAVY)
NIRESIST
HI CHROME (27%)
HI CHROME(25%)
31 OSS
INCONEL
304SS
INCOLOY
(b) SHORT LENGTH

(255mm)
304SS
31 OSS
JKS
NICROFER
0.00 0.50 1.00 1.50 2.00 2.50
$ PER TIP-DAY (@A$600 repair cost)

of Nicrofer only being tested using 2 0 5 m m length). Nicrofer was only 1 3 % of the
cost/tip-day of P A spec cast iron for equivalent conditions despite having an initial cost
5 times that of P A spec.
It should also be noted that most of the plungers that have the longest life also are th
most economic. Thus, selection of most high nickel/chrome materials achieve all the
objectives required of a plunger. Hopefully this data will convince financial personnel
that the initial high cost of implementation of high quality plungers is justified.
9.4 Am SAVINGS
9.4.1 O r e leaks and Air Leaks

Ore leaks are commonly due to rod seal leaks on the crustbreak cylinder of integrated
feeders. The Portland Feeder Reporting System showed an 8 5 % link between ore leaks
and rod seal leaks. Plants T6, T9, T16 and T 1 7 also have traced ore leaks to air leaks
from the cylinder. Section 8.4.8 has suggested that better venting will allow the air
leaks to escape and not affect air leaks; but at what cost?
As noted in Section 1.2.16, air generation is a significant cost in a smelter. The cost
compressing air appears to be minor viz. about $40pa for 1 litre/min continuous flow
(as calculated at Portland and verified by Ingersoll Rand). The cost of a bad (150
1/min) air leak is only $200 pa if metal can be sold, or several times that if it cannot.
The average cost per overhaul of Alcoa feeders was about $718 in 1991 (Table 1-IV).
The cost of 1501/min air leak is small compared to the cost of changeout, as long as the
plant is not power restricted. If the plant is power restricted, the cost of leaking air is
about break-even with repair costs.
Changing the feeder because of air leaks is often uneconomical based on a maintenance
viewpoint, however, if the air leak results in an ore leak, it is unacceptable to let the
alumina fall into the pot unchecked. So there is no choice - the feeder must be removed
if an ore leak results.
If one accepts that it is a fact of life that rod seals will leak sometime, letting the air
vent is often an economic solution to ore leaks for many plants. It is wise to use a
suitable air leak test system to ensure that there is a cost effective choice of when
feeders need to be changed out. This value m a y differ depending on the relative value
of air versus power for each plant at any point in time.
As mentioned in Section 8.4.8, Portland has modified each feeder top mounting flange
to accommodate insulated washers (similar to that used on the mounting bolts) (Figure
8-4 and 8-5). This allows the equivalent air vent area of a 7 5 m m O.D. vent pipe. Ore
leaks almost never occur unless there is a complete blow out of the rod seal housing
from the end of the cylinder. The modification cost about $40/flange and was a once-
off cost. It reduced costs by $70,000 pa just on bolts/insulation usage alone. T o date
the only problem experienced has been alumina being pumped upwards on very bad
leaks in certain situations. About 5 0 % of flanges were modified by the end of January
1993.
Installation of a 300°C rubber skirt halfway down the feeder (below the rod seal area)
prevented air from going d o w n and alumina from going up. After trials in February
1993, full installation commenced mid 1993. Examination of data for a 6 month
period, showed that the frequency of ore leaks for feeders with skirts has been 4 times
less than feeders without skirts for situations where a rod seal leak occurs. Thus the
skirts and flange modifications effectively vent any leaking air away from the alumina.
Figure 9-4 shows that ore leaks have dropped from 50/month to about 10/month. Note
that there was a drop from February 1993 then a rise from June 93-January 1994. This
was due to a very large increase in the number of feeders being overhauled due to
installation of inserts and failures due to P A spec plungers. This had a delay effect on
the perceived benefits of these changes, but one should also inspect Figure 8-8 which
also illustrates the effect of the reduction in j a m m e d objects.
PAGE 339
FIGURE 9 - 4
ORE LEAK HISTORY
OPERATOR OBSERVATION HISTORY REPORT : C-CONTROL CHART

Fa»dera Repaired bBTwewi 1 July 1991 and 30 SBptembw 1995
Obawvcrtion Type; 105 (ORE LEAK)
2<r LlmHa
Fern=h
JUL91 0EC91 MAY92 0CT92 MAR93 AUG95 JAN94 JUN9+ N0V94 APR95 SEP95
Date Repaired
Subgroup Shear n=1
Total Number eA Feeders Repair** 7355 Date Created* 2 October 1995
Total Mumb»r 6f Tesdere wMh an Observotlom 4995 Numb#r matchIng sel»erlon: 999(20%)
Note; GDunts adjuated for Bam pie BJIB
9.4.2 Air Leak Test Method

Venting the air in a controlled manner (so that the ore leak does not occur) is a viable
alternative for integrated feeders. However, one needs a routine air leak test procedure
to ensure the air leaks do not get out of control. Equipment was invented as part of this
study so that Portland and other Alcoa plants could conduct routine surveys. Portland
n o w is the benchmark across Alcoa plants on the lowest air leak rate at 6 1/min per
feeder; a very low figure compared to most smelters. These checks are n o w being done
monthly to ensure that the air leak control is an ongoing activity.
The method to test for pot air leaks is to shut off air prior to the D C V via a ball valve.
Measure the pressure immediately, then again a period of time later - say 30s. The rate
of fall of pressure will indicate the air leak rate by use of the gas law viz.
PiV, = P 2 V 2 .
Vj is the volume of the pipes between the valve and the piston,
PI is the initial absolute pressure w h e n the ball valve is closed and
P 2 is the final absolute pressure after a period of time t.
The change in volume over time t is the leak rate. This technique is easy to implement,
costs very little (about $60/tester) and can assist greatly in minimizing air leaks in the
potrooms.
D10 had a campaign to reduce plant air usage in 1992. They found the largest air leaks
came from their 2 0 0 m m O.D. crustbreaker cylinders. They changed out their worst
ones and dropped the usage by one compressor; a saving of $200,000 pa. D 1 2
conducted a similar investigation in 1993 and also reduced air consumption by 75,000
1/min. The main cause was faulty D C V s . N o w D 1 2 plant air pressure is m u c h higher
than before.
Implementation of routine pot air leak testing at Portland has identified the main cau
were by-passing D C V s , leaks on air hoses and incorrect setting of feeder isolation
valves. Crustbreaker cylinder leakage was not a major cause.
9.4.3 Boyne Feeder Maintenance Strategy

Boyne use a preventive maintenance strategy for their feeders based on air usage
(Dunstan 69). Their cylinders (Parker) leak very badly. Average air leak rate has been
found to follow the empirical equation:
y = 0.37e017x
Where "y" is the leak rate in m3/hr and "x" is the age of the point feeder in months.
Feeders were changed out w h e n air leak rate was such that there was insufficient air
available from the plant air compressors. O n this basis, all feeders in a pot are replaced
after 22 months based solely on air usage. This relates to a leak rate of 70 1/min per
feeder (cf. 90 1/min for a Portland feeder to create an ore leak and average Portland air
leak rate of 61/min).
Boyne believed there were several attractions to this procedure:
(i) controls air leakage to a manageable level,

(ii) allows scheduling of feeder replacement weeks in advance.
(iii) the location of each feeder and its age is known.
(iv) the work can easily be contracted out.
(v) changeout procedures for improvements can be scheduled.
(vi) scheduled overhauls have led to significant price reductions
for components.
(vi) ensures uniform feeder age.
Problems with this philosophy are the lack of knowledge of the life of parts (as none
fail) and, "just to be safe", all parts need to be changed out. Overhaul costs are higher
than necessary.
It is preferable to identify air leaks on pot-to-pot checks on a routine basis and, unl
Boyne, identify which of the feeders in the pot is faulty and repair only that unit. D o
not change all for the sake of one. The other units m a y have years of life left. Boyne's
procedure is not recommended as it is based on an acceptance that severe air leaks are
inevitable.
It is interesting to note that in 1993, Boyne started trials on retrofitting Terry seal
were a design based on the author's recommendations for Portland. These seals were
successful to such an extent that by mid 1994, a decision was m a d e to retrofit Parker
cylinders with Terry seals. Once again, another plant has found Teflon seals are
preferable to Viton. Also, the same basic seals are n o w used in Portland, Boyne, TIO
and trial units at D l 1.
9.4.4 Non-routine Cylinder Overhaul Strategy

Analysis of air leaks on cylinders as they enter the workshop can be used as an aid to a
needs-basis overhaul strategy. The overhaul philosophy to only service cylinders with
an unacceptable leak, has several risks:
(i) What air leak rate at room temperature is unacceptable at operating

temperature in the pot?
(ii) H o w fast will the air leak rise in service?
(iii) Will the temperature cycle of the cylinder result in premature leaks in the
plant?
(iv) Will the tradespersons feel that management is accepting poor quality
out of the workshop and, in future, not be accountable for their
workmanship?
The last item is the one that is the most serious question, as poor attitude by
tradespersons can quickly create a major quality problem with feeders. At Portland
there has been a major improvement in the dedication and quality control of
tradespersons working on feeders; to lose this positive attitude would be most
undesirable. Feedback of the results of trials has shown tradespersons the success of
this procedure, so there has not been a problem despite initial misgivings. This is an
example of the benefits of an open attitude to discussion between tradespersons,
operators, engineers and supervisors.
A trial was conducted at Portland to see if air leaks were different at pot temperature
and at room temperature. Ten feeders had their air leak rate measured on the pot. They
were then removed and tested again at room temperature. This showed six of the ten
had higher air leaks at room temperature, none had less. During the cylinder trial, seal
leak rates increased 7 times at room temperature w h e n cylinders had cooled d o w n
(Section 7.4.3:10).
Hence, repairs of cylinders due to any air leak at room temperature m a y be inflating the
cost of overhaul. In the example above, all 10 cylinders would have been overhauled at
a cost of $600 each, despite the fact that the feeders that were removed were operating
quite satisfactorily before the test.
The average Portland feeder leak rate was 6 1/min, or $ per annum, yet Portland used
to repair 8 5 % of feeders due to "excessive air leaks" at about $600 each.
A procedure and equipment was designed by the author for the workshop using the
previously mentioned Gas L a w method (Section 9.4.2). Gauge capability tests showed
that the repeatability and reproducibility for the method was statistically capable of a
tolerance region of 0-30 1/min. The cost of the equipment was less than $200, but the
assistance it gave to cost effective maintenance was m a n y times greater..
Figure 9-5 shows a flow sheet comparing costs and life of feeders with "full" (cylinder)
overhauls and "quick" (no cylinder) overhauls. Full overhauls only achieved a 1 0 %
greater life but they did so at a cost 3.6 times that of quick overhauls. Also, the cost of
feeders on the second overhaul was not different to the previous quick/full overhaul.
Since the installation of testing equipment in the workshop in January 1993, there has
been a 3 0 % reduction in costs and little change in the life of feeders. Clearly, this
strategy has major cost attractions.
This is a great opportunity for cost savings if proper air leak management and adequate
venting is provided. Three-monthly air leak checks on the pots showed that there was
no problem with feeders having excessive air leaks in the pots despite the cylinders not
being overhauled if the leaks in the workshop were under 30 1/min. The concern of
escalating air usage in the plant proved not to be valid, and the tradespersons concerns
were put to rest. This strategy of overhaul is recommended.
FIGURE 9 - 5
CYLINDER OVERHAUL STRATEGY COSTS
AVERAGE QUICK AVERAGE FULL

OVERALL COST = $189 ea OVERALL COST = $688 ea
QUICK OVERHAULS Vs FULL OVERHAUL

* 10% LOWER LIFE
* 70% LOWER COST
N O T E : Portland feeders with Terry cylinders overhauled 1/6/93 to 31/7/94
9.5 SAVINGS FROM REUSED FEEDERS
If the feeders generally work satisfactorily the day the pot is taken off-line, then it is not
necessary to fully overhaul them. The hazards are almost the same for changing
feeders in a live pot as in a dead pot. The only difference is temperature as dust, fumes
and potential falls are about the same. For pots that have been taken off-line due to
excessive temperature, there is a risk that the feeders m a y have been overheated. It is
wise to overhaul these feeders. These pots are the exception to normal pot failures
(Section 8.4).
The practice of not overhauling feeders from off-line pots should be considered for the
following reasons:
(i) At Portland, "dead pot feeders" in 1991 cost about $420,000 to overhaul.
Alcoa spent $1 million. Probably at least half of this could be saved i.e. they
will be "average" feeders.
(ii) Protect n e w feeders from temperature spikes - so when they are installed in a
pot they stay longer (refer Section 8.4.2:4). High temperature can rapidly
reduce cylinder life.
(iii) Higher temperature of n e w pots tends to tighten up seals and bushes which
slow feeders (refer Section 4.5.2) This causes dagging of feeders on n e w
pots, blocked feeders and unnecessary exposure of operators to dust, heat and
fume.
(iv) N e w integrated feeders are slower and have increased shot size variability
(refer Section 2.5.4:3). This is aggravated by expansion of components at
high temperature.
(v) Smaller plungers (on old feeders) give greater pressure at the crust. N e w pots
have harder crust to normal pots due to lower calcium fluoride content and
higher calcium fluoride/aluminium fluoride ratio. Greater pressure will assist
in preventing blockages.
(vi) Smaller plungers tend to stop plungers jamming between the large anodes that
are in n e w pots.
Portland initiated reusing old feeders in n e w pots for 3 months in 1993 and saw no
change in feeder life. From January 1994, the practice was started again (after
installation of inserts in all pots). B y mid 1994, there was a savings of 1 0 % in costs and
a reduction in feeder changeouts by 1 0 % that could be directly attributed to this
practice. This practice has been used by Tl 1, T16 and T 1 7 with similar success.
9.6 FEEDER TRACKING SYSTEMS
A s there are m a n y pot feeders in a smelter, it is difficult to determine the link between
overhauls and failure. A major benefit of a good pot feeder tracking system is the
identification of cost reduction opportunities and the ability to follow a trial through to
completion. Often the person w h o initiates a trial does not k n o w the results, so trials
are often repeated or lapse. However, if there are no failures, one does not need a
tracking system, therefore, m a n y independent feeder plants m a y not need a tracking
system.
9.6.1 The Purpose of a Feeder Tracking System

At Portland operators changeout feeders, but, in most plants, tradespersons do all
changes in the potrooms. Overhauls are usually conducted in a workshop remote from
the potrooms by tradespersons. Hence, there m a y be up to three groups (or "links in the
chain") which m a y have little to no personal contact with each other. Each group are
often most ignorant of h o w the feeder is being used and his/her effect on h o w it
operates. The supplier and user seldom see (or even k n o w personally) each other as they
are working in quite remote locations
The "reason for removal" may appear to be totally unrelated to the "mechanical fault"
observed by the repairer as the repairer does not have the feeder at elevated
temperatures and there will be no alumina present as this is usually cleaned out before it
reaches the workshop to keep workshop housekeeping to a high standard. Once
overhauled, the feeders are very seldom checked (even at room temperature) and are
almost never checked with alumina to see if they dose correctly. Although this m a y
CHAPTER 9 COST REDUCTION p
seem a pedantic discussion, it is very important when discussing feeder failures because
the tradesperson is not aware of operational situations, and this remote sequence is one
of the flaws in effective feeder performance.
An example of this is the experience at Portland up to mid 1990 where about 30-50
feeders per month did not have any repairs conducted by the tradespersons when they
arrived at the workshop; nothing could be found wrong with them except for signs of
mechanical damage or foreign objects were jammed in the dosing unit; often
superstructure mounting bolts or washers. The operator was frustrated in what was
perceived to be shoddy work because the feeder internals were vibrating uncontrollably.
The repairer considered that the operator was incompetent as there had been no repair to
the feeder and it seldom came back a second time.
Once a tracking system was installed where the reason for removal was linked to
mechanical faults observed in the workshop, it was discovered the cause was incorrect
installation of feeders in the superstructure. If alumina was not swept out of therimof
the superstructure (Figure 1-11), the feeder sat on a bed of alumina. The alumina
gradually vibrated out of the rim area after a period of operation so the mounting bolts
became loose. This in turn resulted in the bolts falling into the alumina and
consequently jamming the dosing spool.
The operators generally correctly tightened the mounting bolts, but only the most
experienced ones were aware of the "trick" to clean out therim.Both parties were
completely unaware of that their effect on each other. Once the link was made, training
and reinforcing of the mounting bolt washers dropped failures from 35 to 25/month
(Figure 8-8 in 1989-1990). Despite the frustration by both parties, other benefits were a
reduction of $3,000/month in repair and transport costs.
It is very important that a tracking system is used to trace feeder performance in order
to link all the people involved otherwise one m a y be solving the wrong problem. U p to
1990, most feeder development was centred on finding a better plunger material despite
the fact that 8 0 % of failures were due to ore leaks. The engineers were not aware of
w h y feeders were removed. They were only aware of what wore out by looking at
feeders in the workshop, so were working on the wrong things. They should have been
working on air cylinders until they were under control, then work on plungers when the
causes of ore leaks were identified and fixed.
Thus it is important to ensure that any "reason for removal" code does not have any
words that are ambiguous and does not imply the cause of failure. For example, one
can observe that the feeder is leaking, but one cannot "observe" that the spring is faulty
as one cannot see the spring. For the tracking system for Portland the reason for
removal is called an "operator observation" and does not assume what are "mechanical
faults" which are independently reported by the repair tradespersons. This terminology
is recommended.
9.6.2 Portland Feeder Reporting System

Since 1987, each feeder overhauled at Portland has had the cylinder numbered and the
parts used listed against the cylinder number. The system was significantly upgraded in
1990 to include feeder assembly numbers, plunger numbers and location in the pots.
With the inclusion of statistical analysis, the Feeder Reporting System (FRS) is possib
the most comprehensive feeder tracking system in the world. T 8 and T17 have
reasonable systems, but not as elaborate as an engineering, analytical, accounting and
inventory tool which has operator and tradespersons directly interacting with on-line
screens. Following the author's Feeder Recommendations to Alcoa plants in 1992, all
Alcoa plants have introduced some type of tracking systems, but none as complete as
Portland.
Note that a plant with low turnover of feeders probably does not require a tracking
system as there is little short term benefit. However, any plant with a less than
acceptable maintenance budget or one that wishes to optimize feeders should seriously
consider a system similar to that at Portland.
Portland has established a computer reporting system on the process V A X computer

system. Computer screens are located in the workshop (for tradespersons) and in the
potrooms (for operators). In this way, both groups directly affected by the feeder have
an input into its performance. The F R S incorporates the following main items with the
ability to add comments if needed at every step in the process:
(i) Feeder LD.

Each cylinder and feeder assembly are uniquely identified by a distinct code
number for tracking performance of parts, costs and modes of failure. Plungers
are coded with the date of casting and the batch number of that day. Each batch
has a chemical analysis from the supplier.
(ii) Test after overhaul
After overhaul, each cylinder is tested for cushioning (by use of an
accelerometer), seal leakage (flow meter), plunger O.D. (calipers) and a number
of other minor items. Comments can be logged for the test data.
(iii) Installation
O n installation in a pot, the operator enters the cylinder number, pot, location,
date and crew. Hence, each feeder is uniquely identified in the potrooms.
(iv) Removal
O n removal, the operator identifies the cause of removal (called "operator
observations") against the cylinder number.
(v) Overhaul
W h e n the tradesperson receives a feeder to overhaul, the cylinder number is
entered onto the computer. The tradesperson is automatically shown any
operator observations, any comments, the age of critical parts and a list of
recommended parts needed to be changed irrespective of the reason the feeder
was removed from the pot. This information m a y be based on Weibull analyses
of past data. In this way, the people w h o service the equipment are aware of
the reason for overhaul and any items that need changing irrespective of failure.
This avoids early failures of items that have almost failed e.g. if partly fatigued.
At overhaul, the tradesperson carries out about 20 standard checks of the feeder
unit and lists "mechanical faults". The feeder is then overhauled and all repair
items are entered along with whether they are "failures" or "preventive
maintenance". (Failure and preventative maintenance data is used for Weibull
analyses.) Plunger O.D. is measured at overhaul. If the plunger is of acceptable
size, it is reused. In addition, if a feeder fails within 100 days, an alarm is raised
to highlight that extra attention is needed on this feeder.
From this database, a complete history is collected for each feeder (or component) of all
parts and where it was placed in the pot. Enquiries are possible by component, by pot,
or by groups of pots to see w h e n services have been done, the operator observations,
mechanical faults, comments, plunger size, plunger wear rate, cost of overhaul, parts
used, current part life, dates of test/install/remove/overhaul and operator/tradesperson
w h o completed these actions.
Reports of feeder performance, cost control and part life are available in tables and
statistical output packages viz. Pareto charts, control charts, histograms, bar charts,
Weibull analyses. This allows detailed assessment of cost and performance control by
engineers, accountants and overhaul contractors.
The database started in 1987, but system integrity improved significantly from June
1991 w h e n the P C data were transferred to the V A X system and standard reporting
procedures were satisfactorily established. Trades entry of parts was commissioned in
August 1991 and the operators in September 1991.
9.6.3 Tracking Systems and Quality Control

Control charts, Pareto charts, Weibull analyses, tables of costs and tables of turnover
rates of parts enables examination of high frequency and/or high cost items in a non-
biased manner. Use of time related graphs such as control charts indicate if changes
have achieved the desired results. There are several examples in this document viz.
Figures 4-6, 5-3, 5-10, 7-5, 8-3, 8-7, 8-8, 8-9, 9-4, 10-3, 11-2 (top) and 11-3. A great
deal of the data presented in tables has been generated from the tracking system.
Some examples of problems identified by tracking systems and quality control checks
include:
(i) After quality checks ex the workshop were initiated, new Terry piston
seals at Portland were found to be 5 0 % of the design thickness and had
been for 5 years. Checks at T 6 found the same problem for the previous
20 years using Terry cylinders.
(ii) The clearances of Terry bushes and seals at Portland were too tight and
affected feeder speed. A similar problem was seen at T16.
(iii) Spigotted flanges installed to improve alignment at Portland proved to
have no effect despite "good engineering judgment". Misalignment in the
spool had a bigger effect which was unnoticed until traced using the
tracking system.
(iv) Backup washers on Parker pistons cost more than the seals they were
installed to protect.
(v) Previous focus to reduce seal costs was changed to optimizing overall
repair costs. Seals only cost about $10 but can be responsible for repair
cost over $700. It was more efficient to pay more for seals to prevent
downstream costs.
(vi) Plunger O.D. measurements identified the wear rates of plungers and led
to better plunger materials.
(vii) Failure rates were linked to pot conditions. The major cause was the pot,
not the repair quality.
(viii) Weibull and 10/50/90% failure analyses gave early indication of the effect
of design changes rather than having to wait for total failures before
making decisions on the success of trials.
(ix) Tradespersons and operators became more aware of costs and the impact
of their jobs resulting in better quality products resulted.
These are only a few of very m a n y benefits obtained by use of feeder tracking systems.
B y far the greatest benefit is the ability to identify the effect of changes. If a
modification is installed, one can compare the benefits against a "control" sample sent
at the same time from the workshop. For this reason, all trials that are conducted at
Portland have a "control" group as well. T o ensure there is an adequate sample size for
statistical comparisons of trial versus control, any trials are conducted on 20 or more
feeders. Thus, the results of trials are quickly either (i) accelerated to production
changes or (ii) the modification is stopped. In all cases, data are used to establish the
findings rather than use subjective judgment.
9.7 MAJOR FINDINGS FROM COST REDUCTION

(i) High nickel alloys are more cost effective than less passive materials
despite the higher initial cost.
(ii) Bimetal and shorter plunger designs optimize the material cost of the
plungers.
(iii) Piston and rod seal air leaks contribute to ore leaks (for integrated
feeders) and to costly air consumption (integrated and independent
feeders).
(iv) Non-routine cylinder overhaul is a cost effective alternative to 1 0 0 %
changeout of seals and plungers with little change in mean time between
failures.
(v) Overhauling feeders from off-line pots is not cost effective.
(vi) Tracking systems are powerful tools in optimizing designs and costs.
9.8 RECOMMENDATIONS FROM COST REDUCTION
(i) Install cast J K S or Nicrofer plungers using the Portland bimetal design.
(ii) Trial low penetration plungers or drop bath levels so plungers do not get
wet.
(iii) Check air leaks often and provide air leak testers for troubleshooting.
(iv) Overhaul cylinders based on quality control guidelines that include
cylinder bypass and rod seal air leak rates.
(v) D o not overhaul feeder from off-line pots.
(vi) Install a feeder tracking system that links where the feeder was located,
reason for removal, mechanical faults and parts changed.
CHAPTER 10 FEEDER DEVELOPMENT PAGE 354
CHAPTER 10
FEEDER DEVELOPMENT
10.1 SUMMARY
In parallel with optimising the existing feeders at Portland, this research investigated
integrated and independent feeder designs. Three feeder designs were invented by the
author and two have been successfully patented.
The A2 integrated feeder proved to be unsuccessful. The Pulse Chute did not progress
past the prototype stage due to interest in the A 3 independent feeder design. The A 3
feeder incorporates the advantages of the independent design but can be retrofitted on-
line to m a n y potlines that have integrated feeders at about half the cost of retrofitting
traditional independent feeder designs. A cheaper Sequential Feed design has been
developed that is a compromise between the A 3 and traditional Alcoa A E D D feeder
designs.
The new feeder designs incorporate many features that enhance long life and ease of
maintenance. T o date, operating results on the trial pots are encouraging.
Comparison of integrated and independent feeders using data and observations from
this research shows that the independent feeders are superior in almost all parameters
except capital cost.

10.2 PROJECT HISTORY

From early 1990, development commenced on feeder designs that achieved long life,
direct feed and reduced maintenance costs. B y mid 1990, the A 2 feeder was invented
(Appendix 6) with trials in the plant commencing in February 1991. Patents have been
approved in Australia, N e w Zealand and the U S A . Improvements to the shot size
accuracy of the existing A E D D feeder (by development of the spool insert shape) was a
byproduct from this development.
Late in 1991, the Pulse Chute was developed and reached prototype standard by late
1992 (Appendix 7). A s the A 3 has significant benefits over the Pulse Chute, the chute
did not proceed past the prototype stage. The A 3 was invented late in 1992 (Appendix
8) and plant trials started late in 1993. The modifications in mid 1993 to the assembly
of the existing A E D D to stop jamming, was a direct result of the A 3 development.
Figure 1-22 shows all these designs and illustrates that the overall dimensions are
similar so they can be installed in pots where A E D D s are currently installed. This was
one of the original objectives. Table 10-1 shows a comparison of the different feeder
designs against features that are important for feeder or pot operation. Table 10-11
shows comparison plant scale data of the trial feeder designs compared to A E D D pots
over the same test period.
From mid October 1993, trials commenced on seven pots at Portland. This was a
"disaster check" to see if these designs had any major drawbacks. Realising the
variability in pot operation, a trial of one or two pots of different design is never
sufficient to prove or disprove a feeder design or pot operating strategy; a trial of about
10 pots would be required. However, a small trial would indicate if the results were
very good or very bad and would assist in tuning equipment and electrical components
and to debug software. Based on this small trial, it was envisaged that a larger trial
would follow w h e n funds were available.
Below is an explanation of the designs, costs and benefits (to date) of feeder
development at Portland.
1-
Z
LU
Q _^
co Z o- co CO CM CO CO CM m
< LU 4^ d 0- c^-
UJ T— CM d d d
Q-
UJ m
Q
Z
_J Q
< UJ
QUENTI
AEDD
CO CO CM in d CM
& LU - c^- o
o d >-
d •
CO d
UJ LU
1- d
CO Z
H
>-
Z
UJ LU
Q CO CO
z CO co CO in CM CD T—
i 1
LU
Z UJ "*" o
o QL
d >-
CM CM m
UJ LU d CJ)
0_ Q
Z
a
LU
HH
H
-
H
§o co
m
d
o
z
m
•<t
csi
d
CM
d
CM
d
co
Oi
<
Z
LU
H
Z
H Q
•ORTLAND
UJ
EXISTING
PQ h-
AEDD
"tf
m o in
en <
- - d
Q o d z CO csi CO z
LU a>
w LL 1-
z
w •4J 2 >.
c cu cu
H cu
TJ
TJ
CU TJ
>>
ro
c
!?r
(trofit Peel
c TJ c TJ
O cu o cu
CD
ro cu
design
O N
a. o o 'co •c Q.
OH 1- J2 CD
a> ro o o
co
CM O) B
TJ ro
c a. £ x:
a. o
<
1- o > o
cu 3 !t
1— TJ TJ sco
O C
M—
o
CO cu
CU K%
CO CU CU TJ
CU >o cu 2%
I— 0 s 4—I
ro CQ ro TJ
CD TJ cu
cu O >
CU -Q ro ro
_c
E
«>
c 4-f
cu
<
Q
h- UJ UJ
_J 1-
LL O >-
LU r-
co
r-
O < < t
LU
CD
LU
o 2
z o UJ z
a.
oS
;IZE STANI
JIOT
Q CO <
MINT
mLU o I— CO or CO gLU
ION (OFF
h-
z Z)
TS
DC LU O OJ 00 LL
Q DC o LL
O O LU UJ ;* H LL LU
LU 2 z 3
CO
< LU
LL
< = o
LU DC LU
oc CO < LL. LU z tr H o_
o >
Z)
\- z< Ul
z r- z CO CO CO H UJ LU <
d O
LL UJ
DC
<
LU
LL. o
X
LU
UJ
LL. <
r-
z
O < r-
Ul fc o
^ X
Q
LU
LU
H <
O >
X UJ
§
UJ
Q
O
Z
DC
DC
Z)
t
Q.
<
DC D
h U
Ul LU
2 -J 2 Q ? co LL. CO o £ < O O DC LL
CHAPTER 10 FEEDER DEVELOPMENT P A G E 357
TABLE 10-H
FEEDER TRIAL DATA

PARAMETER A2 SPLIT HEADER SEQUENTIAL
FEED A3
(1075,1085,2069) (1022,1094) (2075) (2070,2071)
JULY-AUGUST 1994
ALUMINA IN POT
- PERCENTAGE LOW
- STANDARD DEVIATION GOOD
ANODE EFFECTS PER DAY OK OK GOOD FAIR
ANODE EFFECT DURATION TIME OK OK OK OK
PRODUCTION RATE
- m m TAPPED PER DAY OK OK FAIR OK
-TONNES PER DAY OK OK OK OK
APRIL-JUNE 1995
ALUMINA IN POT
- PERCENTAGE LOW LOW
- STANDARD DEVIATION GOOD GOOD
ANODE EFFECTS PER DAY GOOD FAIR
ANODE EFFECT DURATION TIME FAIR GOOD
PRODUCTION RATE
- m m TAPPED PER DAY OK GOOD
-TONNES PER DAY OK GOOD
Notes:
(i) "GOOD/OK/FAIR" rating based on where pot rated against a range of 28 similar pots. " G O O D " means the
pots were in the top 2 5 % of 60 pots in the first period and 28 pots in the second period. Similarly, "FAIR"
were in the worst 2 5 % and "OK" were in the middle range.
(ii) The lower the concentration of alumina in a pot the better the current efficiency but the harder it is to control
operation. " L O W concentration would generally indicate better production rate but potentailly
more tendency for anode effects.
(ii) Between the first and second comparison period, the plungers on the A 3 feeders were increased in length
to increase penetration which m a y be a cause of previously average performance on anode effects.
(iii) Production rate is expressed by two independent methods, " m m T A P P E D P E R DAY" is the distance the
anodes moved during tapping. There is a linear relationship between m m tapped and production rate. This
the most accurate method of assessing production rate. " T O N N E S P E R D A Y " is measured by the crane
that hold the crucible. This parameter m a y be affected by any bath that is mistakenly tapped. Over a long
period and by using comparison data for a group of pots, these data can be used for production rate.
(iv) In the second period, there are no data for A 2 pots as all were changed back to A E D D design. The split
header pots were not considered duetotheir average performance in the first period.
(v) Duetothe small sample size of trial pots, these results can only be used as an indication of trends.
Excursions on such a small sample size can greatly affect comparison results. A much larger sample size
for an extended period is requiredtocompare feeder designs.
10.3 A2 FEEDERS, DRIBBLE FEED AND PULSE CHUTE
The A2 feeder is an integrated feeder that looks like the AEDD, but feeds when the
plunger is up, not d o w n (Figure 1-14). From late 1990, up to 8 pots have been
operating with A 2 feeders. Performance has been quite variable from pot to pot (until
mid 1993) due to fouling of dags on the single feeder chute (shown in Figure 2-1). In
mid 1993, the chutes were changed to thefinaldouble outlet design (Figure 2-17) and
performance improved with results comparable to normal pots. Overall, the benefits of
direct feed, accurate shot size and smaller shot size have not been evidenced possibly
due to tuning factors in the feed control logic. It is also possible that any improvements
to performance are not sufficiently large enough to be seen. Thus, the cost of retrofit is
not justifiable.
The main operating problem with the A2 feeder is that any dagging can foul the chute,
then the pot is starved of alumina. It was decided late in 1993 that this design was not
advanced enough to continue trials. These pots are n o w being converted back to A E D D
as the pots come off-line.
Trials were conducted on a dribble feed continuous feeder chute in 1990-1991. This
proved unsuccessful due to blockages of the small orifices from tramp material. This
material either came from the alumina or was drawn up the nozzles from the pot via
dags. Problems with blockages are c o m m o n in continuous feeders and were a major
reason that other companies also dropped continuous feed. Comalco continues trials on
a pot with continuous feed at Bell Bay. T o date the trials have not progressed past a
one pot trial despite the opportunity presented by the current expansions at Boyne
Island and N Z A S .
Given that continuous flow is prone to blockages, the alternative is to pulse the alum
in short shots of a larger quantity. The Pulse Chute distributed a metered dose of
alumina on an intermittent basis over time (Figure 1-18). In this way, a semi-
continuous feed can be achieved, yet the problems of blockages were not a factor. This
could be achieved by an airslide with 5-10s pulses of aeration. A s it would no longer

be necessary to have very small orifices to control flow, large outlet nozzles could be
used to distribute the alumina thus avoiding blockages.
The design was taken to prototype stage but then dropped in late 1992 due to interest in
the A 3 which was considered a better alternative. The pulse chute m a y also have
problems with alumina sifting into the air plenum when there is no pressure in the
plenum. Although purge holes are part of the design, this is still of concern. Although
patents were applied for world-wide for this design, these applications were withdrawn
in early 1994 due to the high cost of renewing applications for a design that was
unlikely to be taken into production. However, the design of the chute outlets assisted
in the development of the double outlet chutes used for A 2 and A 3 feeders (Figure 2-
17).
10.4 MULTIPLE DCVS PER POT OPTIONS
One major problem with the existing Portland AEDD feeders is that five feeders
operate from the one valve. This results in:
(i) lack of air when all feeders operate, leading to

- poor shot size accuracy,
- different feeder speeds and
- dwell time time-out before the slowest feeder releases dose and/or
breaks a hole,
(ii) time delay across the pot for feeders to operate, resulting in long dwell times
and subsequent dags,
(iii) high air consumption as m u c h air is used tofillthe air line eg. 4 0 % at
Portland.
10.4.1 Split Header

If the existing air header is split in two, then two valves could be used to minimise the
effect of the above factors. This design is called a split header (Figure 10-l(a)).
The two valves operate at different times. Rather than a feed every 3 minutes, a feed
could occur in half that time interval somewhere in the pot. If any mixing occurs across
the pot, then the pot alumina concentration would be more consistent over time. There
is usually a great deal of mixing across a pot due to the agitation of the bath and metal.
Two designs are under trial. One has both valves on the wall of the building and one
has the valve on the pot itself (with the exhaust piped into the pot gas cavity) to
minimise air usage. Results to date from the two pot trial suggest there is not a
significant improvement to pot performance for this design.
10.4.2 Sequential Feed

Sequential feed design is the provision of a separate D C V for each of thefivefeeders so
that each feeder can operate at a different time across the pot (Figure 10-1(b)). This
gives more consistent alumina concentration in the pot over time, allows more
repeatable feeder speed, less dwell time and more accurate shot size.
As with the A3 feeder, the only changes necessary to the pot are installation of electr
across the top of the pot and modifications to the P L C program. The impact on
production is zero for a retrofit of this design.
Shot size accuracy is similar to the A3 feeder (Section 2.4.5). This design has many of
the benefits of the A 3 but at a cheaper cost.
There is only one pot currently on trial. Results show good anode effects per day and
good control of alumina in the bath. A larger sample size is needed to gauge pot
performance. The hardware has been faultless up to date viz. 20 months.
FIGURE 10 -1
MULTIPLE DCVs PER POT FEED OPTIONS
A. SPLIT HEADER 2 DCVs/POT

a existing feeders
„ 2 or 3 feeders operate at one time
existing air header split in two by new valve
EXISTING
DCV NEW
1 ,DCV
-5PLUG 0\ 0 l 0• Q Q EXISTING
WALLBOX
B. SEQUENTIAL FEED 5 DCVs/POT

a existing feeders
„ one feeder operates at one time
DCV
" 0 0 0 0 0 --
WALLBOX
FIGURE 10 - 2
INDEPENDENT FEED OPTION
FEATURES
• five crust break cylinders operate at different times
• crustbreak and feeding occur at different times
• new five feeding cylinders (with hole) under existing cylinder
• feed cylinders operate sequentially across the pot eg. 1,3,5,2,4
• each cylinder can be turned off if required eg. after set, hole mucky
J3 FEED
g DCV
PLUG
O O O O Q FEED CYLINDER
CRUSTBREAK
WALLBOX DCV
CRUST
O00O© BREAK
CYLINDER
10.5 A3 FEEDER
The A 3 independent design is believed to achieve most (if not all) the original
objectives of this research project (Table 1-m and Section 1.3.1). Shot size tests
showed excellent repeatability and control (Section 2.4.5 and 2.6.4). Pot installation
included the sequential operation of feeders across the pot so that not all are fed at the
same time (as is the case with Sequential Feed) and there was a delay between
crustbreak and feed (Figure 10-2) so that the alumina flowed into the open hole. This
section will discuss the results of the A 3 trials to date and its design features.
10.5.1 A3 Pot Performance

The A 3 was invented, scoped and commissioned by the author. Matthew Langmaid and
Colm Fitzpatrick (Mechengineering Services) designed the feeder on C A D and
precommissioned the prototypes. All three people worked as a team to develop the
final design. The author appreciates the skill and cooperation of Matthew and Colm in
achieving what has proved to be a well designed piece of equipment.
Although there have been several problems with some of the dosing cylinders seizing
due to too tight seals/bushes (but never on Parker cylinders) to date (1.11.95), no
mechanical faults have occurred to any part of the A3 design itself or on any crustbreak
cylinders on any of the 10 feeders in almost 2 years of operation. This is most unusual
for Portland feeders as 2 0 % were failing in 3 months and 5 0 % failed in about 3-4 years
when installed at the same time. O n e would have expected 2-5 to have failed by now.
It is even more unusual considering the extra equipment added to the feeder over the
standard A E D D .
Two pots were changed over from AEDD to A3 feeders in December 1993. Results to
date (Table 10-11) indicate the A 3 pots has very good control of alumina concentration
and standard deviation of concentration in the bath. Median metal production is in the
top 2 5 % of pots, but the anode effect rate is higher than normal.
This may be because the concentration is running close to anode effect all the time.
This is desirable for metal production (which is evidenced by the data) but tends to
increase anode effects per day. It m a y also be due to some sticking of the dosing
cylinder for some designs. There has been some problems for some cylinder designs.
Operators often cannot see any problems w h e n the pot has anode effects. Hence, it
could be lean operation or cylinders sticking (or both). Further investigations are
needed.
It is pleasing to note that, even though the frequency of anode effects may not be in the
top 2 5 % of pots, the time in anode effect is small. Once again, if the heat balance was
poor, then the metal production rates would not be in the top 2 5 % of pots. A s time in
anode effect was small, the generation of potentially global warming gases would be
less than normal pots.
As was mentioned with the split header and sequential feed trials, the sample size is too
small to be specific on the ratings of these trials. O n e needs a large sample size to make
any definite conclusions. N o problems have occurred with the feeder design itself.
To date, it appears most of the predictions on the A3 design have proved correct with
the small sample size available.
The main eauipona-tr- problems have been with seizing of the dosing air cylinder at
elevated temperatures and some valve failures (Section 10.7). Cylinder seizing was due
to insufficient clearances on the piston seals on some brands of trial cylinders. There is
still a concern that this sticking problem m a y still exist intermittently and m a y be the
cause of higher than normal anode effect frequency.
10.5.2 Possible Use in Other Smelters

The basic design of the A 3 is to cut a section out of the existing assembly and replace it
with a feed cylinder that has a hole in the centre of the piston rod - a "hollow cylinder"
(Figure 10-3). T h e cylinder operates the dosing unit instead of the spring/collar
operation of the A E D D . Figure 1-15 illustrates h o w it operates and Figure 10-3 shows
h o w the feeder looks compared to an A E D D feeder.
FIGURE 10 - 3
A3 AND SEQUENTIAL FEEDERS
• CYLINDER
CRUSTBREAK
CYLINDER^
FEEDER
DOSING _ ASSEMBLY
CYLINDER
DOSING
DOSING AREA
UNIT
PLUNGER
PLUNGE
& SHAFT
A3 A3 AEDD
ASSEMBLED DISASSEMBLED
AEDD AEDD A3
SEQUENTIAL
FEED
The A 3 design was developed to incorporate the benefits of the independent feeders,
but utilise Alcoa integrated feeder equipment. A s such, the design can be retrofitted in
15 plants using Alcoa feeders - Portland, T2, T3, T4, T5, T6, T7, T9, TIO, Til, T13,
T14, T15, T16, and T17. It could be installed in T 8 with some modifications.
Cylinder dimensions have been chosen such that the one design of cylinder will fit in al
plants, thus reducing the effective cost for mass production, allowing c o m m o n spares
between plants and ensuring the results of trials conducted at Portland are directly
transferable to other plants for quick implementation.
Estimates to install A3 independent feeders in pots at Portland (at 5 feeders per pot)
suggest that the capital cost would be about half that of retrofitting conventional
independent feeders, mainly because the cost of superstructure is m u c h lower on a
retrofit and significantly lower on a n e w pot. At this time it is not known if the cost is
justifiable, but current trials at Portland are afirststep to quantify the cost benefits of
independent feeders over integrated feeders. If there is a cost benefit, then the A 3 is a
cost effective alternative in order to retrofit or to install in n e w pot. In addition, the A 3
can be retrofitted on-line, whereas the integrated feeder needs to have the pot off-line
due to significant superstructure modifications.
10.5.3 A3 Design Features

The A 3 has several attractive design features that have been incorporated from
experience of the A E D D development and A 2 performance.
(i) Spigotted Flanges

The assembly design incorporates spigotted flanges at every join and bronze
bushes are used on moving parts to minimise misalignment. Normal Portland
A E D D deflection of the plunger is 2 5 m m side to side; the A 3 feeder achieves
5 m m which is comparable to other independent feeders (Table 8-III).
(ii) Shot Size Accuracy

Shot size accuracy has been proved to have a standard deviation 3-4-Ws better
than the existing feeders (Section 2.4.5 and 2.6.4). These results m a y be similar
to that achieved by other independent feeders as the cylinder sizes are similar.
However, the mass flow design of the A 3 should give superior results.
Although not mandatory, the trial A3 pots have double outlet chutes rather than
the normal single outlet Alcoa design. This is to give more direct entry of
alumina into the hole. This m a y give better results than the A E D D dust loss
(27) and achieve benefits claimed by Reverdy (26) that direct feed is good for
dissolution (Section 1.2.5).
(iii) Maintenance
The design was developed with maintenance in mind. It is possible to remove
any section of the unit with minimal effect on remaining parts. For example, it
is possible to service the dosing unit without disassembling the crustbreak or
dosing cylinder. It is possible to do air leak and piston bypass checks without
major removal offittingson the top flange.
(iv) Ease of Installation

The top flange equipment (including the valves) is no higher than the top of the
superstructure to avoid any possibility of being hit by an overhead crane. N o
modifications are necessary to the superstructure other than to run conduit and
wiring for the electrics to the valves. N o air line changes, oxy-cutting or
welding are required. These features m a k e it easy to retrofit in operating pots.
(v) Cost Control

It is cheaper to install the A 3 feeder than the other independent designs, as there
is less superstructure work required. The costs of valves, cylinders, electrics,
programming and feeder assemblies are probably similar for the two design
types, but changeout and maintenance costs are less for the A 3 as it is more
compact and easier to service.
In summary, the A 3 feeder achieves at a lower cost the features of independent feeders
viz. low air usage, direct feed, feed and break at different times, immediate crustbreak
pressure, low dwell time, sequential feed across the pot, direct control of dosing unit,
accurate shot size. In addition, the A 3 feeder can be retrofitted in existing smelters with
little to no effect on operations. Thus it is a cost effective competent design.
10.6 END STROKE SENSING
One cause of anode effects in every smelter is blocked feeder holes. If the hole is
blocked, no alumina feeds the pot and an anode effect results.
As mentioned in Section 1.2.12 and illustrated in Figure 1-19, some plants have
developed sensors that detect if the plunger touches the bath by measuring a voltage
across the feeder (52, 53). These require the plunger to get wet to operate. It has been
established that it is not necessary to get the plunger wet to break a hole (Section 3.4
and 5.4.3). M a n y plants (and individual pots within most plants) have a zero plunger
penetration in the bath. This stops the voltage sensing design from working. Dags on
the plunger which cause extra resistance for the voltage signal, also stop the sensors
from working. Trials of this type of cylinder at D l 1 also had problems of arcing out
some electrics if the crustbreak is extended into the bath during a high voltage anode
effect.
At the request of the author, Atlas Copco has designed a mechanical fitting in their
cylinder that will sense the end of the stroke. The piston hits a button in the bottom
block of the air cylinder. This sensor sends an air pressure signal to an electric switch
on the top flange. This, in turn, tells the computer that the cylinder has reached the end
Us
of stroke.
A
Although end stroke sensing does not inform the operator/computer that the hole is
"open" (as such), it does indicate the crustbreaker is fully extended. Provided the crust
is not significantly below the plunger tip (which is usually the case), then the design
will indicate blocked feeder holes. Liquid level control ensures that the total liquid
level does not get too low. Usually, if the crust is low and the hole blocks, the material
will build up over the hole and will quickly prevent the crustbreaker from reaching full
stroke and the alarm is activated. Hence, this design is better than the designs that need
to wet the plunger. At worst, there is only a minor delay before the blockage is
alarmed.
A timer is set at a limit above the normal stroke time of the cylinder. When this time is
exceeded, an alarm is activated by the computer to tell the operator to check the pot. In
this w a y a blocked feeder is detected immediately it happens. This minimises the time
an operator is exposed to the hazards of unblocking holes and removes the cause of
anode effects well before they occur.
Another advantage of the design is a reduction in time that the plunger is wet. Instead
of using a computer dwell time to control w h e n the plunger is raised, the plunger is
raised w h e n it has positively proved to have travelled full stroke. For pots with
multiple feeders per D C V , w h e n all have reached full stroke, all are raised.
For independent feeders, smarter logic can be used. For example, if a blockage is
detected, the feeder can be turned off and the other feeders in the pot feed more
frequently until the blockage is fixed. The crustbreaker can cycle several times to clear
the blockage. Only w h e n the blockage is not cleared after, say, 4 attempts, does an
alarm inform the operator that the problem can not be fixed by the computer. This type
of feature is a very positive w a y of addressing anode effects and the health issues
associated with dags.
Both the A3 feeder and split header pots have end stroke sensing facilities installed in
their control logic and the A 3 feeders have end stroke sensing facilities on the
crustbreak cylinder itself. Trials are underway with end stroke sensing on A E D D ,
sequential feed and A 3 feeder pots, however, to date the system has not been fully
commissioned as the author w a s transferred from the Potrooms in mid 1994. It should
only take a few days to commission it. Despite this lack of interest on use of this design
feature, there have been no mechanical failure of cylinder components in almost 2 years
of continuous operation of 20 cylinders. Although the electrical signals and alarms are
not activated, the mechanical equipment has worked without fault.
It is hoped that this installed design feature which has great potential will be fully
commissioned and tested.
10.7 VALVE AND CYLINDER DESIGN
These trials operate different types of cylinders and a mixture of different valves in t
same pot to give an objective comparison of the different brands for identical pot
conditions. This avoids pot-to-pot or time-related variables.
Atlas Copco have a proven record for DCVs on many plants (Table I-IT). Mac have an
equivalent D C V at half the cost. The top flange of the A 3 and sequential feeders have
provision to use both types interchangeably. Half the valves trialed were Atlas Copco
and half were Mac. The Atlas Copco valves have had no faults in thefirsttwo years of
operation, versus 8 failures for M a c valves.
Four different brands of cylinders are being tested - Parker, Terry, Ortman and Kempe.
In addition, K e m p e have designed a different feed cylinder such that there is no alumina
bypass on the dosing unit. The latter feeders were removed from operation by mid
1994 due to constant attention to pneumatic controls. Initially there were problems
with most types from seals being too tight. O n e needs to accept a little air leakage at
room temperature as seals tighten at elevated temperatures (Refer Section 8.4 on air
leaks at room temperature versus that at operating temperature). Once this w a s
understood and modifications m a d e only two cylinders (Kempe and Terry) have failed
in service after 20 months operation. This has been very good performance for such a
radical design of cylinder. Parker and Ortman cylinders have never had any problems.
Parker and Ortman feed cylinders and Atlas Copco valves should be used for future
pots.
10.8 INTEGRATED VERSUS INDEPENDENT FEEDERS
Table 10-1 compares the main features of the different designs examined in this study.
Table 10-HI summarises the previous references to integrated and independent feeders
to verify conclusions on the relative benefits of each design. It has been established that
independent feeders are superior to integrated feeders in almost all aspects except
capital costs.
It is clear that the independent feeders are more accurate, provide a more repeatable
dose, have lower air usage and have potentially longer life. However, the capital cost is
higher. T h e benefit o n current efficiency and anode effects are subjective as no
comparative studies have been published. Although nothing is published on pot
benefits of independent feeders, dissolution studies show potentially better mixing by
direct feed (Section 1.3.13) as well as the better shot size accuracy verified by Portland
trials (Section 2.4.5). O n e would expect benefits to pot operations. W h a t is unknown
is if these are significant enough to justify three times the capital cost.
It seems inconsistent of the reputation of companies such as Pechiney, Hydro

Aluminium and Reynolds to install independent feeders with poor economic return for
all their installations since 1979. Although no data are published, it appears that these
companies believe the higher cost is justified.
The current trials at Portland have the potential to identify the benefits of the differe
feeder designs. O n e of the features of the trials is that each pot can be changed to
different feeder schemes by turning a switch. In this way, pot-to-pot differences do not
affect comparisons. These comparisons m a y take several years to complete due to
financial restrictions on the number of pots that can be trialled. Conclusions would
come m u c h quicker if more pots were brought on line.
Results to date after 2 years operation show that the A3 has superior performance to
normal A E D D feeder pots with respect to metal production, but have a higher than
normal frequency of anode effects (Table 9-VI).
CHAPTER 10 FEEDER DEVELOPMENT
PAGE 371
TABLE 10 - m
ADVANTAGES AND DISADVANTAGES OF

INDEPENDENT FEEDERS
FEATURE SECTION
FOR
1 DONT NEED TO BREAK A HOLE EACH FEED 1.2.5, 1.2.6
2 LESS AIR USAGE 1.2.5, 1.2.6
3 REDUCED DUST LOSS BETWEEN CHUTE AND HOLE 1.2.5, 1.2.6
DIRECT FEED - BETTER DISSOLUTION 1.2.5, 1.2.6, 2.8
4 - LESS CRATER 2.8, 3.3(vii)
5 - HORIZONTAL VELOCITY VECTOR 2.8
6 LONGEST LIFE AIR CYLINDERS 1.2.15
LONGEST LIFE FEEDERS 1.2.15
7 - CONSISTENT FEED TO POT 1.2.15
8 - LOW MANTENANCE COSTS 1.2.15
8 SPOOL ANGLED TOP & BOTTOM - GOOD FOR SHOT SIZE 2.3.2
MORE ACCURATE SHOT SIZE S.D. 2.4.5, 2.6.4
10 -DCV HAS NO EFFECT ON SHOT SIZE 2.4.5
11 -DWELL TIME HAS NO EFFECT ON SHOT SIZE 2.6.4
12 DAG SCRAPER 2.8, 5.4.7
SLOW PISTON SPEEDS 3.4.1,8.3.4,8.6.1,8.8.12
13 -LOW FATIGUE 3.4.1,8.3.4
14 -LONGER SEAL LIFE 3.4.1,8.3.4
VALVE NEAR CYLINDER 3.4.1,4.6
15 -HIGH PRESSURE AT CRUST 3.4.1
16 -REPEATABLE DWELL TIME 4.6
17 -NOT AFFECTED BY CYLINDER VARIABILITY 4.6
18 LOW NUMBER OF ANODE EFFECTS PER DAY 3.5.2, 10.6
19 LESS EFFECT FOR MUFFLERS BLOCKING 3.5.3.3
LOW DWELL TIME AS NOT WAITING FOR ALUMINA 35.8,5.4.1:1
10 -LESS DAGS 3.5.8, 4.6
LARGE STROKE DISTANCE 6.5.6
12 - COOLER SEALS 6.5.6
13 - LESS PLUNGER PREHEAT 6.5.6
14 NO ORE LEAKS (44% OF INTEGRATED FEEDER FAILURES) 8.3, 8.4.1
NO SPRING 8.3, 8.8.4
15 - LESS JAMMING OF SPOOL 8.3, 8.8.4
16 - REPEATABLE SPOOL STROKE SPEED 8.3, 8.8.4
17 LESS ECCENTRICITY - LONGER SEAL LIFE 7.3.3:4, 8.4.4
18 NO ALUMINA NEAR SEALS - NO INGRESS 8.4
19 BETTER INSULATED 8.7, 8.7.2
20 NO NEED FOR TRACKING SYSTEM AS LOW FAILURE RATE 9.6
AGAINST
1 NO PREHEATING OF ALUMINA 1.2.5, 1.2.6
2 ABOUT 3 TIMES THE CAPITAL COST 1.2.5, 1.2.6
3 NO DIRECT FEED IF FEED EVERY 2-3 TIMES 2.8
4 DEEP PENETRATION - NEED COSTLY PLUNGERS 6.5.5
5 SOME SPOOL SHAFT FAILURES 8.8.12
6 USE VITON - NOT A HIGH TEMPERATURE RATING 7.33.2
This m a y be due to a lower than normal concentration of alumina in the bath. The time
in anode effect suggests that the pot is on the edge of the optimum operating
concentration thus assisting production rate. A s there are only two A 3 pots operating at
present, one cannot m a k e too m a n y conclusions on comparative trials, however, to have
both pots with metal production in the top 2 5 % of pots is very encouraging.
It is the author's opinion that the benefits of independent feeding would only have a
cost benefit for large pots viz. 180kA or larger. This is due to the accurate dosing
achievable on 1-2 feeder systems using a c o m m o n D C V (as explained in Section 2.4).
It is unlikely that the air savings and marginally improved dosing accuracy will make a
difference to the pot economics on smaller pots.
CHAPTER 11 IMPACT OF RESEARCH ON ALCOA AND OTHER SMELTERS PAGE 373
CHAPTER 11
IMPACT OF RESEARCH ON ALCOA AND

OTHER SMELTERS
11.1 SUMMARY
The research conducted by this study has resulted in benefits to Portland and many
other smelters world wide. The recommendations made by the author to Alcoa plants
in 1992 and developments by teams at all Alcoa plants resulted in almost double the
feeder life and half the cost between 1991 and 1994. Portland alone has reduced costs
by $1.4 million per annum.
Feeder development has been a major enabler to halving anode effects across Alcoa
since 1991. This development also addressed safety and health improvements and
involved operators and tradespersons in the implementation of design and procedure
changes.
11.2 INTRODUCTION
The impact of the improvements to existing feeder designs on feeder life, costs and
anode effects are summarised in this Chapter. The importance of designs to assist the
health of people w h o operate, change and service feeders is included.
11.3 PORTLAND RESULTS

Figure 11-1 illustrates the changes m a d e at Portland to improve the A E D D feeder. A
more detailed history is in Appendix 9. In the next sections is an explanation of the
results of this activity.
11.3.1 Plant Performance

The objectives of the research into optimising feeders were three fold:
(i) reduction in feeder changeouts (for health benefits),

(ii) reduction in maintenance costs (for plant economic return), and
(iii) reduction in anode effects (for global warming and pot stability reasons).
Reduction in anode effects is the end result of the other two.
Implementation of this research at Portland has had slower results at Portland than at
several other plants despite using the same ideas of improvement. This has been due to
the more difficult conditions at Portland as shown in Table 11-I. The combination of
higher temperature, 5 feeders/DCV, deep penetration, poor cylinder design, grease
breakdown and high kinetic energy resulted in a frustratingly slow change in
performance.
From mid 1993, a step change occurred in feeder failures (Figure 11-2). The spike in
mid 1993 was a result of smaller shot size from the inserts and the poor choice of P A
spec cast iron that actually reduced plunger life.
a
1
2
o —
to • • -
< J
• •' "•
-1
"•" ' •
o 1
s
* 1
H s —
o> LL 1
o> -
w a Hi
Hi
z
o
60
HI
OH H • —
_ . .
4
-> [
-1
2
<
s
oo LL
O) -t
a
z.
o
CO
<
-1
1
^
S
<
5
CN
U-
Oi -j
o
z
o 1
a)
T— <
CJ ->
->
2
SPOOLS
W POTS)
WASHER
RHAULS
TERRY
TS
rs
LU K
tr >}IU
o
Pz O CO
< ccS LU Z|
m < LU H CO CO 1 X fe
to l-CO Szl^Sco
^ rf *
Q. < _ Q Oi CCui
W 0. iO
s
, a:
? co 5 S m< LU ULU"; ,t I- LU
(_ *- LU K CC K UJ= . z fc!
fe«u- ft(3(jO OS * LU O CO
3 ° K
< ? ^
CL CC II uj 2" o
I-»EOI
CO CO
llllil o -<
jOOUO O l-
co o
J I O u- < to a. 1
a to z LL co
PAGE 376
TABLE 11 -1
PORTLAND VERSUS OTHER FEEDERS

PORTLAND Vs OTHER INTEGRATED FEEDERS
DIFFERENCE MECHANICAL FAULT CAUSE OF FEEDER

FAILURE
P O R T L A N D IS H O T T E R (60°C A V E R A G E ) Cooked seals Ore leak
Springs shrink Ore leak
Plunger preheat Pointed
Dags
5 FEEDERSA/ALVE N O T 2-3 FEEDERSA/ALVE Variable response
Longer dwell time Pointed
Dags
Variable shot size Anode effects
High air volume Air usage high
Sensitive to faults Anode effects
Localised high gas Dags
Pointed
ORIGINALLY VITON S E A L S N O T T E F L O N High wear rate Ore leaks
Requires grease
Grease breakdown Ore leaks
Muffler blocks Blocked feeders
Valve clogs Air leaks
Blocked feeders
PORTLAND Vs INDEPENDENT FEEDERS
C A U S E O F FEEDER
DIFFERENCE MECHANICAL FAULT
FAILURE
P O R T L A N D IS H O T T E R (60°C) see above see above
5 FEEDERSA/ALVE N O T 1 FEEDERA/ALVE see above see above
2-3 TIMES T H E C R U S T B R E A K C Y C L E S 2-3 times lower life of seals Ore leaks

2-3timeslower life of plungers Pointed
2-4 TIMES T H E P L U N G E R D W E L L TIME Pointed

B y late 1994, the failure rate had dropped from the 1993 average of 230/month to under
50/month. B y mid 1994, feeder changeout rates were about half the record lowest
month ever. With a normal life of about a year, it would have been expected that there
would have been a jump in failures by mid 1994 due to the high changeout rates of mid
1993 when changeouts were accelerated to install spool inserts and remove the poor
wear rates of P A spec cast iron. This did not happen. In fact, failures dropped to under
30/month which is the level required for a 5 year feeder. The changeout rate continued
at about this level right through 1995. Thus, the improvements were working.
However, graphs of frequency of failure may give a false impression of performance.

For example, if all the feeders were rapidly replaced, then there would be a period
immediately afterwards when the frequency of failure would be small; the "honeymoon
period". In fact, the performance could be no different to normal.
Similarly, failure rate analysis may be affected by the acceptance criteria of when to
changeout feeders. For example,tfa supervisor decided to change the standard of
pointed feeder from a minimum of 5 0 m m O.D. to 4 0 m m , then there would be a rush of
changeouts that would have no bearing on changes made at repair prior to installation.
FIGURE 11 - 2
PORTLAND FEEDER CHANGEOUT HISTORY
Feeder Repair Rate: C—Control Chart

Date Repaired between 1 January 1991 and 31 December 1995
Component Typei (ALL)
Iff Limits
For n=1:
400 -
£
"2 200 UCL=16S-7
il
C= 144.7
LCU=12Qu6
1DD
1
| • • • i i i i i i i ' i ' ' ' ' ' I I I i ' • • • • " •
JMJ91 JUL91 Jfct|J9a. JUL92 JAW93 JUL95 JANS4 JUL94 JANS5 JUL95
Date Repaired
FEEDER LIFE (YEARS)
1991 1992 1993 1994 1995

A better analysis of the effect of changes in design or operation is to examine the life of
feeders from the date installed, not removed. Figure 11-3 shows the life of 2 0 % , 5 0 %
and 8 0 % of feeders from date installed. This shows a steady median life at about 200
days through to 1991 before quality control actions were initiated. Life rose to 300
days in 1992 but dropped in 1993 back to the original life when the smaller shot size of
inserts and poorer life of P A spec cast iron started to have effects. From mid 1993, life
rose rapidly due to short stainless steel plungers, better grease and holes in assemblies
(to stop jamming of the spool).
Between 1992 and 1994, anode effects at Portland halved partly due to better hardware
and troubleshooting that was developed by this investigation. A lot of work was done
by others on computer control (e.g. Fethon Nahoum), operator attention to detail and a
lift in awareness of the importance of solving excessive anode effects for improved pot
control. Feeder improvement was significant, but was not the only cause. It was an
enabler. Costs and changes dropped accordingly to record low levels.
By the end of 1994, all capital and expense costs for this project have been covered by
savings. Figure 11-4 shows the annual changes to frequency, costs and anode effects
per day for Portland.
11.3.2 Improvements for Health
Aluminium smelting pots have a range of hazards that can affect the people who attend
them - heat, dust, fumes and falls (1,72,73). M a n y of the jobs are affected by strong
magnetic fields that tend to m o v e equipment away from the intended direction. The
magnetic fields are so strong that they clear the magnetic signature on credit cards and
can hold a 25kg crow bar in an almost vertical position. Spanners and wrenches can
stick onto fixed equipment and two hands are required to release them. Thus,
movement of tools is very awkward and can affect the safety of doing what appears to
be simple jobs.
2 => £ -0 v »- i*-o .— a •_ w n
>
—I
CO
CO
3
co ^
UJ
ry
« O
S x
co <d
E
z co
cog
TO Eg
4 ||
3 Q.
LLto
C O
CD O LU
LU -J
>
C > a. z
c 2 <
3 O
IO
? •-
CL co
0
co
o Q
(V
tn Q
O 0 O O C O O O
z K S 8 S K M •»-
0 _| .- H- * r- £ £) O *t IA
X
wy
CO
u.
0
H
O
t,
5
FIGURE 11 - 4
PORTLAND FEEDER PERFORMANCE - SUMMARY
OPERATING FEEDERS ANNUAL REPAIR COST (Aim)
1991 1992 1993 1994 10
NUMBER OF OVERHAULS COST PER OVERHAUL (AS)
0 -: — ^ — ^ t / M m . ^_
FEEDER LIFE (YEARS) C O S T P E R T O N N E O F M E T A L (A$)
A N O D E EFFECTS PER D A Y
Note: (i) Anode effects per day "est 1995" is the estimate for 199S based on the average for Janua(y-June,1995.
(ii) "Improved feeder life will support quantum leap improvements in anode effects, power consumption,
and safety. In addition, longer feeder life offers the opportunity to substantially reduce maintenance costs."
R E . Seymour, 16.4.92.(99)
(iii) These "research activities have been a key enabler to achieving the reduction of anode effects per cell day
in the international locations and, consequently, the significant reduction of perfluorocarbon emissions"
R. R. Taybr,25.7.94.(100)
PAGE 382
Throughout the development of improvements to the existing feeders or n e w feeders,

the designs have been heavily influenced by the people w h o do the job - operators w h o
changeout the feeders and tradespersons w h o overhaul them. S o m e of the items
developed to assist people included better feeder racks and lifting slings. O n e example
of a design that helped operators and tradespersons as well as improving feeder life and
costs was the design of better mounting flange fittings.
A common cause of feeder failure is jammed objects (Table 1-V). This is often from
loose feeder bolts, nuts or washers. The real cause of this is the manner of installation
and/or insulation breaking down. In addition, the nuts, bolts and washers are difficult
to handle in the strong magnetic field, so dropping objects into the alumina in the
hopper is not uncommon. The job of changing feeders is a hot, dusty, fumey operation,
and operators/tradespersons are exposed to fall hazards from heights up to three metres.
It is not a pleasant job.
An alternative design was required that would
(i) allow minimum time on the pot for changeout,

(ii) be simple to modify existing feeders/superstructure,
(iii) have less objects that m a y fall or have to be carried to the top of the
superstructure,
(iv) be non-magnetic if possible and
(v) be cost effective, especially if 2,000 feeders were installed.
Many mounting designs were trialled at Portland...Huck bolts, overcentre clamps, barrel
clamps, G clamps. All these alternatives were expensive and had weight or fatigue
problems.
The author also wanted to find a way to get better venting of air from the feeder to
prevent ore leaks that are caused from aeration of alumina and break d o w n of the seals
at the bottom of the dosing unit. Eventually a simple design was developed of a fixed
insulation washer on the flange (at each of the four bolt holes) with boomerang brackets
(Section 8.4.8 and Figure 8-4 and 8-5).
The result was a once-off modification that solved the installation problem at low cost.
The time required on the pot has more than halved and the number of parts reduced to 2
brackets and two bolts (which had the washers fixed to them) instead of the 28 parts in
the old design. The $50/flange cost ended up saving about $70,000 on parts per year as
almost all items are n o w reused.
Not only has the cost reduced, but the big attraction was that the design allowed any air
leaks from the cylinder to vent from the feeder without causing ore leaks. Ore leaks are
the most c o m m o n cause of failures of integrated feeders (Table 1-V) and a contributing
cause of failures for independent feeders. A s explained in Section 8.3.1, the main cause
of ore leaks is a leak from the rod seal of the cylinder aerating the alumina nearby and
causing alumina (ore) to fall into the pot unmetered. Since the boomerang brackets and
fixed insulation was installed at Portland, ore leaks have almost disappeared (Figure 9-
4). N o w only excessive cooking of seals causes ore leaks.
As well as the change to the flange fittings, non-magnetic lifting slings and racks to
hold feeders safely in a vertical position assisted the operators from having to force
tools against the magnetic forces and gave them a stable place to store and remove new
feeders without any hazards. These racks also were used to store all nuts, bolts and
tools for the changeover job,. The racks contained covered boxes to store installation
and removal tags for the tracking system. Thus, all items required to change feeders
were located in easy reach of the person doing the job.
Hence, astute asessment of the process and people's needs have solved hazards as well
as saving costs.
11.3.3 Worker Involvement
In this project, operators and tradespersons have been directly involved both in
development and in normal maintenance. Operators and tradespersons have unlimited
access to all screens in the computer tracking system, so they take an interest in the
history of feeders.
Tradespersons were frequently consulted on h o w things were going and what they
thought of any trials. They often identified opportunities and were a key indication of
trends before data were sufficient to identify the trends. Although they were not always
correct, it was useful data to keep management and engineers on track. They also
appreciate that they have ownership of, and are a valuable input into the process and are
rightly proud of improvements. It is interesting that, even when their jobs were at some
risk as failures dropped, there were only positive statements from tradespersons of h o w
well "we" have done. It was a team effort. This is the only way to get sustained
improvement.
Modifications for health developments have been warmly received by operators and
tradespersons. They can see their job is safer and easier. This also reflects in a more
positive attitude to a job that is often boring, hot, dusty and fumey. A s operators can
input w h y they removed the feeder, they k n o w they have an influence on the process.
The success of feeder development at Portland (and other smelters that have followed
Portland's conclusions and recommendations) has been largely affected by the
involvement by these operators and tradespersons. Their contribution cannot be under-
emphasised.
11.2 ACTIONS AT ALCOA SMELTERS
As discussed in Section 1.3.4, in 1992 the Alcoa Primary Metals Quality Manager Bob
Seymour (who was in charge of all international and U S A smelter quality improvement
teams) directed that each plant follow the advice of the author on feeder improvements.
His letter to all Smelting Managers is in Appendix 4 (99). Following the Kissane
Feeder Recommendations (Table 1-VH), action was taken at all Alcoa plants on some
or all of these items. Figure 11-5 shows 1991 to 1994 data from all Alcoa plants, and
Table 11-11 shows highlights from these plants.
Across Alcoa there was a reduction of 44% in changeouts and reduction of 48% in
costs....over $ 2 m savings. Anode effects dropped by 7 0 % (Figure 11-6).
CHAPTER 11 IMPACT OF RESEARCH O N ALCOA AND OTHER SMELTERS PAGE 385
FIGURE 11-5
ALCOA POT FEEDER PERFORMANCE - BY PLANT
OPERATING FEEDERS ANNUAL REPAIR COST (A$m)
NUMBER OF OVERHAULS COST PER OVERHAUL (A$)
B1991
• 1993
• 1993
0199*
• 1994
• 1995
ill 9 Jt- HCL
FEEDER UFE (YEARS) COST PER TONNE OF METAL (A?)
-
Q1994
a 1995
Tl—'
(__ •£••'••}• - 1-- .zn — ( 1

TABLE 11-H
COMMENTS FROM ALCOA PLANTS ON FEEDER DEVELOPMENT
PORTLAND - increased wear by 30% when reduced shot size

- dropped plunger life by 4 0 % on change to P.A. cast plunger
- plunger life 5 times better for short 310 cast stainless and 15 times better for
Nicrofer
- increased frequency of changeouts to address A.E's
- not changing feeders on off line pots
- air leak rate average is <oy/r>vn/feeder...3oi/mlrVp0t
- rapid improvement in failure rates from start 1994
- costs dropped by $1.5m per annum by 1994 and failures by 8 5 %
D12 - air leak audits found bad DCVs air consumption reduced by 2500 cfm
- piston rod failures in 1994 due to loosefittings...changingto rolled threads
- increased length of plungers in 1994 due to need to lower bath levels
T6 - dropped plunger life by 40% on change to P.A. cast plunger
- increased frequency of changeouts to address A.E's and pointed feeders
- air leaks and hoses were major problems
- repeaters - 3 % of total had 3 0 % of changes.
- drop in failures in 1994 due to improved cylinder components from 1992
(Portland spec seals)
T7 - 40% reduction in failures from 1992 due to establishing procedures and training
- team to address feeder operation on a daily basis
- improved life by improving cushioning
T9 - dwell time halved
- early failures reduced by m o d s to guides and shafts
- Hanna cylinders replaced by Lindberg
- testing alternative plungers
- extra costs in 1993 from m o d s to assemblies and cylinders evidenced results
from 1994
T11 - troubleshooting team to address anode effects found most causes were feeder
problems
- mufflers blocking caused extreme back pressure. A new muffler had about
100kPa of back pressure whereas an old/plugged one was about 350-400kPa.
Installing new mufflers from 1995.
T13 - insulated plungers stopped arcing
- less problems/overhaul
- s o m e plungers were changed to add 5 0 m m to the length in 1994 due to
reducing bath level
T15 - dwell time halved
- a Dover seal (similar to T17) w a s sourced but not implemented
- late 1994 dropped bath levels and increased feeder changeouts accordingly
T16 - retrofitted rod seals, piston seals, wiper and springs
- improved communication, tracking, testing into workshop and procedures
- testing alternative seals from two suppliers
- comparing failured feeders and ones from off-line
CHAPTER 11 IMPACT OF RESEARCH ON .ALCOA AND OTHER SMELTERS PAGE 387
FIGURE 11 - 6
ALCOA FEEDER PERFORMANCE - SUMMARY

OPERATING FEEDERS A N N U A L REPAIR C O S T (ASm)
NUMBER OF OVERHAULS C O S T PER O V E R H A U L (AJ)
Note: (I) Anode effects per day "est 1995" is the estimate for 1995 based on the average for January-Jun«U995.
(il) "Improved feeder life will support quantum leap improvements in anode effects, power consumption. _
and safety. In addition, longer feeder life offers the opportunity to substantially reduce maintenance cost*
R E. Seymour, io.4.a£.v*'/
(iii) These "research activities have been a key enabler to achieving the reduction of anode effects per celday
in the international locations and, consequently, the significant reduction of perfluorocarbonOTissons^ ^
The results of some plants were comparable to Portland. T 7 for example, reduced
turnover by 8 2 % and costs by 9 0 % . T 1 6 had the highest life at about 5 years. Most of
the plants that had poorer performance in 1991 (Portland, T 6 and T9) had to invest in a
rebuilding campaign to obtain control of performance, but showed evidence of the
projected benefits by early 1995.
Less exposure of operators and tradespersons to dust, heat, fumes and potential fall
hazards has been a pleasing benefit. A very important feature of the investigations has
been the extensive involvement of tradespersons and operators. This had a significant
impact on results.
A better understanding of what causes anode effects has resulted. This has reduced the
variability of feeding so that computer changes can fine tune the pots. It is considered
that feeder improvements were contributors to the reduction in anode effects across
Alcoa according to Dick Taylor (Director of Technology, Primary Metals, Alcoa)
(Appendix 5). Taylor also verified the impact of this investigation on the
"improvement in the design, control and functionality of the alumina feeder" and that
these "research activities have been a key enabler to achieving the reduction of anode
effects per cell day in the international locations and, consequently, the significant
reduction of perfluorocarbon emissions" (100).
Experience has shown that two of the 10 recommendations did not achieve what was
expected. Spool inserts proved to be beneficial on the large feeders at Portland, but
were not as significant at smaller plants. The P A spec cast iron plungers (used at
Portland and T 6 ) proved to give worse life than the previous H R cast iron. Short 310SS
plungers were recommended to all plants in mid 1993 but Portland is the only Alcoa
plant which has changed to non-cast iron plungers to date, despite limited trials in T 6
andT9.
The other eight original recommendations proved very successful. All plants found that
blocked feeder holes were mainly due to covering after set, not plunger speed. Dwell
time was halved at T15 and T9. D 1 2 saved 250cfm of air by addressing leaking D C V s .
Portland, T 1 6 and T 7 used their tracking systems and procedures in a very proactive
manner and identified m a n y problems that were never appreciated before. Except for
Portland and T6, most plants did not proceed with dual dwell time due to computer
restrictions.
Clearly, not all the results are due to the Recommendations per se, but concentrating on
feeder hardware (which had not been done as intensively before) showed many ways
that the plants could be improved. The key to any further improvements and cost
cutting is an effective tracking system. Not m a n y plants had this area under control.
B y 1995, most plants had difficulty in even doing a Pareto chart on w h y feeders were
removed (though most had lists for what was done in the workshop). Portland has
found that the pot conditions are usually the major reasons for failure...not the way the
feeder was repaired. It is important that the cause is traced and a good tracking system
is required to do this.
There is still considerable scope for plants to make significant savings in feeder
maintenance. Installation of stainless steel plungers, use of lubrication, air leak venting,
using breathing holes in the assembly and addressing optimum plunger depth show
great promise in m a n y plants (both Alcoa and non-Alcoa). Note that use of cast
plungers is unwise from an economic point of view and cast plungers have potential to
create dags initially due to imperfections in surface finish. A n y plunger trials should be
comparative trials against a control plunger in the same pot, for the same time and
starting from the same original diameter.
Although a lot of progress has been made, there is still a wide gap between the 2 year
Alcoa feeder life and the 5-6 years average claimed by most independent feeder plants.
11.3 ACTIONS AT NON-ALCOA SMELTERS

Close contact with other smelters has resulted in many actions which have been
significantly affected by this research. Some items are listed in Table 11-IH.
TABLE 11-ID
ACTIONS AT NON-ALCOA PLANTS
T2 universal joint for reduced misalignment between piston

rod and plunger shaft
T3 alternative plungers
independent feeder trials
T7 change in piston rod shape to stop fatigue

trials on slow plunger speed to stop fatigue
co-operative trial with Portland of alternative seals
T8 retrofit of Terry FEC in Parker cylinders
T10 implementation of Terry cylinders

implementation of Terry F E C
changed from routine feeder changeout to repair on
failure
D3 alternative plungers
do not break every feed
D11 implementation of Terry cylinders
In addition to the above, Terry Fluids are installing their cylinders at Capral (formerly
Alcan) at Kurri Kurri, NZAS (Comalco in New Zealand), Comalco Bell Bay, Boyne
Island and Dubai (United Arab Emirates) as a result of the seal research at Portland.
The cost/benefits and extent of life improvements are not available, but comments by
representatives of these plants testify to significant improvement in feeder life.

CHAPTER 12 SUMMARY OF RESULTS PAGE 391
CHAPTER 12
SUMMARY OF RESULTS
12.1 RESEARCH OBJECTIVES
From early 1990, this research into feeders has considered two basic objectives
(Section 1.3.1):
(i) optimise existing feeder designs (especially the AEDD feeder),

(a) reduce failures
(b) reduce anode effects
(c) reduce costs
(ii) compare integrated and independent feeder/crustbreaker units and

develop a state-of-the-art feeder which has the following features.
(a) life of over 5 years,

(b) cost effective in both capital and maintenance costs, and
(c) can be retrofitted in integrated pots easily.
Consider the results of this research against these objectives.
12.1.1 Feeder and Pot Performance

Chapter 11 illustrated that, compared with a 3 year average from 1991-1993, in 1995
the Portland results were as follows (Figure 11-4):
(a) feeder life increased by 83%,

(b) anode effects reduced by 6 8 % , and
(c) feeder repair costs reduced by 8 2 % or $1.4m per annum.
From 1991 to 1994, Alcoa results were as follows (Figure 11-6):
(a) feeder life increased by 52%

(b) anode effects reduced by 7 2 % , and
(c) feeder repair costs reduced by 4 8 % or $3.4m per annum.
Note that the reduction of anode effects was not solely due to feeder life/performance
improvement but to m a n y actions taken by others. Feeder research was a significant
enabler to this improvement (as per Taylor note (Appendix 5)). About 1 0 % of the cost
reduction was due to the reduction of feeders on line because of the global reduction in
aluminium manufacture in 1994-1995. However, the costs in 1995 are considered to be
much lower than in 1994 as a result of implementation of major changes that inflated
costs at several smelters in 1994. Plant benefits are the result of sustained efforts by
many at each plant. This research helped direct actions to the most appropriate areas.
Although data is not available, it is believed that there have been significant
improvements to non-Alcoa smelters because of this research.
12.1.2 Feeder Development
Compilation of information in literature, data and comments from plant contacts and
data developed in this study (Table 10-111) showed that the independent feeder is
superior to the integrated feeder in almost all aspects except capital cost. The
development of the A 3 and Sequential Feeder designs achieved most of the benefits of
the commercially available independent feeders yet at reduced costs and they could be
retrofitted into integrated feeder plants on operating pots.
Thus, the objectives of capital cost and ability to retrofit were attained, but it is sti
early at the time of writing to identify if life and maintenance costs are adequate for
these designs. Until a larger sample size of test pots is placed on line, it will not be
possible to gauge feeder life and the effect on pot performance for these designs.
12.1.3 Conclusions on Achievements Against Objectives
The objectives of this project were met with respect to optimising feeders...reduced
failure rates, reduced anode effects, reduced repair costs. Feeder designs were
developed that have the potential to be cost effective and can be retrofitted in operating
pots. Thus, the objectives originally set were largely achieved.
12.2 FUTURE WORK
There are several ideas that can achieve cost savings as well as improve life of existing
feeder/crustbreaker units in all smelters (whether integrated or independent):
(i) stainless or alloy plungers

(ii) small penetration in the bath
(iii) reuse feeders from off-line pots
(iv) special attention to repeaters and feeders with high cost repairs
(v) air leak tests on overhaul
(vi) air audits in potrooms
Optimisation of the integrated feeder at Portland is basically on "hold" from 1994 as the
recent changes to grease, guides, seals, bush wiper, inserts and plungers take effect.
Most of the above points are long term design changes, so time is required to collect
enough data to measure effectiveness before further changes are made.
The main direction of future development of feeder schemes at Portland will be the
testing of the n e w feeder schemes; sequential feed, A 3 feeders and end stroke sensing.
The second stage of an additional ten A 3 pots rests largely on the results of these trials
and the availability of m o n e y in the current poor metal price situation. There is
considerable interest by operators and technical personnel in these ideas, but these will
only progress if there is a clear picture of enhanced pot efficiency. Unfortunately, this
is difficult to see unless a large number of pots is operational, due to the large
variability in pot performance.
At other Alcoa smelters, there are n o w teams of tradespersons, engineers and

supervisory people continually optimising their feeders. Trials or implementation of
m a n y items raised in this report are continuing at other non-Alcoa plants, especially for
cost control in the current climate of poor profitability of aluminium production.
12.3 CONCLUDING REMARKS
O n e item that B o b Seymour (99) mentions in his note to the Alcoa Smelting Managers
in 1992 should be stressed in assessing the results of feeder optimisation work.
"Improved feeder life will support quantum leap improvements in anode effects, power
consumption, and safety. In addition, longer feeder life offers the opportunity to
substantially reduce maintenance costs."
This needs to be stressed. The main target for this work in smelters is not a
maintenance activity, though a lot of the work to achieve the target improvements is
intimately involved in feeder repairs and feeder design. It is very easy to measure the
priorities, success and failure of feeder research on reduction in maintenance costs. The
real measurements of success are feeder life, shot size accuracy and ability to deliver
the dose to the bath. These are the main objectives of pot feeders. If these are
achieved, then other computer and process people can optimise the pots knowing that
the alumina supply is consistent. Reduced maintenance costs are side benefits.
Feeder operation is a critical item to the control of anode effects and pot performance.
It is no use having the smartest computer feed logic if the correct alumina quantity is
not delivered effectively into the bath. Feeder costs are significant both in capital and
operating aspects. For these reasons, feeder development is very important for the
industry.
It is hoped that this study has made some contribution to addressing the challenge of
optimising alumina feeders in aluminium smelting pots.
REFERENCES PAGE 395
REFERENCES
1. Isaken, I., "An Assessment of the Role of CF4 and C2F6 as Green House
Gases", University of Oslo, Centre for International Climate and Energy
Research Oslo (CICERO), 1992, Table 4.7, p24.
2. Kissane, J.P., Portland Aluminium (Alcoa) "Apparatus for Controlled Supply
of Alumina", Patents AU-B-77582/91, NZ-240101, U S 5,324,408, 5.10.90.
3. Kissane, J.P.,Portland Aluminium (Alcoa),"Alumina Supply Apparatus for
Electrolytic Smelter", Patents S A 93/5050, NZ-253652, AU-662829, U S
5,423,968, International Application PCT/AU93/00332, 14.7.92.
4. Kissane, J.P., Portland Aluminium (Alcoa), "Method and Apparatus for
Continuous Supply of Alumina", Patent application, PCT/AU91/0016829.4.91
5. Prider, R T . (editor) "Mining in W A " . U W A Press, 1979, pp 218-228 and pp
100-108
6. Grojotheim, K and Welch, B.J. "Aluminium Smelter Technology".
Aluminium, Verlog G m b H , 1980

7 Grjotheim, K , Krohn, C , Malinovsky, M., Matiasovsky, K., Thonstad, J.,
"Aluminium Electrolysis", 2nd Ed, Aluminium - Verlog G m b H , 1982
8. Naussbaum, A.L, "Aluminium Smelter Technology for the Nineties", Light
Metal Age, Feb 1993, pp 8-77

9 Fredman, E.M., Merken, M.A., Baltacka, P.M., Neketen, U.M., Tenlakow,
F. K , Dnepropetrovsky Metallurgical Plant, "Machine for Servicing
Aluminium Electrolysers", Patent, S U 1301875, 7.4.87.

10 Libidiw, A.F., Bybnow, N.E., Friman, E.M., Zaharow, B.F., Tkachenko,
O.P., Boikob, B.C., Dnepropetrovsky Metallurgical Plant, "Machine for
Servicing Aluminium Electrolysers", Patent S U 1236002, F R 138361, 7.6.86.
11. Surpas, D., D y k m a n , J.P. and Lelewicz, V., Aluminium Pechiney, "Removing
Gases from Molten Aluminium - during production by adding alumina to melt
and aspirating off gases as they leave the surface", Patent FP 7911195, 24.4.79
12. Dalen, K.M., Kualavag, A. and Nagell, B., Norsk Hydro AS., "Point Feeder
for Aluminium Electrolysis Cell"., Patent U S 5,045,168, 3.7.90

REFERENCES PAGE 396
13. Sem, M.O., Electrokemisk A/S, "Method of Charging Aluminium Furnaces",

Patent U S 3,006,825, 31.10.61
14 Koyigin, V.K., Szimit, U.W., Mintis, M.T., Zawozin, U.D., Gladkih, G.L,
Irkutsk Aluminium Plant, "Electrolyser for Achieving Aluminium", Patent S U
1178799
15. Anon, Champagnie Pechiney and Electrification, "Alumina Supply for
Preparation of Aluminium by Electrolysis", Patent F R 418,408, A U 12,840/70,
14.04.70
16. Stefanidis, D . and Georgantonis, D., Light Metals, A I M E , 1973, pp 271-276
17. Reverdy, M., Aluminium Pechiney, "Alumina Feeding Systems (Mechanical)".
Australian Aluminium Smelter Technology Course, Sydney, 1987, pp 17.1 -
17.53.
18. Capitaine, W., Swiss Aluminium Ltd., "The Alusuisse Retrofit Concept -
Retrofit and Modernization of Existing Smelters", Light Metals, A I M E , 1982,
pp 499-511.
19. Langon, B. and Varin, P., Aluminium Pechiney, "Aluminium Pechiney 280kA
Pots", Light Metal Age, October 1986, pp 10-14

20. Keinborg, M . and Cuny, J.P., Aluminium Pechiney, "180 k A Prebake Pot
from prototype to potline". Light Metals, A I M E , 1982, pp449-460.
21. Kiley, A.J. and Shiver, H.T., Reynolds Metal Co., "Electrolytic Reduction Cell
with Crust-breaking and Ore Feeding Means", Patent U S 3,371,026, 27.02.68
22. Gerphagnon, J X . and Wolter, C., Aluminium Pechiney, "Assembly for Spot
Feeding Alumina to an Electrolytic Tank for the Production of Aluminium",
Patent U S 4,437,964, A U B-14958/83, F R 8209700, 20.03.84

23. Bonney, P. and Gerphagnon, T., Aluminium Pechiney, "Process and
Apparatus for Accurately Controlling the Rate of Introduction and the Content
of Alumina in an Igneous Electrolysis Tank in the Production of Aluminium.",
Patent U S 4,431,491, A U B-73198/81, E P 0,044,794, 14.02.84

24. Anon, "Aluminium Electrolyser Alumina Feed Device", Patent - S U 985-154A
(SU 317369 13.07.81), Metallurgy, week 8344, Alum. Magn. Electr. Ind.,
30.12.82
REFERENCES PAGE 397
25. Anon, "Aluminium Producing Electrolyser", Patent S U 1,560,636A (SU

328970 17.11.88), Metallurgy, week 9050, Irkut Alum. Magn. Ele., 30.04.90
26. Reverdy, M., Aluminium Pechiney, "High Amperage Cells". 8th Interational
Course on Process Metallurgy of Aluminium, 1989
27. Racunas, B.J. and Seger, E.J., (Internal Alcoa Report), "Effects of compacted
E S P - butt bath pot cover, compacted E S P feed and P-100 anode geometry on
pilot Hall cell fluoride evolution", Division report #06-92-107, 30.12.92
28. Bagshaw, A.N., Kuschel, G.I., Taylor, M.P., Tricklebank, S.B. and Welch,
B.J., "Effect of Operating Conditions on the Dissolution of Primary and
Secondary (Reacted) Alumina Powders in Electrolytes", Light Metals, A I M E ,
1985, pp 649-659
29. Bagshaw, A.N. and Welch, B.J., "The Influence of Alumina Properties on its
Dissolution in Smelting Electrolyte", Light Metals, A I M E , 1986, pp 35-39
30. Maeda, H., Matui, S. and Eta, A., "Measurement of Dissolution Rate of
Alumina in Cryolite Melt"., Light Metals, A I M E , 1985, pp 763-780
31. Roach, G.D., "Final Report of the Alcoa Sponsored Project by G. Kuschel and
Prof. B Welch, University of Auckland", Internal Alcoa Report, February 1990.
32. Lowe, R.L., Alcoa, "Alumina Feeder", Patent U S 3,681,229, A U 31274/71, F R
71.26172,01.08.72
33. Friedli, H., Arnold, E., Swiss Aluminium Ltd., "Storage Bunker Device for
Feeding Electrolytic Cells", Patent U S 4,332,660, E P 0025414, 01.06.82
34. Friedli, H., Gut, E., Aeschbach, P., Swiss Aluminium Ltd., "Device for
Servicing Electrolytic Cells", Patent U S 4,328,085, F R 8019460, 04.05.82
35. Jaggi, E., Swiss Aluminium Ltd, "Device for Controlled Batch Feeding of a
Fluidizible Particulate Material and Process for its Operation", Patent U S
4,525,105, E P 0069057, 25.06.85

36. Casdas, S., Aluminium de Grece sa Industrielle et Commerciale, "Process and
Apparatus for Controlled Supply of Alumina", Patent U S 2,483,964, A U B-
71338/81, F R 8013028, F R 539892, 11.12.81 & 18.10.84

REFERENCES PAGE 398
37. Niizeki, K., Watanabe, T., Takeuchi, A., and Kubota, T., Nippon Light
Metal Co. Ltd., "Apparatus For Feeding and Distributing Particulate Material",
Patent U K 1,452,540, 464735, A U 66481/74, F R 740847, 18.01.77
38. Hanrot, J.P., and Volpeliere, J., Aluminium Pechiney, "Distributing

Fluidizable Powder", Patent A U B-51884/86, F R 579437, E P 0190082,
17.07.86
39. Nagell, B., Norsk Hydro A.S., "Pneumatic Dosimeter", Patent Au-A-23488/88
N O 874220, 9.10.87.
40. Bjarnason, G., Eggertsson, T., Swiss Aluminium Ltd, "Device for Batch
Feeding of a Fluidizable Particulate Material", Patent U S 4,498,818, EP
0062605, 12.02.85
41. Boudreault, G., Alcan International Ltd., "Method for Feeding Particulate
Material", Patent E P 0206555, U S 4919303, 26.01.83
44. Arnold, P.C., University of Wollongong, "Flow Properties and Hopper
Geometry Parameters for Reacted Alumina Samples", February, 1991
45. Aalbu, J., Moen, T., Aalbu, M., Borg, P., Ardal O g Sunndal Verk A S ,
"Method of Controlling the Alumina Feed into Reduction Cells for Producing
Aluminium", Patent W O 86/05008, 28.8.86

46. Moen, T., Aalbu, J. and Borg, P., "Adaptive Control of Alumina Reduction
Cells with Point Feeders", Light Metals, A I M E , pp 459-469.

47. Tabereaux, A.T., Watts, W.E., Wilson, C.A., Reynolds Metal Co., "Method
for Improved Alumina Control in Aluminium Electrolytic Cells Employing
Point Feeders", Patent U S 4654130, 31.03.87.

48. Merz, W., Alusuisse, "Automatic Charging of Electrolysis Furnaces for
Manufacturing Aluminium via Pneumatic Convey System Controlled by
Computer; Reduces Energy Requirements", Patent E P 0026735, 28.8.79
49. Hanrot, J.P. and Volpeliere, J., Aluminium Pechiney, "Closed Apparatus
Providing Potential Fluidization for Horizontally Conveying Powdered
Materials", Patent 4,659,263 (close to U S 4,747,732), F R 8217859 (1982),
21.04.87
REFERENCES PAGE 399
50. Hanrot, J.P. and Volpeliere, J., Aluminium Pechiney, "Closed Apparatus with
Potential Fluidization for Horizontally Conveying Powdered Material in a
Dense-Bed", Patent U S 4,747,732, W O 85/04676, E P 122925 (1986), F R
8217859 (1982), 31.05.88
51. Hanrot, J-P., Volpeliere, J., and Pitour, A., Aluminium Pechiney,
"Fluidizated Bed for Continuous Separation of T w o Mixed Solid Phases", Patent
U S 4,741,443, E P 0187730, 03.05.88
52. Heinzmann, U. and Braun, W., Swiss Aluminium, "Process and Device for
Controlling a Crust Breaking Facility", Patent A U B-24373/84, 05.11.87
53. Anon, Aluminium Pechiney, "Process and Appartus for Controlling Solid
Electrolyte Additions to Electrolytic Cells for Aluminium Production", Patent
A U 14784/88, 27.10.88
54. Westerman, E.J., and Harrison, C.B., "Plunger Tip Alloys for Aluminium
Reduction Cells", Journal of Metals, pp 26-32, N o v 1984

55. Jain, R.K., Tricklebank, S.B., Welch B.J. and Williams D.J., "Interaction of
Aluminas with Aluminium Smelting Electrolytes", Light Metals, A I M E , 1983,
pp299-305
56. Kuschel, G.I. and Welch, B. J., "Further Studies of Alumina Dissolution under
Conditions Similar to Cell Operation", Light Metals, A I M E , 1990, pp299-305.
57. Kuschel, G.I. and Welch, B.J., "Effect of Alumina Properties and Operation of
Smelting Cells on the Dissolution Behaviour of Alumina", Proc. 2nd Int.
Alumina Quality Workshop, 1990

58. Barrillon, E., "The Merits and Demerits of Various Types of Bayer Aluminas",
Erzmetall, Bd. 31, 1978 H 11, pp 519-522.

59. Welch, B.J., "Aluminium Quality and Cell Feeding", 4th Australian Aluminium
Smelter Technology Workshop Sydney, 25-30th Oct. 1992, pp 272-286

60. Taylor, M.P., Lin, X., Fraser, K. and Welch, B.J., "The Dynamics and
Performances of Reduction Cell Electrolytes", Light Metals, A I M E , 1990, pp
259-265
61 Thomstad, J., Johnsen, P. and Kristensen, E.W., "Some Properties of
Alumina Sludge", Light Metals, A I M E , 1980, pp 227-239.

REFERENCES PAGE 400
62. Liu, X., Purdie, J.M., Taylor, M . P . and Welch, B.J., "Measurement and
Modelling of Alumina Mixing and Dissolution for Varying Electrolyte Test and
Mass Transfer Conditions", Light Metals, A D M E , 1990, pp 289-298
63. Haverkamp, R G . , Metcon, J.B. and Welch, B.J., Proceedings of the Molten
Salt Symposium of the Electrochem Society, M a y 1992
64. Richards, R L . , Banakes S.T. and Griswold, V.D., Alcoa (Warrick) internal
report, "Alumina Feeding Investigation", 1984
65. Bertand, Y. and Lectard, A., Aluminium Pechiney, "Aluminium Pechiney
Specifications for Optimising the Aluminas Used in Sidebreak and Point
Feeding Reduction Pots", Light Metals, A I M E , 1984, pp 667-686
66. Reverdy, M., "Modern High Amperage Cell Technology", 4th Australian
Aluminium Smelter Technology Workshop, Sydney, 25-30.10.1992
67. Chervonin, V.M., Stepanov, V.T., and Afrakova, T.F., Irkutsk Branch of the
Federal Scientific Research and Project Institute of Aluminium, Magnesium and
Electrode Industry. U S S R State Committee for Inventions and Discoveries,
"Electrolyser to obtain aluminium", Patent S U 1,258,883

68 Mirkin, M.A., Fridman, Y.M., Goldin, A.A., and Tkachenko, O.P., Dne
Propetrousic Factory for Metallurgical Equipment,Patent S U 1,258,884, S U
1191491,03.09.86
69. Dunstan, M . R , Boyne Smelters Ltd, "Development of a Point Feeder
Maintenance Stategy", Australian Smelter Maintenance Workshop, February
1992.
70. Pfiste, H., Swiss Aluminium, "Electropneumatic Drive System for
Crustbreaking Devices and Process for Operating the Same", Patent U S
4,606,257, E P 0124480, A U A-27128/84, 19.08.86

71. Shapiro, J.P., "Alcoa Gravimetric Feeder", Internal Alcoa Report, 14.12.81
72. Hughes, J.P., Proceedings of a Seminar, "Health Protection in Primary
Aluminium Production". International Primary Aluminium Institute
Copenhagen, 28-30.06.77
R™NCES PAGE402
89. Anon, Seal Manual:, "Competitive Selling of Viton for Industrial Seals"
Supplied by Terry Fluids Ltd, Undated
90. Andersson, K., Atlas Copco, Personal correspondence, 8.7.91

91. Stark, H.L., "Spools", Unisearch, University of N S W , August 1991
92. Stark, H.L., "Spools Fatigue", Unisearch, University of N S W , August 1991
93. Stueland, M., Hydro Aluminium (Karmoy Fabrikker), personal

correspondence, 21.10.93.
94. Pawlek, R , "Primary Aluminium Smelters of the World", Aluminium-Verlag,

Updated April, 1993.
95. Frankenfeldt, R E . , "Report from the Chairman of the IPAI Environmental

Committee to the IPAI Board", 2.11.93.
96. Juran, J.M., " Juran's Quality Control Handbook", 4th Edition, M c G r a w Hill,
1974, pp24.8-24.9 and AH.3.
97. Cover of Light Metals Age, (Hydro Aluminium, Karmoy, Norway), Vol 50 N o
11 & 12, Dec 1992.
98. Jones, P., and Roy, R K . , " A n Improved Actuating Cylinder", Comalco
Aluminium Limited and Terry Fluid Controls, Patent application
PCT/AU95/00268, 15.7.94.
99. Seymour, R E . , "Alumina Feeder Improvement", M e m o to Alcoa Smelting

Managers, 16.4.92.
100. Taylor, R R , "James P. Kissane - Portland", personal correspondance, 25.7.94.
101. Hoi nisi, P., and Leonard, B., (Pechiney and Tomago respectively) discussions
at Fifth Australian Aluminium Smelter Technology Workshop, Sydney,
27.10.95.
102. Knapp, L., (Boyne Smelters Limited) discussions at Fifth Australian
Aluminium Smelter Technology Workshop, Sydney, 27.10.95.

103. Purdie, J., (Comalco), "Effect of Feeder Design on Alumina Concentration
Control and Associated Process Variation", Fifth Australian Aluminium Smelter
Technology Workshop, Sydney, 27.10.95, pp628-650.

APPENDICES
Appendix 1 Examples of Shot Size Control Charts Page 1
APPENDIX 1
EXAMPLE OF SHOT SIZE CONTROL CHARTS
1520
^1500
^1480
N
5)1460
O1440
x
W1420 1400
1400 1 2 3 4 5 6 7 8 9 101112131415161718192021222324 25
SHOT NUMBER
1 3 5 7 9 11 13 15 17 19 21 23
0. 800
o 600
400
200
UJ
> 0
UJ
1 3 5 7 9 11 13 15 17 19 21 23 25
SHOT NUMBER
APPENDIX 2
OFF-SITE SHOT SIZE DATA

Appendix 2 Off-site Shot Size Test Data Page 1
EFFECT OF ALUMINA TYPE, FEEDER DESIGN, INSERT AND SPOOL SHAPE

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA ]
D W E L L TIME (s)
TYPE TIME (s) 0.6 1.9 1.6 2.0 2.S 3.0 3.5 4.0 46 5.0
1 10-Jan STANDARD PARKER FINE KEMPE 175mm + 1.0/1.1 A V E R A G E 878 890 1063 1195 1372 1483 1557
30mm SPACER STDOEV 4 8.5 9 10 a.5 a.s 8.5
2 10-Jan PA#1 53mm INSERT PARKER FINE KEMPE 175mm* 1.0/1.1 A V E R A G E 1398 1374 1384 1442 1445 1435 1423
(CAST IRON) 30mm SPACER S T D DEV 1.5 1.5 2.5 1 1 2 1
3 10-J an PA#1 53mm INSERT PARKER FINE KEMPE 175mm* 1.0/1.1 A V E R A G E 1414 1448 1464 1470 1447 1424 1455
(ENAMEL) 30mm SPACER S T D DEV 2 1.5 2.5 1 1.5 1 5
4 10-Jan A2 STANDARD PARKER FINE KEMPE 175mm + 1.0/1.1 AVERAGE 690 856 852 853 852 854 85B
30mm SPACER S T D DEV 8 1.5 1.5 1 1.5 1.5 1
5 10-Jan TS#1 STANDARD TERRY FINE STAINLESS ST. 0.6/0.9 AVERAGE 1498 1594 1638
240mm STDOEV 1.5 1 1.5
e 10-Jan PA#1 BOLT HOLES PARKER FINE K E M P E 175mm + 1.0/1.1 AVERAGE 742 955 1150 1394 1451 1510 1613
BLOCKED 30mm S P A C E R S T D DEV 3.5 3.5 6.5 6.5 4.5 2.5 3.5
7 10-Jan PA#1 STANDARD PARKER NORMAL KEMPE 175mm* 1.0/1.1 AVERAGE 823 962 1112 1298 1408 1442 1527
30mm SPACER STD DEV 3 3 4 7.5 4.5 3.5 3
8 10-Jan PAH 53mm INSERT PARKER NORMAL KEMPE 175mm* 1.0/1.1 AVERAGE 1431 1471 1486 1489 1484 1471 1490
(CAST IRON) 30mm SPACER STD DEV 3 1 1 1 1 0.5 0.5
9 10-Jan PA#1 53mm INSERT PARKER NORMAL KEMPE 175 m m + 1.0/1.1 AVERAGE 1380 1491 1444 1462 1474 1510 1472
(ENAMEL) 30mm SPACER STD DEV 3.5 1.5 1 1 1 1 15
to 10-Jan A2 STANDARD PARKER NORMAL K E M P E 175mm + 1.0/1.1 AVERAGE 509 855 864 856 857 862 854
30mm S P A C E R S T D DEV 16 15 0.5 1 1 1 75
11 10-Jan T6#1 STANDARD TERRY NORMAL STAINLESS ST. 0.6/0.9 AVERAGE 1403 1554 1808 1639 1664 1674 1685
240mm S T D DEV 1.5 1 1 1 0.5 0.5 0.5
12 10-Jan PA#1 BOLT HOLES PARKER NORMAL K E M P E 175mm + 1.0/1.1 AVERAGE 735 952 1142 1319 1502 1560 1708
BLOCKED 30mm SPACER S T D OEV 2 4.5 4 4 3.5 3 3.5
13 10-Jan PA#1 STANDARD PARKER UNREACTEO K E M P E 175mm + 1.0/1.1 AVERAGE 716 980 1182 1288 1419 1554 1687
30mm SPACER STD DEV 2.5 4 6 4 5.5 2.5 3
14 10-Jan PA#1 53mm INSERT PARKER UNREACTED! K E M P E 175mm * 1.0/1.1 AVERAGE 967 1428 1431 1431 1440 1450 1436
(CAST IRON) 30mm S P A C E R S T D DEV 18 1.5 1.5 0.5 2 1 1
15 10-Jan PA#1 53mm INSERT PARKER U N R E A C T E D K E M P E 175mm + 1.0/1.1 AVERAGE 1388 1391 1405 1416 1433 1428 1392
(ENAMEL) 30mm S P A C E R S T D DEV 2 1 0.5 1 0.5 0.5 1
IS 10-Jan A2 STANDARD PARKER U N R E A C T E D K E M P E 175mm + 1.0/1.1 AVERAGE 528 801 806 837 834 824 826
30mm SPACER S T D DEV 14 2.5 2 2 1 3 1
17 10-Jan T6#1 STANDARD TERRY U N R E A C T E D STAINLESS ST. 0.6/0.9 AVERAGE 1324 1471 1539 1585 1606 1626 1633
240mm S T D OEV 3 1 1 0.5 1 0.5 0.5
18 10-Jan PA#1 BOLT HOLES PARKER UNREACTED; K E M P E 1 7 5 m m * 1.0/1.1 AVERAGE 716 1149 1128 1435 1405 1584 1671
BLOCKED 30mm S P A C E R S T D DEV 2.5 4.5 6 5.5 5.5 2 1.5
EFFECT OF SECOND FEEDER (SIZE.STROKE TIME.CYLINDER)

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME (s)
TYPE TIME(s) o.s 1.0 1.6 2.0 2.6 3.0 3.S 4.0 4.6 6.0
19 29-Jan PA #2 STANDARD PARKER FINE KEMPE 175mm* 1.0/1.1 AVERAGE 782 I 1001 1191 1381 1404 1555 1584 1670 1712 1715
30mm SPACER S T D DEV 2.5 5 4 4.5 5 3 3 2 1 1.5
20 02-Feb T6#2 STANDARD TERRY FINE STAINLESS ST. 1.6/1.9 A V E R A G E 1031 1426 1504 1584 1607 1633 1857 1686 1692 1703
240mm STDOEV 3 2 2 1.5 1 1 0.5 1 0.5 1.5
21 29-Jan PA #2 STAN0ARD PARKER FINE KEMPE 175mm* 1.6/1.9 AVERAGE 677 979 1205 1405 1530 1569 1604 1660 1676 1720
30mm SPACER STDOEV 5 3 5.5 4 5 2.5 2 1.5 1 2
22 29-Jan PA #2 STANDARD PARKER NORMAL KEMPE 175mm* 1.0/1.1 AVERAGE 795 1052 1323 1491 1566 1633 1679 1753 1791 1793
30mm SPACER S T D DEV 2 3 3.5 3.5 2 2 2 1 1 1
23 31-Jan T8#2 STANDARD TERRY NORMAL STAINLESS ST. 1.6/1.9 AVERAGE 1144 1564 1631 1673 1692 1699 1694 1726 1729 1727
240mm STD D E V 3 1 1 0.5 1 0.5 0.5 1 0.5 0,5
24 10-Feb PA #2 53mm INSERT PARKER FINE KEMPE 175mm* 1.6/1.9 AVERAGE 560 940 1235 1280 1300 1340 1320
(CAST IRON) 30mm SPACER STD DEV
25 10-Feb PA #2 53mm INSERT PARKER NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 608 1033 1326 1348 1358 1361 1361
(CAST (RON) 30mm SPACER S T D DEV 4.5 2 1.5 0.5 0.5 0 0
26 11-Feb PA #2 53mm INSERT PARKER NORMAL KEMPE 175mm* 1.3/1.5 AVERAGE 584 1074 1298 1344 1362 1364 1374
(CAST IRON) 30mm SPACER STDOEV 2.5 2 1 3 1.5 2 4
27 12-Feb PA #2 53mm INSERT PARKER FINE KEMPE 175mm + 1.3/1.5 AVERAGE 477 957 1245 1287 1305 1353 1384
(CAST IRON) 30mm SPACER S T D DEV 2 2 1 1 0.5 0.5 0.5
EFFECT OF SIZE ON 81mm STROKE

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME |s)
TYPE TIME (s) 0.5 1.0 1.6 2.0 2.5 3.0 3.6 4.0 4.6 6.0
28 14-Feb PA #2 81mm PARKER NORMAL KEMPE 175mm* 1.3/1.5 AVERAGE 1324 1944 2100 2119 2120 2119 2127
30mm SPACER S T D DEV 11 36 0.5 1 0.5 0.5 1
(NO INSERT)
29 13-Feb PA«2 81mm PARKER FINE KEMPE 175mm + 1.3/1.5 AVERAGE 1302 1245 2048 2080 2097 2083 2080
30mm SPACER S T D DEV 4 2 2 0.5 0.5 0.5 0.5
(NO INSERT)
EFFECT OF STROKE TIME AND DIFFERENT CYLINDERS

RUN ALUMNA SPRING STROKE DATA D W E L L TIME (a)
DATE FEEDER DESIGN CYLINDER
TIME (3) 0.6 1.0 1.8 2.0 2.5 3.0 3.5 4.0 4.6 6.0
TYPE
- 1069 1438 1618 1745 1843 1724
' 3 d 28-Feb PA #2 STANDARD PARKER NORMAL KEMPE 175mm * 1.7/1.9 AVERAGE 823
STDOEV 3.5 2 3 1.5 1 0.5 0.5
30mm SPACER
KEMPE 175mm* AVERAGE 821 1174 1526 1736 1687 1833 1848
31 25-Feb PA #2 STANDARD PARKER NORMAL 1.2/1.7
2.5 2 1.5 1.5 0.5 1 0.5
30mm SPACER S T D DEV
AVERAGE 922 1602 1817 1851 1867 1B73 1869
32 28-Feb PA #2 STANDARD TERRY NORMAL KEMPE 175mm + 1.7/1.9
2.5 1 0.5 0.5 0.5 0.5 0.5
AVERAGE 992 1659 1818 1834 1852 1850 1850
33 28-Feb PA #2 STANDARD TERRY NORMAL KEMPE 175mm* 1.2/1.7
S T D DEV 4.5 0.5 0.5 0.5 0.5 0.6 1
30mm SPACER
AVERAGE 705 1084 1481 1703 1756 1669 1852
34 24-Feb PA #2 STANDARD TERRY NORMAL KEMPE 175mm* 1.5/1.7
2 2.5 2 1 0.5 0.5 1
Appendix 2 Off-site Shot Size Test Data
EFFECT OF SPOOL STROKE (NO INSERT) WITH DIFFERENT STROKE TIME

TYPE TIME (S)
35 20-Feb PA #2 HIGH S P O O L TERRY NORMAL KEMPE 175mm* 1.3/1.5 A V E R A G E 1802 2085 2075 2079 2069 2079 2071
80mm
30mm SPACER STDOEV 1 0 5 0.5 0.5 0.5 05
36 15-Feb PA #2 HISBWKM. TERRY NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 1186 1809 2180 2201 2202 2196 2189
80mm 30mm S P A C E R 8.5
S T D DEV 1 1 0.5 1
37 19-Feb PA #2 HIGH S P O O L TERRY NORMAL KEMPE 175mm* 1.3/1.5 AVERAGE 1667 1974 2044 2071 2066 2069 2073
80mm 30mm SPACER STDOEV 6 0,5 0.5 0.5 0.5 0.5 0.5
38 17-Feb PA #2 HIGH S P O O L TERRY NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 1463 2011 2068 2052 2110 2058 2062
80mm 30mm SPACER STDOEV 5.5 6.5 1 1 1.5 0.5 0.5
39 15-Feb PA #2 HIGH S P O O L PARKER NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 1224 1677 2061 1174 2166 2172 2170
80mm 30mm S P A C E R STDOEV
EFFECT OF DIFFERENT FEEDERS FROM OFF LINE POTS

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME (a)
TYPE TIME (a) 0.6 1.0 1.6 2.0 2.6 3.0 3.6 4.0 4.6 6.0
40 04-Mar PA #2 *1029 PARKER NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 856 1253 1668 1793 1819 1822 1866
30mm SPACER STDOEV 3.2 1.5 1.2 1.4 0.6 0.4 2.2
41 04-Mar PA #2 #1975 PARKER NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 748 1093 1519 1678 1728 1765 1803
30mm SPACER S T D DEV 2.2 2.6 1.6 0.95 0.6 0.6 0.5
42 06-Mar PA #2 #1282 PARKER NORMAL KEMPE 175mm + 1.6/1.9 A V E R A G E 789 1136 1531 1694 1780 1807 1801
30mm S P A C E R S T D DEV 2.4 2.3 2.2 0.8 1.2 0.5 1.5
43 OS-Mar PA #2 *590 TERRY NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 739 1083 1434 1556 1659 1652 1683
30mm SPACER S T D DEV 3.3 2 1.4 0.7 2.5 0.4 0.5
44 16-Mar PA #2 #474 TERRY NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 786 1235 1614 1690 1722 1743 1772
30mm SPACER STDOEV 2.8 1.7 0.8 0.4 0.4 0.6 2.7
45 10-Mar PA #2 #650 TERRY NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 940 1667 1622 1847 1848 1829 1844
30mm S P A C E R S T D DEV 3.2 1.1 0.4 1.4 0.4 0.5 0.4
EFFECT OF DIFFERENT CYLINDERS WITH INSERT

TYPE TIME (s) 0.5 1.0 1.5 2.0 2.6 3.0 3.6 4.0 4.6 6.0
46 19-Mar PA #2 5 3 m m INSERT PARKER NORMAL KEMPE 175mm* 1.7/1.9 A V E R A G E 724 1217 1402 1437 1451 1430 1427
(STEEL) 30mm SPACER STD DEV 3 2.8 0.6 0.5 0.3 1 1.1
47 20-Mar PA #2 5 3 m m INSERT PARKER FINE KEMPE 175mm* 1.7/1.9 A V E R A G E 713 1131 1350 1399 1420 1445 1422
(STEEL) 30mm SPACER S T D DEV 6.4 2.2 1.7 0.9 0.4 0.9 1.8
48 21-Mar PA #2 5 3 m m INSERT TERRY NORMAL KEMPE 175mm* 1.7/1.9 A V E R A G E 400 987 1288 1327 1346 1349 1342
(STEEL) 30mm S P A C E R S T D DEV 20.5 2.7 1.6 0.6 0.4 0.8 0.8
49 21-Mar PA #2 5 3 m m INSERT TERRY FINE KEMPE 175mm* 1.7/1.9 A V E R A G E 609 1054 1280 1334 1352 1354 1364
(STEEL) 30mm SPACER STDOEV 2.2 3.9 1.6 0.5 0.3 0.3 0.2
EFFECT OF STROKE TIME.SIZE AND INSERT ON T6 FEEDERS

TYPE TIME (a) 0.6 I 1.0 1.6 2.0j 2.6 3.0 3.5 4.0 4.6 6.0
50 17-Mar T6#2 STANDARD TERRY NORMAL STAINLESS ST. 0.6/0.9 A V E R A G E 1365 1646 1667 1680 1684 1682 1682
240mm STDOEV 2.4 0.6 0.4 0.4 0.3 0.2 0.6
51 18-Mar T8#2 STANDARD TERRY NORMAL STAINLESS ST. 1.1/1.3 A V E R A G E 1251 1651 1683 1693 1692 1691 1689
240mm S T D OEV 2.3 0.4 0.2 0.4 0.2 0.2 0.2
52 17-Mar T6#2 STANDARD TERRY FINE STAINLESS ST. 0.6/0.9 A V E R A G E 1396 1609 1717 1677 1774 1783 1786
240mm S T D DEV 6.4 1.1 3.9 1 1.4 1.2 1.2
53 18-Mar T6#2 STANDARD TERRY FINE STAINLESS ST. 1.1/1.3 A V E R A G E 1056 1628 1663 1679 1697 1680 1660
240mm S T D DEV 2.9 0.6 1.2 0.7 1.2 0.7 0.7
54 22-Mar T6#2 5 3 m m INSERT TERRY NORMAL STAINLESS ST. 0.6/0.9 A V E R A G E 966 1284 1287 1291 1288 1284 1284
240mm STDOEV 9.8 0.4 0.4 0.4 0.4 0.3 0.4
55 23-Mar T6#2 5 3 m m INSERT TERRY FINE STAINLESS ST. 0.6/0.9 A V E R A G E 1024 1292 1333 1281 1287 1291 1296
240mm STDOEV 22 0.4 0.4 0.4 0.5 0.6 0.7
EFFECT OF DIFFERENT FEEDERS. SIZE. AND STROKE TIME ON T11 FEEDERS

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME HI
TYPE TIME |s) 0.5 1.0 1.S 2.0 2.5 J.O 3.6 4.0 4.6 6.0
56 24-Mar T11 #1 STANDARD LINDBERG NORMAL STAINLESS ST. 0.8/0.8 AVERAGE 1103 1565 1630 1653 1659 1673 1673
240mm S T D DEV 5.6 0.6 0.4 0.4 0.4 0.4 0.4
57 23-Mar T11#1 STANDARD LINDBERG FINE STAINLESS ST. 0.8/0.8 AVERAGE 1078 1494 1554 1577 1594 1604 1612
240mm STDOEV 9 0.5 0.3 0.4 0.4 0.3 0.3
58 24-Mar Til #2 STANDARD LINDBERG NORMAL STAINLESS ST. 0.8/0.8 AVERAGE 966 1100 1247 1366 1459 1519
240mm S T D DEV 2.4 1.8 2.4 1.6 1 0.6
59 01-Apr T11#2 STANDARD LINDBERG FINE STAINLESS ST. 0 8/08 AVERAGE 905 1406 1493 1563 1575 1600 1619
240mm S T D DEV 6.9 1.2 0.4 0.4 0.6 0.8 1.2
60 T11 #2 STANDARD STAINLESS ST. 1.3/1.3 AVERAGE 684 990 1215 1340 1429 1498 1550
31-Mar LINDBERG NORMAL
240mm STDOEV 3.4 2 2 1.5 1.2 1.6 0.8
81 STAINLESS ST. 1.3/1.3 AVERAGE 696 963 1151 1318 1426 1494 1530
31-Mar T11#2 STANDARD LtNOBERG FINE
S T D DEV 3.2 2.5 2.5 1.2 1.2 1.6 1
240mm mmm
EFFECT OF SIZE ON T11 WITH INSERT

RUN ALUMNA SPRING STROKE DATA D W E L L TIME ll)
DATE FEEDER DESIGN CYLINDER
TYPE TIME (s) 0.5 1.0 1.6 2.0 2.5 3.0 3.5 4.0 4.5 5.0
STAINLESS ST. 0.8/0.8 A V E R A G E 695 1178 12S1 1304 1324 1324 1326
62B 02-Apr T11 #2 5 3 m m INSERT LINDBERG NORMAL
STD DEV 2.7 2.8 0.4 0.2 0.4 0.4 0.3
240mm
0.8/0.3 AVERAGE 607 1141 1266 1303 1311 1319 1327
82A 02-Apr Til #2 5 3 m m INSERT LINDBERG FINE STAINLESS ST.
S T D DEV 2.3 2.2 0.4 0.4 0.4 0.2 0.4
240mm
EFFECT OF STROKE TIME AND CYLINDER ON A2 FEEDFR

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME Is)
TYPE TIMEIS)
S3 25-Mar A2 STANDARD PARKER NORMAL K E M P E 175mm + 1.2/1.4 A V E R A G E 497 8S6 839 849 860 843 857
30mm SPACER STDOEV 14 0.5 0.3 0.9
64 23-Mar A2 STANDARD PARKER NORMAL KEMPE 175mm* 1.7/1.9 A V E R A G E 437 893 857 858 897 857 846
30mm SPACER STDOEV 1 0.5 0.3 0.5
65 01-Apr A2 STANDARD TERRY NORMAL K E M P E 175mm + 1.7/1.9 A V E R A G E 407 915 912 909 915 917 918
30mm SPACER STDOEV 0.9 0.5 0.4 0.6
68 04-Apr A2 STANDARD TERRY NORMAL KEMPE 175mm* 1.2/1.4 A V E R A G E 652 910 910 914 911 912 908
30mm SPACER S T D DEV 1.8 0.3 0.4
67 04-Apr A2 STANDARD TERRY FINE KEMPE 175mm* 1.7/1.9 A V E R A G E 506 936 933 938 942 946 940
30mm SPACER STD DEV 1.1 0.4 0.4 0.6 0.5 0.8
EFFECT OF SPOOL SHAPE

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA D W E L L TIME (af
TYPE TIME (SI 0.6 j 1.0 1.6 2.0 2.6 3.0 16 4.0 4.6 6.0
68 24-Apr PA #2 SHORT SPOO PARKER NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 356 309 1220 1609 Hi ,9
30mm SPACER S T D DEV 1,3 1 0.6 0.5
69 24-Apr PA #2 81mm INSERT PARKER NORMAL K E M P E 175<nm* 1.6/1.9 AVERAGE 851 1486 1512 1520 1516 1519 1534
30mm S P A C E R STDOEV 2.2 0.2 0.2 0.3 0.5 0.2
70 27-Apr PA #2 HOLES PARKER FINE KEMPE 175mm* 1.6/1.9 AVERAGE 846 1192 1601 1709
1642 1695 1719
SHOWING 30mm SPACER STD DEV 4 2.3 1 2.1 0.2 0.6
71 26-Apt PA #2 HOLES PARKER NORMAL K E M P E 175mm + 1.6/1.9 AVERAGE 982 1614 1712 1721 1755 1753
SHOWING 30mm SPACER STDOEV 3.2 0.4 0.2 0.2 0.4 0.2
EFFECT OF SIZE AND INSERT ON T6 FEEDERS

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA DWELL TIME (a)
TYPE TIME(S) 0.6 1.0 1.6 2.0 2.6 3.0 1.6 4.0 4.6 6.0
72 28-Apr T6#2 STANDARD TERRY NORMAL STAINLESS ST. 0.6/0.9 A V E R A G E 1399 1656 1891 1706 1713 1713 1721
240mm STD DEV 4.2 0.4 1.4 0.4 0.2 0.4 0.2
73 29-Apr T6#2 STANDARD TERRY FINE STAINLESS ST. 0.6/0.9 A V E R A G E 1454 1663 1702 1711 1719 1731 1728
240mm STDOEV 2.1 0.8 0.4 0.4 0.4 0.4 0.3
74 30-Apr T6#2 8 1 m m INSERT TERRY NORMAL STAINLESS ST. 0.6/0.9 A V E R A G E 1292 1365 1373 1381 1386 1401 1406
240mm S T D DEV 0.3 0.4 0.2 0.2 0.3 0.2 0.2
75 29-Apr T6#2 81mm INSERT TERRY FINE STAINLESS ST. 0.6/0.9 A V E R A G E 1285 1339 1363 1366 1375 1380 1389
240mm STD DEV I 0.4 0.3 0.4 0.4 0.2 03 0.2
EFFECT OF SIZE AND INSERT ON T11 FEEDERS

TYPE TIME (a) 0.6 1.0 1.6 ZO 2.6 3.0 J-6 4.0 4.6 6.0
76 01-May T11 #2 STANDARD LINDBERG NORMAL STAINLESS ST. o a/o 8 A V E R A G E 840 1249 1480 1596 1643 1676 1689
240mm S T D DEV 3.1 1.4 0.6 0.4 0.4 0.4 0.6
77 08-May T11 #2 STANDARD LINDBERG FINE STAINLESS ST. 0.8/0.8 A V E R A G E 842 1308 1631 1614 1665 1690 1701
240mm STDOEV 2.8 1.2 0.4 0.4 0.2 0.2 0.3
78 07-May T11 #2 81mm INSERT LINDBERG NORMAL STAINLESS ST. 0 8/0 B A V E R A G E 833 1326 1345 1357 1362 1367 1369
240mm S T D DEV 3 0.4 0.4 0.4 0.2 0.3 0.3
79 01-May T11 #2 81 m m INSERT LINDBERG FINE STAINLESS ST. 0.8/0.8 A V E R A G E 771 1321 1355 1355 135S 1356 1364
240mm STD DEV | 7.4 0.3 0.3 0.2 0.2 0.2 0.5
EFFECT OF BATH SYSTEM DUST ON STANDARD AND INSERT FEEDERS

RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING STROKE DATA DWELL TIM: <s>
TYPE TIME (a) 0.6 1.0 1.6 2.0 2.6 3.0 3.6 4.0 4.6 6.0
SO 27-May PA #2 STANDARD PARKER NORMAL K E M P E 175mm + 1.6/1.9 A V E R A G E 483 923 1110 1224 1373 1431 1490
30mm SPACER S T D DEV 8.1 4.8 3 7.6 8.7 8.4 8
81 18-May PA #2 STANDARD PARKER 10% DUST KEMPE 175mm* 1.8/1.9 A V E R A G E 667 847 1042 1228 1393 1481 1566
30mm SPACER STDOEV 4.2 3.2 4 5 3.4 2.6 1.5
82 19-May PA #2 81mm INSERT PARKER 1014 D U S T K E M P E 175mm + 1.6/1.9 A V E R A G E 528 892 1096 1269 1312 1348 1348
30mm SPACER STDOEV 4 3.1 3.6 2.7 2 0.8 1
83 25-May PA #2 STANDARD PARKER 3 0 % DUST K E M P E 175mm * 1.6/1.9 A V E R A G E 548 959 1119 1319 1369 1580 1513
30mm SPACER STDOEV 8 7 10 12 18 6 13
84 25-May PA #2 61mm INSERT PARKER 30% OUST K E M P E 175mm + 1.6/1.9 A V E R A G E 300 562 744 912 1023 1181 1230
30mm SPACER STDOEV 11 25.5 18.5 11,5 9.5 12.5 11
86 28-May PA #2 STANDARD PARKER NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 483 923 1111 1224 1380 1431 1491
30mm SPACER STDOEV 9.3 4,7 2.9 7.6 9.2 8.4 4
EFFECT OF STROKE DISTANCE WITH INSERT (PARKER Vs TERRY. FINE Vs NORMAL)

TYPE TIM: (a) 0.6 1.0 1.5 2.9 2-6 3.0 3.6 4.0 4.6 6.0
102 22-Jul PA #2 STANDARD PARKER FINE K E M P E 175mm * 1.6/1.9 A V E R A G E 866 1559 1695 1670 1684 1729 1753 1806 1733 1812
30mm S P A C E R STDOEV 4.9 1.9 1 0.6 1 0.5 0.9 0.9 1 0.8
103 24-Jii PA #2 STANDARD TERRY FINE KEMPE 175mm* 1.8/1.9 A V E R A G E 535 1489 1724 1654 1766 1826 1702 1886 1711 1688
30mm SPACER S T D DEV 3.2 1.2 0.9 0.4 0.8 1 0.4 1.4 0.8 0.4
104 24-Jul PA #2 STANDARD TERRY NORMAL K E M P E 175mm + 1.6/1.9 A V E R A G E 750 1579 1672 1691 1695 1702 1714 1715 1711 1714
30mm SPACER STDOEV 6.2 1.1 0.5 0.4 0.3 0.3 0.3 0.4 0.3 0.4
105 28-Jtf PA #2 5 3 m m INSERT KEMPE 175mm* 1.6/1.9 A V E R A G E 468 1066 1338 1344 1344 1346 1351 1352 1355 1354
TERRY FINE
30mm SPACER STDOEV 2.3 3.2 0.2 0.2 0.3 0.3 0.2 0.3 0.2 0.2
106 29-Jul PA #2 53mm INSERT TERRY NORMAL K E M P E 175mm + 1.6/1.9 A V E R A G E 532 1304 1391 1391 1392 1349 1339 1335 1346 1337
30mm S P A C E R S T D DEV 1.9 1.3 0.4 0.5 0.7 0.2 0.2 0.2 02 0.2
107 04-Aug PA #2 53mm INSERT PARKER FINE K E M P E 175mm • 1.6/1.9 A V E R A G E 649 1051 1346 1368 1379 1365 1377 1384 1381 1382
30mm S P A C E R S T D DEV 2.3 2.5 1.1 0.6 0.3 0.9 0.3 0.3 0.3 0.2
108 05-Aug 53mm INSERT PARKER NORMAL KEMPE 175mm* 1.6/1.9 A V E R A G E 701 1086 1368 1367 1373 1375 1379 1377 1377 1384
PA #2
30mm S P A C E R S T D DEV 2.9 1.5 0.8 0.3 0.3 0.3 0.2 0.3 0.2 0.3
109 07-Aug PA #2 STANDARD PARKER NORMAL K E M P E 175mm * 1.6/1.9 A V E R A G E 775 1056 1349 1516 1543 1599 1639 1668 1688 1703
30mm S P A C E R STDOEV 4.2 2.2 1.9 2.1 1.5 1.6 0.9 0.8 0.5 0.4
110 11-Aug STANDARD KEMPE 175mm* 1.6/1.9 A V E R A G E 813 1034 1233 1455 1504 1650 1598 1652 1784 1739
PA #2 PARKER FINE
30mm S P A C E R STODEV 3 1.7 4.5 34 2.3 1.6 2.6 1.4 0.7 2
111 12-Aug PA #2 8 1 m m INSERT NORMAL STAINLESS ST. 1.8/1 9 A V E R A G E 902 1439 1472 1469 1468 1456 1434 1463 1457 1418
PARKER
240mm STD DEV 8.9 0.4 0.2 0.3 0.2 0.3 0.4 0.4 0.3 0.4
112 13-Aug PA #2 81mm INSERT PARKER FINE STAINLESS ST 1.8/1.9 A V E R A G E 742 1461 1496 1503 1504 1505 1510 1507 1508 1507
240mm S T D DEV 5.9 0.8 0.4 0.2 0.3 0.4 0.4 0.3 0.5 0.5
EFFECT OF SPRING TYPE (KEMPE 175mm + 30mm SPACER: KEMPE 210mm: SS 240m: S
INSERT STROKE (81mm and 53mm); TERRY Vs PARKER
RUN DATE FEEDER DESIGN CYLINDER ALUMNA SPRING 3TROKE DATA D W E L L TIME (a)
TYPE TIME (a) 0.5 1.0 1.6
113 17-JUlg PA #2 81mm INSERT TERRY NORMAL STAINLESS ST. 1.6/1.9 AVERAGE 753 1401 1416 1416 1417 1418 1418 1418 1418 1417
240mm STDOEV 3.1 0.4 0.3 0.3 0.3 0.2 0.2 0.3
114 18-Aug PA #2 81mm INSERT TERRY FINE STAINLESS ST. 1.6/1 9 AVERAGE 687 1450 1450 1450 1449 1448 1451 1448
240mm STDOEV 5.4 0.4 0.3 0.3 03 0.3
115 17-Aug PA #2 81mm INSERT TERRY NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 878 1444 1436 1444 1442 1446 1444 1442 1448
30mm SPACER S T D DEV 6.3 0.4 0.3 0.3 0.3 0.4 0.3 03
116 24-Aug PA #2 81mm INSERT TERRY FINE KEMPE 175mm * 1.8/1.9 AVERAGE 794 1446 1448 1449 1451 1449 1464 1459 1458
30mm SPACER STD DEV 3.2 0.8 0.5 0.5 0.3 0.3 03
117 26-Aug PA #2 81mm INSERT PARKER NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 867 1462 1516 1536 1486 1472 1502 1542 1487
30mm SPACER STD DEV 2.1 03 0.7 0.6 0.3 0.2 0.6
118 28-Aug PA #2 81mm INSERT PARKER FINE KEMPE 175mm* 1.6/1.9 AVERAGE 722 1521 1582 1584 1584 1584 1564 1564 1580 1581
30mm SPACER S T D DEV 2.4 0.8 0.7 0.4 0.2 0.3
119 01-Sep PA (2 STANDARD PARKER NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 861 1208 1529 1673 1720 1736 1734 1744 1747
30mm SPACER STDOEV 3.3 2 2.9 0.8 0.6 0.4 0.7 0,3 0.3 02
120 03-Sep PA #2 STANDARD PARKER NORMAL KEMPE 210mm 1,6/1.9 AVERAGE 877 1221 1536 1679 1720 1734 1758 1776 1786 1783
STDOEV 1.8 1.3 1.8 1.0 1.1 0.4 05 0.6 0.2 0.4
121 03-Sep PA #2 53mm INSERT PARKER NORMAL KEMPE 210mm 1.671.9 AVERAGE 721 1219 1336 1358 1362 1370 1368 1365 1371 1362
S T D DEV 3.4 2.6 0.2 0.2 0.1 0.2 0.2 0.3 0.3 0.5
122 09-Sep PA #2 53mm INSERT PARKER NORMAL STAINLESS ST. 1.6/1.9 AVERAGE 753 1194 1333 1369 1377 1387 1387 1391 1386 1390
220mm STDOEV 2.3 2.1 0.6 0.4 0.2 0.3 0.3 0.3 0.3 0.2
EFFECT OF CUSHION QUALITY

TYPE TIME (a) 0.6 1.0 1.6 2.0 2.8 3.0 3.6 4.0 4.5 6.0
122A 10-Sep PA #2 STANDARD PARKER NORMAL STAINLESS ST. 1.6/1.9 AVERAGE 883 1198 1560 1636 1684 1700 1742 1742 1755 1746
220mm STD OEV 2.6 2.5 1.2 2.0 0.1 1.6 0.5 0.4 0.3 0.8
123 14-Sep NORMAL KEMPE 175mm +
PA #2 STANDARD
N O CUSHION
PARKER
30mm SPACER
1.671.9 AVERAGE 844 1448 1531 . 1583 1844 1675 1784 1702 1700
123 A 16-Sep PA #2 STANDARD PARKER FINE KEMPE 175mm* 1.6/1.9

STDOEV 3.7
667
3.1 0.9 .
0.3 0.7 0.4 03 0.3 0.3
944 1282 1447 1518 1608 1836 1665 1693 1713
AVERAGE
N O CUSHION 30mm SPACER STDOEV 3.4 3.4 3.6 2.7 2.1 3.7 2.3 1.6 1.9 OS
124 FINE KEMPE 175mm*
16-Sep PA #2 5 3 m m INSERT PARKER
30mm SPACER
1.6/1.9 AVERAGE 471 1092 1329 1341 1349 1346 . 1348 1350 1422
125 21-Sep PA #2
N O CUSHION
8 1 m m INSERT PARKER FINE KEMPE 175mm* 1.6/1.9
STDOEV
AVERAGE
2.7
597
6.0
1368
0.6 0.7 0.4 1.0 - 0.3 0.5 0.8
1426 1426 1424 1439 1442 1451 1448 1451
N O CUSHION 30mm SPACER STDOEV 2.5 1.3 0.8 0.5 0.5 0.2 0.2 0.6 0.4 0.2
126 23-Sep PA #2 STANDARD PARKER NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 780 1092 1441 1551 1631 1623 1657 1702 1712 1721
N O CUSHION 30mm SPACER S T D OEV 4.8 2.9 3.9 1.6 1.3 1.9 0.8 1.2 0.3 0.2
127 29-Sep PA #2 5 3 m m INSERT PARKER FINE KEMPE 175mm + 1.8/1.9 AVERAGE 709 1045 1285 1347 1359 1364 1366 1368 1366 1365
30mm SPACER STDOEV 2.7 3.5 2.7 0.7 0.3 0.3 0.2 0.5 0.1 0.1
128 01-Oct PA #2 STANDARD PARKER NORMAL KEMPE 175mm + 1.8/1.9 AVERAGE 809 1081 1351 1569 1604 1660 1684 1711 1713 .
30mm SPACER S T D DEV 2.1 2,2 2.7 1.6 1.5 OS 0.9 0.5 0.6 .
129 OS-Oct PA #2 STANDARD PARKER FINE KEMPE 175mm * 1.6/1.9 AVERAGE 717 1055 1328 1478 1614 1658 1888 1741 1750 1746
30mm SPACER STDOEV 3.2 2.9 2.7 2.4 1.6 1.1 0.6 0.7 0.4 0.5
EX-PLANT EFFECT OF SPOOLS & SEAL AGE

TYPE TIME (a) 0.6 1.0 1.6 2.0 2.6 3.0 3.5 4.0 4.6 6.0
130 07-OM PA #2 EXPLANT PARKER NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 805 1107 1396 1628 1655 1740 1651 1681 1791 1788
1248 30mm SPACER STDOEV 2.9 3.4 3.4 1.6 3.9 1.0 0.5 1.2 0.7 1.1
131 08-Ocl PA #2 OLD SEALS PARKER NORMAL KEMPE 175mm + 1.8/1.9 AVERAGE 610 1056 1320 1349 1355 1372 1370 1388 1369 1371
53mm INSERT 1248 30mm SPACER STD DEV 2.5 2.4 2.1 1.3 0.7 0.3 0.3 0.3 0.4 0.3
132 10-Ocl PA #2 NEW SEALS PARKER NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 872 1146 1394 1509 1565 1589 1635 1861 1704 1699
STANDARD 1248 30mm SPACER STDOEV 3.1 3.0 2.0 1.3 1.6 1.2 0.6 0.3 2.6 0.3
133 21-Oct PA #2 NEW SEALS PARKER NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 696 1313 1419 1415 1392 1497 1419 1386 1415 1413
53mm INSERT 124B 30mm SPACER S T D DEV 2.1 1.5 0.5 0.5 0.3 0.6 0.6 0.8 0.2 0.2
ATLAS COPCO
TYPE TIME (a) 0.6 1.0 1.5 2.0 2.5 3.0 3.6 4.0 4.6 6.0
142 27-Oct PA #2 STANDARD 9001 NORMAL KEMPE 175mm + 1.6/1.9 AVERAGE 857 1142 1511 1602 1660 1703 1730 1758 1771 1776
30mm SPACER STDOEV 3.6 2.0 1.3 1.8 1.5 1.0 0.7 0.7 0.5 0.7
143 27-Ocl PA #2 53mm INSERT 9001 NORMAL KEMPE 175mm* 1.6/1.9 AVERAGE 763 1105 1330 1359 1371 1372 1376 1388 1377 1384
30mm SPACER S T D DEV 2.4 2.4 1.3 0.5 0.8 0.3 0.4 0.5 0.7 0.7
144 28-Oct 53mm INSERT 9001 FINE KEMPE 175mm* 1.6/1.9 AVERAGE 701 1109 1361 1356 1392 1418 1423 1419 1419
PA #2
30mm SPACER S T D DEV 2.8 1.8 1.8 1.1 1.1 0.5 0.5 0.5 0.4
145 9001 FINE KEMPE 175mm* 1.6/1.9 AVERAGE 878 1181 1490 1638 1770 1794 1819 1861 1868 1881
28-0« PA #2 STANDARD
30mm SPACER S T D DEV 3.7 3 3.8 1.9 0.6 1.3 0.6 0.6 0.8 0.5
Notes: (I) All shot size m a s s units are g

(ii) Average and standard deviations are based on 25 consequitive shots of alumina.
(iii) Stroke time is "downstroke time/upstroke time" in s.
APPENDIX 3
ALUMINA SIZE AND DENSITY

Appendix 3 Alumina Size and Density Pagel
(a) A L U M I N A SIZING
(a) N O R M A L A L U M I N A R U N S (JAN - SEP.92) (b) FINE A L U M I N A R U N S (JAN -AUG,92)
RUN +100# +325# D10 D50 090 <21 FFT RUN +100# +325# D10 D50 D90 <21 FFT
MICRON MICRON
7 8.0 96.8 52.4 114.1 172.1 5.0 1 8.0 96.8 18.1 105.8 167.1 10.0 10.0
9 8.0 96.4 47.9 112.7 170.6 5.6 3 7.3 96.0 16.0 104.3 165.6 12.9
10 7.2 96.8 60.5 115.8 173.0 5.2 4 7.2 94.4 33.5 109.4 171.6 7.3
11 7.6 96.0 61.1 115.3 178.3 4.5 5 6.8 96.0 36.2 109.6 173.1 7.4
12 6.4 96.4 56.0 114.2 169.9 5.3 6 8.0 96.8 34.2 110.1 170.4 7.2
22 8.8 93.6 56.5 112.3 171.3 4.2 3.8 19 6.8 96.0 60.7 116.7 181.3 6.0 5.2
23 8.8 96.4 68.9 117.1 179.3 3.8 20 7.2 96.0 38.5 111.2 172.4 5.7
25B 6.8 96.4 65.4 114.2 168.3 3.1 3.5 25A 8.8 96.4 57.1 112.3 168.4 4.4 4.5
26 8.4 97.2 63.2 107.1 158.3 1.5 3.6 29 8.0 98.0 57.1 110.5 164.9 4.1 4.3
27 9.6 98.0 70.1 112.6 170.4 3.9 3.4 37 9.6 98.4 59.0 110.0 164.5 3.7 3.6
28 8.4 99.2 69.5 112.3 169.8 1.2 3.5 38 9.2 98.0 57.2 109.9 165.0 4.1 4.1
30 8.3 96.9 74.7 119.8 173.4 1.7 3.4 39 8.8 98.0 57.9 110.2 166.0 4.1 4.2
31 8.2 97.5 71.4 113.9 171.8 3.2 47 8.8 97.2 63.9 110.5 163.1 2.7 3.5
32 7.9 97.4 77.3 119.1 177.2 3.3 49 8.4 97.2 69.2 112.0 170.0 1.1 3.4
33 8.2 97.5 71.4 113.9 171.8 3.2 52 9.6 96.4 62.6 113.2 170.9 3.7 3.9
34 8.9 97.4 95.5 118.0 177.8 3.3 53 9.2 97.2 65.5 114.2 167.1 3.1 3.8
35 9.6 98.0 70.1 112.6 170.4 3.9 3.4 55 7.6 97.2 71.9 113.5 171.2 3.8 3.3
36 10.0 97.6 1.2 3.4 56 7.6 98.0 74.9 131.7 373.4 3.7 3.3
40 8.4 97.1 75.7 117.8 177.3 0.8 3.3 59 10.8 98.4 74.4 116.7 169.0 3.8 3.3
42 7.6 98.8 69.6 110.1 176.6 3.5 3.2 61 8.4 97.6 72.3 115.2 167.1 3.7 5.4
44 8.8 99.2 73.4 113.1 163.9 3.5 3.2 62B 7.6 98.0 73.3 114.9 166.2 3.8 3.3
46 8.0 98.8 69.9 110.4 167.0 2.8 3.2 67 7.6 98.4 71.0 112.3 163.2 3.7 3.3
50 8.8 98.8 76.5 113.7 163.1 3.3 3.2 70 10.6 98.3 72.5 113.7 173.1 3.4 3.3
54 11.2 98.8 72.7 113.4 170.8 3.4 3.3 74 9.6 98.4 75.4 116.5 176.1 3.3 3.3
57 7.6 98.0 74.6 115.7 166.5 2.6 3.3 77 9.1 98.3 75.1 115.6 173.5 3.4 3.2
58 8.0 98.4 78.8 120.8 174.9 2.7 3.3 79 10.2 99.1 75.1 115.7 170.1 3.2 3.2
60 8.4 98.8 3.5 3.2 AVE 8.5 97.3 58.6 112.9 177.1 3.8 4.8
62A 10.8 98.8 76.7 116.4 169.4 3.4 3.3 S.D. 1.1 1.1 17.6 4.9 39.5 1.8 2.3
63 8.8 98.8 73.6 113.8 170.5 3.5 3.2 102 8.1 97.5 40.7 107.0 170.5 7.3 7.2
64 8.8 98.4 70.6 111.2 161.9 2.7 3.2 103 9.4 95.0 39.7 105.2 169.4 7.3 6.2
66 8.8 99.2 3.4 3.2 105 6.5 95.4 52.3 107.2 166.5 5.2 4.7
71 10.5 97.9 76.1 116.9 175.6 3.4 3.3 107 6.3 95.6 57.7 109.5 174.1 4.1 4.2
72 10.5 99.2 82.0 120.4 172.5 3.4 3.2 110 7.3 96.4 53.0 105.1 163.6 4.4 4.7
76 10.2 98.9 77.0 117.7 170.8 3.4 3.2 112 7.9 96.2 57.5 106.0 164.3 4.1 4.7
78 9.6 98.3 74.5 115.0 172.4 3.4 3.1 116 6.7 96.4 54.3 104.7 165.4 4.0 4.7
AVE 8.7 97.8 70.4 114.7 171.2 2.8 3.6 118 7.4 97.4 54.2 104.5 160.7 4.1 3.4
S.D. 1.1 1.2 9.1 3.1 4.8 1.1 0.7 A V E 7.5 96.2 51.2 106.2 166.8 5.1 5.0
101 5.0 96.0 50.7 107.5 174.2 5.7 4.9 S.D. 1.0 0.8 6.6 1.6 4.0 1.3 1.1
104 7.5 97.0 53.3 105.5 163.9 5.0 4.4
106 6.5 98.0 59.9 107.6 169.6 3.4 3.9
108 6.2 97.4 70.0 112.2 170.1 1.7 3.7
109 7.0 97.6 56.0 107.5 154.5 1.6 3.6
111 7.6 97.1 66.3 109.4 166.8 1.3 3.5 (c) U N R E A C T E D A L U M I N A R U N S (JAN.92)
113 6.6 97.1 70.3 110.9 165.1 1.0 3.5
115 7.4 97.4 65.5 106.0 162.5 1.2 3.4
117 8.0 97.9 66.4 106.2 166.5 1.2 3.7 RUN + 100# +325# D10 D50 D90 <21 FFT
119 7.5 98.0 65.4 104.8 161.3 1.1 3.5 MICRON
120 8.0 97.6 59.4 109.4 166.7 0.6 3.4 13 9.4 94.2 36.5 113.6 170.5 6.1
121 7.2 97.3 72.5 113.1 171.6 0.7 3.4 15 7.8 93.9 38.6 113.6 170.2 6.5
122 8.0 97.7 67.7 106.5 162.2 0.6 3.3 16 6.8 94.8 46.4 113.3 174.8 5.4
122A 8.3 98.1 70.8 106.6 166.4 0.5 3.4 17 7.6 96.8 30.3 106.9 172.9 5.1
123 7.0 98.5 67.2 105.2 161.4 0.4 3.3 18 7.6 94.8 54.1 115.4 174.1 5.5
AVE 7.2 97.5 64.1 107.9 165.5 1.7 3.7 AVE 7.8 94.9 41.2 112.6 172.5 5.7
S.D. | 0.8 0.6 6.5 2.5 4.7 1.6 0.4 S.D. 0.9 1.0 8.3 2.9 1.9 0.5
Notes: (i) +100* and +325# refer to Tyler mesh. Values are cumulative percent oversize.
(ii) D10, D50 and D90 refer to the micron size above of which 10%, 5 0 % and 9 0 % of the sample is gre
(iii) "FFT" is funnel flow time ie. the time in minutes that 10Og of alumina flows from a standard funnel.
The longer the time, the finer the size. This is a standard analysis within the alumina industry.
There is almost a linear relationship between sub 20 micron sizing and FFT.
Appendix 3 Alumina Size and Density Page 2
(b) ALUMINA DENSITY COMPARISON

(RUNS 1-18)
ALUMINA LOOSE PACKED RUN
TYPE DENSITY DENSITY
NORMAL 1030 1163 22

1026 1153 23
1020 1200 7
1020 1220 9
1033 1180 10
1030 1181 11
1034 1185 12
AVERAGE 1028 1183

STD DEV 5 21

FINE 1030 1270 1

1040 1270 3
1058 1186 4
1055 1185 5
1063 1203 6
1034 1210 19
1030 1182 20
AVERAGE 1044 1215

STD DEV 13 36

UNREACTED 990 1200 13

980 1210 15
1008 1157 16
1015 1145 17
1008 1146 18
AVERAGE 1000 1172

STD DEV 13 28
Note: Density measured as kg/m3

APPENDIX 4
BOB SEYMOUR NOTE ON FEEDER

RECOMMENDATIONS
Appendix 4 Bob Seymour M e m o on Feeder Recommendations Pagel
R. E. Seymour Smelting Managers

Primary Metals
Knoxville 12 1992 April 16
Alumina Feeder Improvement
A "main-ring event" at the Feed Alumina Network meeting in

Wanatchee April 9-10 was art update by Jim Kissane (Portland) on
alumina feeders. Earlier work w a s contained in the report Pot
Alumina Feeder Design Review. J. P. Kissane. August 1991. Jim did
an excellent job of informing Feed Alumina contacts using a standup
presentation, a notebook to document findings, and a videotape.
The best companies in the world are able to operate alumina feeders
for 5-7 years (the life of the cathode) without replacement. Alcoa
plants range from 6 to 36 months. W e are definitely not a
benchmark. It appears that Karmoy (Norsk Hydro), w h o holds the
benchmark for anode effect frequency (0.03 AE/PD), also holds the
benchmark for feeder life (about 7 years). I wonder if that is
coincidence?
As i see it, Alcoa has managed feeders even worse than we have
managed pot life. W e know the average life to failure, but little if
anything about failure age distribution or root causes. Not unlike
potlining, our focus has been on improving change out procedures
rather than systematically eliminating causes of failure.
Attached are recommendations from the work done at Portland that
can be acted upon. Most recommendations apply at all locations,
although there m a y be exceptions. Here is an opportunity to rapidly
transfer technology important to a Quantum Leap Milestone.
One common theme in the recommendations is targets. I suspect
that few locations have ever measured Feeder Downstroke Time or
Feeder Dwell Time and that even fewer locations have specification
nominals and tolerances. It would seem to be a natural for pot
operators to check these characteristics when they investigate
multiple anode effects. Sampling and control charting or regular
auditing also seem to be in order.
Appendix 4 Bob Seymour M e m o on Feeder Recommendations Page 2
Feeder Improvement Recommendations

1. Establish a system to track alumina feeder maintenance by
failure type, pot, and component and a performance measurement
system using Pareto and Weibull analyses. Portland has developed a
tagging and computer system that is to be adopted by Point Henry.
There is probably an advantage in adopting the Portland
classification of failures, since w e could then compare performance.
2. Establish a system to require suppliers of critical feeder
components - pluggers, seals, springs, etc - to provide evidence of
that product specifications are being m e t Are their manufacturing
processes in control and capable?
3. Measure distribution of feeder downstroke time (time from
solenoid energized to cylinder full down) for each feeder system.
Establish downstroke time targets for each system and reduce
variability around the target. A suggested target is:
Downstroke Time - 1.5 seconds
Tuning will be required to assure that the crust is broken reliably.
Each combination of pot type, cylinder type, and valving is a unique
system. Causes of variability can include blocked mufflers, piston
seal leaks, solenoids, etc. Avoid tuning the system to take care of
special cases - pots with leaky seals or plugged mufflers. High
cylinder velocity increases wear by a cubed factor.
4. Survey blocked (mounded, dry packed) feeders and analyze the
causes several ways including potroom location such as section or
room, work activity, time of day, and anything else you can think of.
5. Reduce feeder dwell (time that plugger is in full down position) to
minimize plugger wear. A suggested target is:
Dwell Time - Downstroke time + ( 1.2 to 1,5 seconds).
6. Test dual dwell times - sv^ry third feeder cycle has a long dwell
dwell time to break the crust. T h e first two cycles are short dwell
times - just enough to let the alumina get dumped. Not all systems
may have the hardware to implement dual dwell times.
7. Establish measurable criteria for new/rebuilt feeder performance

and establish a system of quality checks to assure that criteria are
met. Spring length, air cushion adjustment, plugger diameter, rod
seal leakage, piston seal leakage and mounting bolt tightness (loose
bolts cause arcing) are important criteria.
8. Install "spool inserts" in feeders on a change out basis to make
feeder shot weights more robust to changes in alumina flowability.
Portland will recommend a design for each plant
9. Change specifications for pot pluggers to the following:
Chemical Composition
Element C Si Mn Cr Mo S P Cu
Nominal 3.1 1.7 0.7 1.4 0.5 0.0 0.0 0.16
USL 3.2 1.4 0.5 1.2 0.4 0.1 0.1 0.18
LSL 3.0 1.2 0.8 1.5 0.6 0.0 0.0 0.20
Hardness BHN(300O)
Nominal 350
USL 400
LSL 300
N O T E : The new specification will produce a much harder iron
that cannot be machined. The holes will have to be cast in
place, or a steel body will have to be cast into the plugger.
Shape
Flat Bottom
10. Survey pots for piston seal leakage, which contributes to high
air consumption and low air pressure.
11. Implement location specific recommendations relating to piston

seals, washers, springs, and other cylinder components. These can
be found in the notebok supplied to your location by Jim Kissane.
A reasonable goal for feeder life should be to equal the goal for
potlife. Improved feeder life will support quantum leap
improvements In anode effects, power consumption, and safety. In
addition, longer feeder life offers the opportunity to substantially
reduce maintenance costs.
USMS plants should review the attached recommendations, the
Portland videotape and the summary from Jim Kissane's notebook.
Be prepared to discuss your ideas and plans for these
recommendations during the M a y round of S S L T visits. For each
aspect of feeders - purchased components, rebuilt superstructures,
operation, reliability - accountability should be clearly defined.
S o m e questions to oonsider are:
W h o Is accountable?
What systems make accountability operational?
What data is monitored?
R. E. Seymour
distribution:
^melting Managers
L.E. Tate - Badln
V.C. Adorno - Massena
W.J. Drake - Rockdale
CJ. Siizewski - Tennessee
D.W. Willett - Warrick
D.J Carney - Wenatchee
G.A Turnbow - Sao Luis
D.G. Judd - Portland
L.B, Davey - Point Henry
copies:
G.J. Pizzey - Melbourne
F.G- Tigre - Sao Paulo
B.C. Raw© - Knoxville 12
R.R. Taylor - Knoxville 12
K.A. Isakson - Knoxvilte 10
J.L. Roddy - Knoxville 10
C.A. St. Clair - Knoxville 10
APPENDIX 5
DICK TAYLOR NOTE ON RESULTS OF FEEDER

RESEARCH
Appendix 5 Dick Taylor Memo on Results of Feeder Research Pagel
A L U M I N U M C O M P A N Y OF AMERICA
12W RfVgRV^W TOWER
900 SOUTH GAY m~m£~
KN0XVSJJ;. T£H?C0SE£ 37902~984!J a
1994 July 25 ALCOA
HE: J A M E S P. K I S S A N E - P O R T L A N D
The efficiency 0* the eieeueiytie process of converting aluminium oxide (alumina) to

aluminium is primarily dependant on the control of alumina thai is dissolved in the
cryolite bath {electrolyte) T h e ideal percent of alumina in solution is approximately
2.0 to 2,5 percent in most cells- W h e n the alumina in solution increases significantiy
above the ideal level, the process becomes unstable and unproductive for a long
period of time, W h e n the alumina in solution decreases below the idea percent, the
resistance of the ceil dramatically increases to a level requiring immediate corrective
action. This latter condition is named an 'anode effect". Fortunately, an anode
effect can be readily corrected.
Historically, the electrolytic process has been controlled by reducing the percent
alumina in solution below ideal forcing an anode effect several times per day. During
an anode effect w h e n the cell resistance increases and the pot voltage exceeds
approximately 8 volts tetrafiuromethane and hexaflluooethane gases are generated.
These gases are not generated during normal operation of the celL Studies by
atmospheric scientists indicate that these gases contribute to global warming.
A program was established at the Alcoa international smelting locations to reduce the
number of anode effects per operating cell day • » % -to 1990- An objective
suggested by the United States E P A w a s a 3 0 % to 6 0 % reduction of anode effects
per cell day by the year SQO0. Improved operating efficiency w a s a secondary
objective of this program.
Alumina is fed into the cett by an alumina feeder which is controlled by a computer
based on cell resistance, T h e functionalfy of the alumina feeder is a key enactor to
the control of the percent alumina in solution and, consequently, a key enablertothe
reduction of anode effects.
Each of the nine international locations Initiated a program to improve the
functtonaJty of the alumina feeder. J a m e s P. Kissane participated in this activity.
Over time and on his o w n Initiative, Jim Kissane conducted an in-depth research
project on the functionality of the feeder and also on the computer control logic and,
to s o m e extent, the related ceil chemistry. H e assisted in establishing an alumina
feeder information network between the international locations and travelled
extensively conducting investigations and seminars. Jim is recognized as the leader
of our programs related to alumina feeders.
Appendix 5 Dick Taylor M e m o on Results of Feeder Research Page 2
James P. Kissane
Page T w o
1994 July 25
The improvement in the design, control and functionality of the alumina feeder is a
success story and w e recognize Jim Kissane as the primary leader of this program.
More importantly has been his contributions to the understanding and philosophy of
cell operation relative to alumina feeding and the feeder control functions of the
computer control system.
Jim Kissane's research activities have been a key enabler to achieving the reduction
of anode effects per cell day in the international locations and, consequently, the
significant reduction of perfluorocarbon emissions. A graph indicating our experience
in minimizing anode effects per cell day is enclosed for your reference.
Sincerely,
/^£&*^*JSI A~* yz
Richard R. Taylor
Director of Technology
Primary Metals Division
A N O D E EFFECTS PER POT DAY

INTERNATIONAL PREBAKES
1.5
0.5
BASE J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J J A S O N D J F M A M J
1991 1992 1993 1984 1995
\SHARE\BISHOPVAlFUKWHUNTAE-PD
APPENDIX 6
A2 U.S.A. PATENT
APPENDIX 6 A2 U.S.A. PATENT PAGEI
The Commissioner of Patents

and Trademarks
Has received an application for a patent
for a new and useful invention. The title
and description of the invention are en-
closed. The requirements of law have
been complied with, and it has been de-
termined that a patent on the invention
shall be granted under the law.
Therefore, this
United States Patent

Grants to the person or persons having
title to this patent the right to exclude
others from making, using or selling the
invention throughout the United States
of America for the term of seventeen
years from the date of this patent, sub-
ject to the payment of maintenance fees
as provided by law.
Commwionet of Patents aad Trademarks
Amu
APPENDIX 6 A 2 U.S.A. P A T E N T
PAGE 2
US005324408A
U n i t e d States Patent [19] [it] Patent Number. 5,324,408
Kissane [45] Date of Patent: Jun. 28, 1994
[54] APPARATUS FOR CONTROLLED SUPPLY

OF ALUMINA FOREIGN PATENT D O C U M E N T S
[75] Inventor: James P. Kissane, Portland, 31274/71 7/1971 Australia .
Australia 66481/74 3/1974 Australia .
73198/81 7/1981 Australia
[73] Assignee: Portland Smelter Services Pry. Ltd., 14784/88 10/1988 Australia
Victoria, Australia
OTHER PUBLICATIONS
[21] Appl.No.: 30,167 Patent Abstract, Metallurgy, p. 18, Aliunimum electro-
[221 PCT Filed: Apr. 29,1991 lyser 83-806124/44. ( D e c 1982).
Patent Abstract, Metallurgy, p. 23, Altuninjrum produc-
[86] PCT No.: PCT/AU91/00169 ing electrolyser .... 90-374562/50 (Apr. 1990).
§ 371 Date: Apr. 2, 1993 Primary Examiner—Donald R. Valentine
§ 102(e) Date: Apr. 2,1993 Attorney, Agent, or Firm—Birch. Stewart, Rolasch &
Birch
[87] PCT Pub. No.: WO92/06230
[57] ABSTRACT
PCT Pub. Date: Apr. 16, 1992
Apparatus for the controlled supply of alumina to an
[30] Foreign Application Priority Data electrolysis tank having an electrolyte crust breaking
Oct. 5, 1990 [AU] Australia PK2658 plunger (2) includes a supply chamber (19) connected to
the entry port (16) of a dose holder (10). Alumina leav-
[51] Int. a.5 C25C 3/14 ing an exit port (15) of the dose holder (10) passes via an
[52] U.S. CI 204/245 inclined wall (9) to a delivery chute (21) which directs
[58] Field of Search 204/67, 243 R-247 the alumina to a hole formed in the crust by the plunger.
[56] References Cited Valve means (14) movable with the plunger (2) controls
the opening of the dose holder entry and exit ports (16,
U.S. PATENT D O C U M E N T S 15), closing one port as it opens the other, and allowing
3,371,026 2/196S Kiley et al 204/245 alumina to flow through the delivery chute (21) as the
3,901.787 8/1975 Niizeki et al 204/245 plunger (2) is retracted from the crust. T h e plunger
4.049.529 9/1977 Golla 204/245 X movement required to control the valve means is such
4,321,115 3/1982 Rebmann et al 204/245 X that alumina can be fed into the tank substantially con-
4,328,085 5/1982 Friedl et al _. 204/245 tinuously without meeting interference from the
4,435,255 3/1984 Casdas 204/245 X plunger.
4,437,964 3/1984 Gerphagnon et al 204/245
4,525.105 6/1985 Jaggi 204/245 X 5 Claims, 1 Drawing Sheet
5,045,168 9/1991 Dalen et ai 204/245
APPENDIX 6 A2 U.S.A. PATENT PAGE 3
U.S. Patent June 28, 1994 5,324,408

5,324,408
1 2
crust formed on the surface of molten electrolyte, the
APPARATUS FOR CONTROLLED SUPPLY O F crust breaking mechanism including a plunger with a
ALUMINA cutting edge mounted on a reciprocable plunger shaft,
and an alumina storage container adapted to release
B A C K G R O U N D O FT H E INVENTION 5 alumina as required for entry into the electrolyte
I. Field of the Invention through the hole in the crust, characterized in that the
This invention relates to an apparatus for the con- storage container feeds alumina through an alumina
trolled supply of alumina or other solid materials to an supply passage and an entry port into a supply chamber
electrolytic tank in which the alumina is converted to defined between an inner wall of the feeder assembly
10 an a n
aluminium. ^ outer supply chamber wall; a supply chamber
2. Description of Related Art exit port controlled by a valve means connects the sup-
In the electrolysis of alumina, solid alumina is dis- ply chamber to a dose holder having an inner wall
solved in a tank or pot containing a molten electrolyte mounted around and concentrically with the plunger
such as cryolite and it is desirable to maintain the alu- shaft; the inner wall is urged downwardly towards the
mina concentration in the electrolyte within a predeter- " head of the plunger; an entry port in the dose holder is
mined range. In current practice for the electrolysis of immediately adjacent to the supply chamber exit port so
alumina, the alumina is fed in successive doses of prede- that when the valve means opens the supply chamber
termined size into one or more holes which are made in exit port, it simultaneously opens the dose holder entry
the electrolyte crust so that the alumina can be admitted port and alumina in the supply chamber is able to flow
when required. A s the electrolysis of the alumina pro- 20 directly to the dose holder, the valve means is opera-
ceeds continuously, it would be desirable if the alumina pvely associated with the inner wall so as to m o v e in
consumed in the electrolysis process could be continu- response to the movement of the inner wall between a
ously replaced so as to maintain the optimum alumina { j ^ position in which the dose holder is closed to the
concentration in the electrolyte. However, the optimum supply chamber and a second position in which the dose
operating conditions are such that the electrolyte crust 15 holder is opened to the supply chamber, the valve
continuously reforms on the surface of the electrolyte m e a n s being open in itsfirstposition to a flow passage
making it difficult to continuously supply alumina to the defined between the inner wall and the valve means and
molten electrolyte beneath the crust. For this reason, m {ts ^^4 position closing off the dose holder from
known alumina feeding procedures involve the use of a the fl<JW p^g,.. Ae d o s e holder i, a chamber defined
crust breaker which is operated intermittently to break 30 fe Quter w a U > t w j radially m w a r t U y directed end
the electrolyte crust and form a hole through which the w a ] k a n d a radiaUy m w a r d m o v a W e wall f o r m e d b y th e
solid alumina can be fed However, the action of the ^ ^ m e M ^ m o v a b l e ^ definingvtilh ^ TCSptc.
crust breaker is necessarily such that the crust breaking dve ^ ^^ allcriiativrfy. depending on the position
mechanism, such as a pneumatically operated shaft with ^ ^^ ma^s g d M e ^ ^ ^ leading tQ
an appropriate chisel means O-eranafter referred to as a 35 of , ^ ^ ,y
plunger) at «s free end. will be moved m and out of the cbgmbeT^^t w h e n ^^ j ^ * m y ctose d, the
hole formed by the plunger. h jm ^ the l o w e f ^ ^ o f ^ d o s e
fa one k n o w n feedmg procedure, two separate pneu- downwardly and inwardly inclined towards a
mane systems are employ^ « « P ^ t e c « . fonned ^ ^ wa] , ^ definin ^
breakmg mechanism ^*"J^J^**J*^ * lower part of the dose holder exit port; the valve seat in
mina feeding system. In th* procedure, it is pebble for P £ ^ ^ ^ the
the mechamsrn operating the crust h r ^ g mechanism ^wLILvel .JÛ ^ v a l v e P m e a n s a g a m s t the
to form the necessary hole w the deôlyte crust and £ Qf ^ ^ ^ ^ ^ ^ the
retract the crust breaker so that the feeding system can OUWUWMU «SUI6 ftlrfW wn^nâniiv tn
then be operated to place a charge of alumina into the 45 P ^ c shaft m a y be driven further downwardly to
break
^ 7 fit^HKv tn7 i™«breaker * « electrolyte crust; striker means on the plunger
"In a T ^ r p r o ^ u S S P-umatic system is shaft which meets the lower edge of the inner wall as
u s S to opiate the c n ^ brealdng m e c h a n i s m , ^ the the plunger shaft ts raised from t t s , ™ l ^ Z ° % * -
discharge of alumina from a storage device is co- *>n and raises the mner wall and « s j a s ^ t t d valve
ordinatfd with the downward movement of the crust 50 means to close the entry port and " P « * J « * * « ^
breaker. In this procedure, the alumina charge is thus the dose hoWer. and an mchned w a U connected adja-
released w h e n the crust breaker is through the crust so cent to the lower end of t h « ™ « J ^ f * ' f « f "
that the alumina is not free to enter the hole in the crust «*«nbly and terminating at its lower free edge at or
untilftecrust breaker is retracted. While this procedure within the entry pernor.of a delivery chute adapted to
has the advantage of a single pneumatic system, it is 55 b* mounted below the feeder assembly and to provide a
obvious that not all of the alumina will be able to pass funnel-like action to direct alumina which leaves the
through the hole into the electrolyte immediately when dose holder to one or more outlets terminattng m use
,u. *£* h r M i , „ « „*„^t-i above the hole m the electrolyte crust.
the crust breaker is retracted. ^ ^ ^ a M e m b l y rf ^ pr(sent m v e n t i o n ^ ^
SUMMARY OF THE INVENTION w a crusl breaking mechanism which is preferably pneu-
It is an object of the present invention to provide an matically operated. The crust breaking **£™*» £
unproved alumina feeder assembly which can utilize a eludes a plunger with a cuttmg edge for b " * ^ * * 6
sinale pneumatic mechanism but avoid the disadvan- crust mounted on a reciprocable plunger shaft T h e
tages of the k n o w n system using such a single mecha- plunger shaft preferably carries striker means which
J ^ * 65 m a y consist of a collar adjacent the plunger or a shoul-
Accordingly the present invention provides a feeder der by the junction of the plunger shaft and the plunger.
assemWyfof an alumkTelectrolysis tank including a T h e feeder assembly further includes at least one
crust breaking mechanism operable to break a hole in storage container comprising a hopper or like vessel tor
APPENDIX 6 A2 U.S.A. PATENT
PAGE 5
5,324,,408
3 4
finely divided alumina. Other storage containers m a y be exit port leading to the flow passage or an entry port
associated with the feeder assembly for other additives leading to a supply chamber. T h e nature of the port in
to the electrolysis tank such as aluminium fluoride, the dose holder is controlled by the movement of the
calcium fluoride, crushed bath, soda ash, or cryolite. valve means so that when either port is fully closed, the
The other storage containers m a y be adapted to feed 5 other is fully open.
their contents into the tank in a similar manner to that Preferably the lower end wall is substantially down-
described below for the alumina. wardly and inwardly inclined at an angle greater than
A supply chamber provided between the storage the angle of respose of the alumina powder which is to
container and a dose holder includes a preferably sub- be fed through the dose holder. This inclination of the
stantially cylindrical inner wall mounted around and 10 lower end wall ensures that all the alumina powder
concentrically with the plunger shaft. T h e inner wall is (other than that held in the annular seat) will flow from
urged downwardly towards the head of the plunger, the dose holder w h e n the exit port is open.
preferably by spring pressure exerted between a radially T h e inclination of the upper end wall is substantially
outwardly extending flange on the inner wall and a downwardly and outwardly. T h e upper end wall is
feeder assembly outer wall which is also mounted con- 15 preferably also inclined at an angle greater than the
centricaUy with the plunger shaft T h e feeder assembly angle of repose of the alumina powder which is to be
outer wall m a y include a radially extending flange more fed through the dose holder. This inclination of the
remote from the plunger head than the flange on the upper wall ensures that the dose chamber will be filled
inner wall so that a coil spring mounted between the with alumina, thus providing the desired accurately
respective inner wall and outer wall flanges can exert 20 reproducible dosage.
the desired pressure urging the inner wall downwardly T h e annular seat in the lower end wall not only pro-
until its d o w n w a r d movement is terminated. T h e spring vides a means of sealing the exit port of the dose holder.
is mounted in the upper portion of the supply chamber It also provides a stop to terminate the downward
so that alumina in the supply chamber will not interfere travel of the valve means and the associated inner wall
with the spring operation. 25 which occurs when the plunger shaft is lowered in
The supply chamber is defined between the inner response to the downward urging of the spring or other
wall of the feeder assembly and a preferably substan- pressure exerting means. The valve means is held in the
tially cylindrical outer supply chamber wall. T h e sup- lower end wall seat by the downward pressure while
ply chamber includes an entry port connected to an the plunger shaft m a y be driven further downwardly if
alumina supply passage below the inner wall flange and 30 the crust is to be broken.
an exit port controlled by a valve means. T h e capacity W h e n the plunger shaft is raised, means consisting of
of the supply chamber is preferably at least that of the the plunger head itself, or the preferred striker means,
dose holder. T h e inner wall at the supply chamber is meets the lower edge of the inner wall and raises it and
preferably supplemented by a substantially down- the associated valve means to close the entry port and
wardly and outwardly directed supply chamber side 35 open the exit port of the dose holder. T h e upward
wall which terminates at its lower edge by the supply movement of the inner wall is terminated w h e n the
chamber exit port. Preferably the supply chamber side upper end edge of the valve means seats within the
wall is inclined at an angle greater than the angle of annular seat in the upper wall of the dose holder.
repose of the alumina which is to pass through the sup- T h e feeder assembly further includes an inclined wall
ply chamber. This ensures that the alumina will flow 40 connected adjacent to the lower end of the inner wall.
freely through the chamber. T h e inclined wall is preferably of substantially frusto-
The supply chamber exit port is immediately adjacent conical form and terminates at its lower, free edge at or
to an entry port in the dose holder so that w h e n the within the entry portion of a delivery chute.
valve means opens the exit port of the supply chamber, T h e delivery chute is adapted to be mounted below
it simultaneously opens the entry port to the dose 45 the feeder assembly and is adapted to provide a funnel-
holder, and alumina in the supply chamber is able to like action to direct alumina which leaves the dose
flow directly to the dose holder. holder to one or more outlets which terminate in use
T h e valve means is operatively associated with the above the hole in the electrolyte crust. T h e delivery
inner wall so as to m o v e in response to the movement of chute preferably directs all the alumina leaving the
the inner wall between a first position in which the dose SO lowar edge of the inclined wall at the base of the inner
holder is closed to the supply chamber, and a second wall, towards one or more delivery outlets 5.
position in which the dose holder is opened to the sup- T o assist a further understanding of the invention,
ply chamber. In its first position, the valve means is reference is n o w m a d e to the accompanying drawing
open to a flow passage defined between the inner wall which illustrates one preferred embodiment of the pres-
and the valve means. In its second position, the valve 55 ent invention. It is to be appreciated that this embodi-
means closes off the dose holder from the flow passage. ment is given by w a y of illustration only and that the
T h e valve means is preferably substantially cylindrical invention is not to be limited by this illustration.
and is connected to the inner wall between itsfreeend BRIEF DESCRIPTION OF THE DRAWING
edges. Each of the respective free end edges of the FIG. 1 is a diagrammatic sectional view of the pres-
cylindrical valve means is adapted to seat in an annular 60 ent invention.
seat defined at the opposite ends of the dose holder. DETAILED DESCRIPTION OF PREFERRED
T h e dose holder is a chamber defined by an outer EMBODIMENT
wall which is preferably substantially cylindrical and The drawing shows, somewhat diagrammatically,
has two radially inwardly directed end walls in which one half only of a sectional view of a preferred form of
the respective annular seats are defined, and a radially 65 feeder assembly. Plunger shaft 1 is connected to plunger
inward movable wall formed by the valve means. De- 2, and shoulder 3, which is at the junction of plunger
pending on the position of the valve means, the dose
holder will always include an open port constituting an
APPENDIX 6 A2 U.S.A. P A T E N T PAGE 6
5,324,408
5 6
shaft 1 and plunger 2, abuts striker means 23 on inner L A feeder assembly for an alumina electrolysis tank
wall 4 in the position shown. Inner wall 4 is urged including a crust breaking mechanism operable to break
downwardly by spring 5 which is held between flange 6 a hole in crust formed on the surface of molten electro-
on inner waD 4 andflange7 on outer wall 8. Inclined lyte, the crust breaking mechanism including a plunger
wall 9 at the lower end of inner wall 4 is connected 5 with a cutting edge mounted on a reciprocable plunger
adjacent to the junction of inner wall 4 and the striker shaft, and an alumina storage container adapted to re-
^ ^ „ lease alumina as required for entry into the electrolyte
The chamber forming dose holder 10 is defined be- through the hole in the crust, characterized in that the
tween side wall 11 and end walls 12 and 13, together storage container 8 adapted to feed alumina Ora-j|h»o
wnTvalve means 14 which comprises the moveable "> d ™ ? " * * J W "* « W p o « % ° * £ « £
wall connected to inner wall 4. In the drawing, valve chamber defined between an inner wjdl of the feeder
open, while entry port 16 is closed. Annular seats 17 and ^ dose holder, the inner wall is
18 for m e respectiveend edges o ^ t a : meanswall 14 ^ ^ J ^ S ^ « J 2 S * t f y with the plunger
are formed m the respecOve end walls 12 and 13. ^ ^ ^ waU a
Supply chamber 19 isfilled generally below the levd ^ , ^ w a r d l toward$ ^ j , ^ of ^ pi^ger; an
offlange6 by alumma entering as mdicated by arrow 20 • fa ^ dQschoJ<Jer înmKdiatcly adjacent to
from an alumina storage container. Inclined wall 24 ^ £ «~ chamber exit port so that when the valve
supplements inner wall 4 to direct the alumma in cham- ^ ^ ^ Q ^ ,y chamber exit p^t,h simoita-
ber 19 to entry port 16. Delivery chute 21 ^connected neously ^p^ ^ dose h o i d e r entry port and alumina in
as indicated to the outermost wall of the feeder assem- ^ m ]y chamber js a b j e to flow directly to the dose
bly and is adapted to direct the alumina leaving dose holder. tne v a i v e means ;s operatively associated with
holder 10 and flowing viaflowpassage 22 and down ^ y^ wa], ^ M to ^y,, m response to the move-
inclined wall 9 into the hole in the electrolyte crust 2$ ment ofthemnef w a H ^tween afirstposition in which
which has been formed by plunger 2. It will be appreci- the dose no i de r is closed to the supply chamber and a
ated that it was only necessary for the valve means to ggcon^ position in which the dose holder is opened to
move the distance A for a charge of aluminatobe re- ^ s u p p i y chamber, the valve means being open in its
leased from the dose holder, the movement required for g ^ poô,, to a flow passage defined between the
the plunger to break through the electrolyte crust is 30 inneT ^ j m& w e ^vc means and in its second posi-
considerably greater than that represented by distance ôn dosing 0ff the dose holder from the flow passage;
A. the dose holder is a chamber defined by an outer wall.
It will be appreciated that the present invention al- ^y,, radially inwardly directed end walls and a radially
lows the design and operation of a feeder system which inward movable wall formed by the valve means, the
utilizes only a single pneumatic mechanism co- 35 movable wall defining with the respective ends walls
ordinated with the supply of alumina to the hole in the alternatively, depending on the position of the valve
electrolyte crust formed by the crust breaking mecha- means, a dose holder exit port leading to the flow pas-
nism and that the alumina can be fed directly into the j ^ or an entry port leading to the supply chamber, so
hole when the crust breaking mechanism is retracted that when either port is fully closed, the other is fully
from the hole. Although some alurnina flows directly 40 opened; the lower end wall of the dose holder B down-
through the dose holder while the valve means is being wardly and inwardly inclined towards a valve seat
moved from the position in which the exit port is open formed in the lower end wall and defining the tower
to the position where the entry port is open, substan, part of the dose holdeu exit port; the valve seat m the
tially all of the alumina released from the dose holder is lower end wall provides a stop to terminate the down-
able to flow directly into the hole in the electrolyte 4S w a r d travel and hold the valve means against the down-
crust ward urging of the associated inner wall while the
It is a further advantage of the present invention that plunger shaft may be driven further downwardly to
the downward movement of the plunger can be limited break the electrolyte crust; striker means on the plunger
when it is desired only to activate the valve means so as shaft which meets the lower edge of themner wau as
to recharge the dose holder. It is not necessary for the 50 the plunger shaft is raised from its crust breaking opera-
plungcr to travel downwardly to thefiillextent re- tion and raises the inner wall and its assoctatodivatve
quiredtobreak the crust, each time some downward meanstoclose the entry port and open the exitport ot
movement is requiredtorecharge the dose holder. The the dose holder, and an mchned ^ ? ^ ^ ^
dose holder may thus be recharged and the plunger cent to the lower end of the un^r w a U o f **Mer
retracted to release the charge of alumina, without the 55 assembly and terminating at its lew^free edge at or
nhmaer travelling fully through the electrolyte crust within the entry portion of a delivery chute adapted to
K p i r g e ^ ^ ' c o n s i d l r a b l y reduced. be mounted below the feeder ~ ^ J * » £ " * , "
The ancles of the dose holder end walls are greater funnel-like actiontodirect alumina which leaves the
t h a n ^ a S l e ofrepo^ of alumina, so the alumina dose holder to one or more outlets tenmnatmg » use
charmns rrocedure is not affected to the same extent as 60 above the hole in the electrolyte crust.
L i t K e ^ ^ e e d i n g procedures by Nations in the 2. The feeder assembly as claimed m clarm J. wherein
" « X P o T S S S K 5 S which leadstomore the lower end wall nofithesdose bold*-^smnuafly
consent
nriate dose charge
holdervolume
volume precision.
allows frequent
Selection
ûmmamtothede^rolytobathmchargessmrilerthim
current
^^concentration
charge sizes thus
more assisting
substantially
in feeding
of appro-
maintaining
constant. of 65 than
the Jdownwardly
j the
3.The ^ angle
- Jand
feederof«*?
^repose^f
X° Sy* ithe
fawardly^chned*«»gkêr
2alumma
£ , - powder
* wnicn
greaJris
"f™g^^^J^*^2,„tecin
.
- substantially
5,324,408
8
than the angle of repose of the alumina powder which is annular valve seats formed in the upper and lower end
walls of the dose bolder.
to be fed through the feeder assembly.
5 T h e feeder assembly as claimed in claim 1, wherein
4. T h e feeder assembly as claimed in claim 1, wherein the supply chamber is formed with an inclined inner
the valve means is substantially cylindrical, is connected wall which terminates at its lower edge by the supply
to the inner wall between its free end edges, and seats in chamber exit port.
U.S. Patent June 28, 1994 5,324,408

APPENDIX 7
PULSE CHUTE PATENT APPLICATION

APPENDIX 7 PULSE C H U T E PATENT APPLICATION PAGE
(Th« foilowioi 11 io Dt ruled (a oy m e receiving Office i

INTEaWATlONAL
INTERNATIONAL APPLICATION APPLICATION Ne:
UNDER THE
INTERNATIONAL
PATENT COOPERATION TREATY R U N G DATE:
REQUEST
THE USDE15IG1WD BEQUESTS THAT THE PlESEnT (Stamp/

INTERNATIONAL APPLICATION BE PIOCESSED _Nameof receiving Office and ' P C T International Application"
ACCORDING T O THE PATENT C0OPEIAT1ON TBEATY
Applicants or Agent's File Reference
(indicated by applicant if desiredl
B.i N o . I TITLE O F INVENTION
METHOD AND APPARATUS FOR CONTINUOUS SUPPLY OF ALUMINA
BoiNo.II APPLICANT (WHETHER. OR NOT ALSO INVENTOR); DESIGNATED STATES FOR WHICH WF/c.HF/iTrc
The person identified in this box is (check one only)- I { applicant and inventor*
Name aaaf address:**
PORTLAND SMELTER SERVICES PTY. LTD.

535 BOURKE STREET,
MELBOURNE, VICTORIA 3000,
AUSTRALIA
Telephone number: Telegraphic address: Teleprinter address

(including area code)
Country of nationality: AUSTRALIA Country of residence-*"
The person identified m this box is applicant for the purposes of (check one only):
i tall designated States (Ti''1 designated States except j [the United States i 'the Slates indicated
1—I L£_ithe United Sutes of America i——lof Arnenca only iw the 'Supplemental
I Jin "Supplementa Boi'
Bar No. 01 FURTHER APPLICANTS. IF ANY: (FURTHER) INVENTORS. IF ANY; D E S I G N A T E D STATES FOR
W H I C H T H E Y A R E APPLICANTS (IF APPLIC A B LE). A icparalc sub-box has to be filled in in respect of each person i includes wnere
applicable, a legal entity), if the following two sub-boxes are insufficient, continue in the "Supplemental Box." (giving there for each addi-
tional person the sane indications as (hose requested in the following two sub-boxes) or by using a "continuation sheet."
The person identified in this tub-box i> (check one only) |Xj applicant and inventor* f~J applicant only mvemoronly"
Name anal address:**
KISSANE, Janes Patrick

3 MOFFATT COURT,
PORTLAND, VICTORIA 3305,
AUSTRALIA
If the person identified in this sub-bjf is af pliant (or applicant and manor j. indicate also:
Country of nationality: AUSTRALIA Country of residence:**"
and whether that person is applicant tot the purposes of I check one only):
Gall deaienatact Statsta I l»" designated States except [~7]the United Stales j Ithe States indicated
* I I the United Stales of America Lilof America only I lin the "Supplemental Box"
The person identified in this sub-box i»< check one only I ) [ applicant and inventor" 1 applicant only , ' mvemoronly*
Name ana addreu:*"
If the person identified in this sub-oox is applicant tor applicant an4 micntor), indicate also:
Country of nationality Country of residence:***
and whether thai person is applicant tot the purposes of (check one only):
.u dnisBsatid suu. 1 1*" »«*Mjneied Susies eaeept I I the United Stales I I the Slates indicated
•
all designated Stales | [ ^ •j r J2* d j t i l t l ^ A m J ! n c a 1 lof America only I—I in the "Supplemental Box
If the perxaat m d r utrl at atypiicaeu end inventor" or as -inventor only" is not aa lanwauerfbr the purposes of all the designated States.
free the necessary wdfcauoaa in U M -Svpsteeneatal box.*
Indicate the name ofanattirel person by|hriB«hii/herfamilyiiamtni*foUowe^
. ijs full official designation la the address, include both the postal coda (if any) and the country (name).
If residence if not indicated, il will be assumed that the country of residence it the sane as the country indicated in the address.
F o n a r X T / R O / 1 0 1 <ll« she«) (January ll»JO> See notes on accompanyme sheet
APPENDIX 7 PULSE CHUTE PATENT APPLICATION-
PAGE 2
Sheet number
ff W COMMON :
CERTAIN C A ^ < J )°» " * " " E N T A T I V E ( I F A N Y ) ; ADDRESS FOR NOTIFICATIONS IN
^haêêX^ *•" *»""«•<' a. agent or common representative » a:,

Name ana address, including postal code and country: If the space below is used instead for an
PHILLIPS ORMONDE & FITZPATRICK address for notifications, mark here
367 COLLINS STREET,
MELBOURNE, VICTORIA 3000,
AUSTRALIA
Telephone number. (°3) 614 1944 Telegraphic Teleprinter

• including area code) address: address.
OR T R E I ™ 0 ! * ! ^ 1 ? T ° F ? * 0 C P S °FSTATES ° * STATtS''': C H 0 I C E ° F C E R T A I N K I N D S 0 F P R O T E C T I O N
O R T R E A T M E N T . The following designations are hereby made (please mark the applicable cheek-boxeai:
Regional Patent
and any other Slate which is a Contracting Sute of the European Patent Convention and of the PCT
:
—' °* Feneg*U.*TS»:of'"'"' ,MAiM ?"°' Cameroon- Cen,raJ African Republic, Chad. Congo. Gabon. Mali. Maur.unu.
and any other Slue which is a Contracting Sute of OAPI and of the PCT; Mother OAPI title desired, specify on dotted line'J >•
National Patent (if other kind of protection or treatment desired, specify on dotted line' *>)
I [ A T Austria'1' I j K R Republic of Korea''>
X ' AL' A ustralia*3 > I | LK Sti Lanka
BB Barbados |_j L U Luxembourg' 1 )
1
~ B G Bulgaria' ) \~J M C Monaco* 1 "
~3T B R Brazil'1) i j M G Madagascar
Xi C A Canada ^~] M W Malawi'1)
. C H and LI Switzerland and Liechtenstein 1 j N L Netherlands
; D E Germany (Federal Republic aty*) .... [X\ N O Norway
! | R O Romania
DK Denmark j ] S D Sudan
[ j E S Spain'1) | S E Sweden
j X; Fl Finland | ) S U Soviet Union'1)
I ; G B Untied Kingdom
~2 HU Hungary X; U S United States of America'1).
j f JP Japan'1)
: j RP Democratic People's Republic of Korea'J'
Space reserved for designating States (for the purposes of a national patent) which Have become party to the PCT after the issuance of this sheet.
(I) The applicant's choice of the order of designations may he indicated by marking the check-hoses with sequential arable numerals (see
also the 'Noies to Box No. V*).
(2) The selection of particular States for a Eurooean patent can be made upon entering the national (regional) phase before the European
Patent Office (sec also (he "Notes to Box No. V*).
(J) - -If another kind of protection or a title of addition or. in the United Stales of America, treatment as a continuation or a continuanon-in-pan
Form PCTYRO/IOI
is desired, (second sheet) (January
specify according iy*0)
to the instructions given in the "Notes to Boa No. V." See notes on accompanying sheel
APPENDIX 7 PULSE CHUTE PATENT APPLICATION PAGE 3
Sheet n u m b e r .
B o x N o . VI P R I O R I T Y C L A I M (IF A N Y ) . The priority of the following earlier application!si is hereby claimed

Country (country-rn which it Filing Date Application N s Office of Fitinglflll in only :f
*ei 'lied if national application: (day. month, year) the earlier application s in
one of the countries for which it international ippucatian
was filed if regional of interna- or a regional application >
tional aophcauoa)
Hi AUSTRALIA 5 OCTOBER 1990 PK 2658
1*1
Ui •
i Letter codes may be used io indicate country and/or Office of fiiing)
When the earlier application was filed with the OITiee whieh. for the purposes of the present international application, is the rece-ung Office.
ihe applicant may, against paymtnt oftnt nquini ftt. ask the following
?~r the receiving Office is hereby requested to prepare and transmit to the international Bureau a certified copy of the above-mentioned
r • earlier application/of the earlier applications identified above by the numbers I insert the applicable numbers i
Box No. VII EARLIER SEARCH (IF ANY). Fill m «nere a search iinternational, international-type or atheri by tne International
Searching Authority has already been requested lar completed I and the said Authority is now requested to base the international search.
to the extent possible, on the results of Ihe said earlier search .Identify such search or request either by reference to the relevant applica-
tion tor the translation thereof) or by reference io the search request.
International application number ai International/regional/national

number and country (or regional filing date
Office i of other application.
Date of request for search N u m b e r (if available)

given to search request
Box No. VIII SIGNATURE OF APPLICANT(S) OR AGENT

PHILLIPS ORMONDE & FITZPATRICK
Agents for Applicant:
rs*-s-^nj>
John A. Waters
If the present Request form is signed on behalf of anv n~ cam ?y an agent, a separate power of attorney appointing the agent and signed by
the applicant is required. If tn such casertit detired to, -sale i « of a general power of attorney (deposited with the receiving Office i. a copy
thereof must be attached to this form.
B o x No.lX C H E C K L I S T (To he Client m by tne \ - S . . C J M . This international application ai filed is accompanied 5y the items
checked below-
This international application contains the follow ng i.-ser of
meets I I • separate signed power of attorney
I request -rets ! | | copy of general power of attorney

I. description •rets 3 PH pnontydocumentltXteeBoxNo. vii
). claims -ee*.l
a i J receipt of the fees paid or revenue stamps
i abstract _ •es'.s
-ejtt 5 fx] cheque for the payment of fees
S.' drawings
Teial t4 -rets 6 | j request to charge deposit account
• r~] other document (tpeeify)

Figaro n u m b e r ofU»dra«nngs!rian» * digested
to accompany the abstract for publication.
(The lellewiee » •• M "lied la by Use reeervtstg Office)
I. Date of actual receipt of the purported international application
2. Corrected data af actual receipt due to later hut timely received papers
or drawings completing the purported international application
J. Data of timely receipt of the required corrections under Article 11 of the P C T .
4. Drawings
• deceived
• N o Drawings
(Tto Mtotrltii is t* be flUed In by tx* InlirailUael B a

ABSTRACT
Apparatus for supplying alumina to an electrolysis

5 tank includes a supply chamber in which alumina fed
intermittently from a storage container is fluidised by gas
injected through a gas-pervious wall- (lo) in the supply
chamber beneath the alumina and flows through a discharge
conduit (8) to a hole formed in the electrolyte crust. The
10 fluidised alumina flows upwardly to reach the. inlet of the
discharge conduit (8) past a hood (7) which prevents
non-fluidised alumina from entering the inlet. A method of
supplying alumina using the facilities provided by such
apparatus is also disclosed.
METHOD AMD APPARATUS FOR COMTTHTJOnS SUPPLY pp AL-DMIWA
This invention relates to a method and apparatus for

the continuous supply of alumina or other solid materials to
5 an electrolytic tank in which the alumina is converted to
aluminium.
In the electrolysis of alumina, solid alumina is
dissolved in a tank or pot containing molten electrolyte
such as cryolite and it is desirable to maintain the alumina
10 concentration in the electrolyte within a predetermined
range. In current practice for the electrolysis of alumina,
the alumina is fed in successive doses of predetermined size
into one or more holes which are made in the electrolyte
crust so that the alumina can be admitted when required. * As
15 the electrolysis of the alumina proceeds continuously, it
would be desirable if the alumina consumed in the
electrolysis process could be continuously replaced so as to
maintain the optimum alumina concentration in the
electrolyte. However, the optimum operating conditions are
20 such that the electrolyte crust continuously reforms on the
surface of the electrolyte making it difficult to
continuously supply alumina to the molten electrolyte
beneath the crust. For this reason, known alumina feeding
procedures involve the use of a crust breaker which is
25 operated intermittently to break the electrolyte crust and
form a hole through which the solid alumina can be fed.
However, the action of the crust breaker is necessarily such
that the crust breaking mechanism, such as a pneumatically
operated shaft with an appropriate chisel means (hereinafter
- 3 -
referred to as a plunger) at its free end, will be moved in

and out of the hole formed by the plunger.
In one known feeding procedure, two separate pneumatic
systems are employed, one operating the crust breaking
5 mechanism and the other operating the alumina feeding
system. In this procedure, it is possible for the mechanism
operating the crust breaking mechanism to form the necessary
hole in the electrolyte crust and retract the crust breaker
so that the feeding system can then be operated to place a
10 charge of alumina into the hole formed by the crust breaker.
In another procedure, a single pneumatic system is
used to operate the crust breaking mechanism, and the
discharge of alumina from a storage device is co-ordinated
with the downward movement of the crust breaker. In this
15 procedure, the alumina charge is thus released when the
crust breaker is through the crust so that the alumina is
not free to enter the hole in the crust until the crust
breaker is retracted. while this procedure has the
advantage of a single pneumatic system, it is obvious that
20 not all the alumina will be able to pass through the hole
into the electrolyte immediately when the crust breaker is
retracted.
It is an object of the present invention to provide an
apparatus allowing substantially continuous addition of
25 alumina to the electrolyte.

The feeder assembly of the present invention is used
with a crust breaking mechanism. Any known or other
appropriate crust breaking mechanism can be used. The crust
breaking mechanism may be pneumatically operated. The crust
- 4 -
breaking mechanism may include a plunger with a cutting edge

for breaking the crust mounted on a reciprocable plunger
shaft.
The feeder assembly further includes at least one

5 storage container comprising a hopper or like vessel for
finely divided alumina. Other storage containers may be
associated with the feeder assembly for other additives to
the electrolysis tank- such as aluminium, fluoride, calcium
fluoride, crushed bath, soda ash, or cryolite. The other
10 storage containers may be adapted to feed their contents
into the tank in a similar manner to that described below
for the alumina.
Each storage container is adapted to release a
predetermined amount of its contents as required to feed the
15 electrolysis tank. The mechanism employed to measure and
release the storage container contents may be any known or
other appropriate mechanism.
The present invention relates to the supply of alumina
from an alumina supply chamber to which predetermined
20 amounts or charges of alumina are fed from the alumina
storage container.
Accordingly the present invention provides apparatus
for supplying alumina to an alumina electrolysis tank
including an. alumina storage container adapted to release
25 alumina as required for entry into the electrolyte through a
hole formed in crust on the electrolyte by a crust breaking
mechanism, characterised in that the alumina storage
container is connected to an entry aperture of an alumina
supply chamber having a base wall from which inner and outer
APPENDIX 7 PULSE C H U T E PATENT APPLICATION PAGE 8
- 5_
side walls lead towards the upper portion of the chamber to

define the entry aperture; a transverse gas-pervious wall
extends between the side walls of the chamber spaced from
the base wall.and below the chamber inlet to define a plenum
5 into which gas may be introduced through a gas inlet, the
gas-pervious wall allowing gas flow sufficient to fluidise
alumina but not allowing alumina to enter the plenum; at
least one discharge conduit extends through the base wall
and the transverse wall to an outlet below the supply
10 chamber, the outlet is positioned in use above the hole
formed in the electrolyte crust by the plunger and tihe
opposite, inlet, end of the discharge conduit is located
above and spaced from the transverse wall; and a hood member
is spaced above the discharge conduit inlet but extends
15 below the level of said inlet.
The present invention also provides a method for the
supply of alumina from a storage chamber adapted for the
intermittent supply of predetermined amounts of alumina,
characterised in that the predetermined amounts of alumina
20 are fed to a supply chamber, the alumina in the supply
chamber is fluidised by gas injection through a gas-pervious
wall below the alumina, and the fluidised alumina flows from
the supply chamber through a discharge conduit.
The alumina supply chamber which forms a
25 characteristic feature of the present invention is partly
defined by a base wall and inner and outer side walls which
abut or merge with the base wall. The side walls lead from
the base wall towards the upper portion of the supply
chamber where they may merge towards one another to define

APPENDEK 7 PULSE CHUTE PATENT APPLICATION PAGE 9
- 6-
an entry aperture through which alumina fed from the alumina

storage container can enter the supply chamber.
The supply chamber further includes a transverse
gas-pervious wall between the inner and outer side walls.
5 The gas-pervious wall is preferably inclined to the
horizontal plane for the reason discussed below.
The transverse wall is spaced from the base wall
between the base wall and the chamber inlet so as to define,
with the base and side walls a plenum into which a gas,
10 preferably air, can be introduced in controlled manner. A
gas inlet is provided to enable introduction of gas into the
plenum.
The gas-pervious material used to form the transverse
wall may be any material appropriate to this application.
15 The material concerned must allow a sufficient flow to
fluidise the alumina but at the same time should not allow
the alumina to enter the plenum. The material should be
suitably resistant to the temperatures it will encounter in
use. One suitable material is a perforated or woven
20 stainless steel plate.
At least one discharge conduit extends through the
base wall and the transverse gas-pervious wall of the supply
chamber and extends below the supply chamber to an outlet.
The discharge conduit outlet is positioned in. use so as to
25 direct alumina into the hole formed in the electrolyte crust
by the plunger. The opposite, inlet, end of the discharge
conduit is located above and spaced from the transverse wall.
The construction of the invention includes a hood
member which is spaced above the inlet of the discharge

- 7 -
conduit and extends below the level' of the inlet. The

spacing of the hood member and the inlet is such as to
prevent gravity flow of alumina from the supply chamber into
the conduit as it enters the supply chamber or fills the
5 supply chamber above the level of the inlet. The hood
member preferably has a depending skirt, the lower edge of
which extends below the level of the conduit inlet.
The alumina supply chamber may be a substantially
separate vessel located in use to one side of the plunger
10 shaft. In this embodiment, the inner and outer side walls
of the supply chamber as referred to above may form a
continuous circumferential wall to the chamber. Another
form of supply chamber is substantially annular in form and
is located concentrically with the plunger shaft. The
15 supply chamber has one or more discharge conduits.
In use, the alumina supply chamber of the invention is
operated by first supplying a predetermined quantity of
alumina from the alumina storage container. The alumina in
the supply chamber is then fluidised by injecting gas,
20 preferably air, through the gas inlet to the plenum. The
injected gas passes through the gas-pervious transverse wall
to fluidise the alumina and allow it to flow past the hood
member into the inlet and through the discharge conduit.
The gas injection is controlled in relation to its
25 pressure, the time period of injection and the time between
successive injections. 3y controlling these parameters for
the injected gas in relation to the freguency and amount of
predetermined alumina supply to the supply chamber, it is

possible to maintain the desired continuity of alumina flow
- a-
from the discharge conduit. Preferably the gas will be

injected into the plenum at spaced time intervals to provide
fluidising pulses. The pulse length and time between pulses
can be varied as required. For example, it is possible to
5 maintain a substantially continuous flow of alumina through
the discharge outlet at a rate which is calculated to
maintain a supply of alumina within the supply chamber until
the release of the next predetermined amount of alumina from
the storage container. The flow rate through the discharge
10 outlet from the supply chamber can be varied, depending on
the frequency of release of alumina from the stonage
container. The flow of alumina through the discharge outlet
can be temporarily suspended while the plunger is operated
to make or keep clear the necessary hole in the electrolyte
15 crust.
The pressure of the gas injected into the plenum may
be in the order of five kiloPascal but will vary depending
on the permeability of the gas-pervious wall, the height of
the alumina above the wall, and the grain size of the
20 alumina which is fed into the supply chamber. PreferabL
the gas is injected into the plenum at a pressure and ga
flow rate such as to avoid over-f luidisation of the
alumina. The gas injection into the plenum can be
controlled by any appropriate means, for example
25 solenoid-control on the gas pressure inlet valve can be used
to control both the opening and closing of this valve.
Alternatively, solenoid-control can be used to open the gas
inlet valve against a mechanical pressure applied by a
spring or other means tending to keep the valve closed.

APPENDED 7 PULSE CHUTE PATENT APPLICATION PAGE 12
- 9 -
in order to assist a clearer understanding of the

present invention, reference is now made to the accompanying
drawing which illustrates one preferred embodiment. It is
to be appreciated that this embodiment is given by way of
5 illustration only and that the invention is not to be
limited by it.
The drawing shows, somewhat diagrammatically, the
preferred annular form of alumina supply chamber mounted
concentrically with plunger shaft 1 and crust breaking
10 plunger 2. The supply chamber is defined by inner wall 3,
outer wall 4 and base wall 5. The upper portion ol outer
wall 4 converges inwardly towards inner wall 3 to farm the
entry aperture of the supply chamber through which alumina
from a storage container (not shown) may enter the supply
15 chamber, diverted, if necessary, by inclined wall 6. Hood
member 7 prevents incoming alumina from passing directly
into discharge conduit 8. Depending skirt 9 on hood member
7 further, prevents the normal build up of incoming alumina
from entering the inlet of discharge outlet 8.
20 Gas-pervious wall 10 defines, with the lower portion
of the side walls 3,4 and base wall 5 of the supply chamber,
a plenum 11. Gas injected through pipe 12 into the plenum
in a controlled manner causes fluidisation of the alumina
which has entered the upper part of the supply chamber, thus
25 allowing the fluidised material to flow beneath hood 7,
along the body of discharge conduit 8 and through the outlet
13 of this conduit. Outlet 13 is directed towards the hole
in the electrolyte crust which has been formed by plunger 2.
Gas-pervious wall 10 is- preferably inclined to the
- 10 -
horizontal plane as illustrated. When gas pressure, is

applied through the plenum 11, alumina above wall 10 is
fluidised but heavier materials which may be present as
contaminants in the alumina charge settle to the lower part
5 of the wall 10 and may be cleared periodically.
Accumulation of any contaminants at the lower part of the
gas-pervious wall minimises the- potential for blockage of
the gas flow by these materials and also assists in
preventing the contaminants from being carried into the
10 discharge conduit.
The present invention makes it possible to extend the
time over which a measured charge of alumina is fed into the
electrolyte. Thus a relatively large charge can be fed
into the electrolyte at an adjustable flow rate and in a
15 substantially continuous manner. The use of relatively
large alumina charges reduces the wear on the dispensing
equipment and makes more accurate charge measurement
possible.
- ii -
1. Apparatus for supplying alumina to an alumina

electrolysis tank including an alumina storage container
adapted to release alumina as required for entry into the
electrolyte through a hole formed in the crust on the
5 electrolyte by a crust breaking mechanism, characterised in
that the alumina storage container is connected to an entry
aperture pf an alumina supply chamber having a base wall
from which inner and outer side walls lead towards the upper
portion of the chamber to define the entry aperture; a
10 transverse gas-pervious wall extends between the side walls
of the chamber spaced from the base wall and below the
chamber inlet to define a plenum into which gas may be
introduced through a gas inlet, the gas-pervious wall
allowing gas flow sufficient to fluidise alumina but not
15 allowing alumina to enter the plenum; at least one discharge
conduit extends through the base wall and the transverse
wall to an outlet below the supply chamber, the outlet is
positioned in use above the hole formed in the electrolyte
crust by the plunger and the opposite, inlet, end of the
20 discharge conduit is located above and spaced from the
transverse wall; and a hood member is spaced above the
discharge conduit inlet but extends below the level of said
inlet.
25 2. Apparatus as claimed in claim 1 characterised in that

the supply chamber is substantially annular in form and is
located concentrically with the plunger shaft.
- 12 -
3. Apparatus as claimed in claim 1 or claim 2

characterised in that the gas-pervious transverse wall is
inclined to the horizontal plane.
4. A method for the supply of alumina from a storage

chamber adapted for the intermittent supply of predetermined
amounts of alumina, characterised in that the predetermined
amounts of alumina are fed to a supply chamber, the alumina
in the supply chamber is fluidised by gas injection through
10 a gas-pervious wall below the alumina, and the fluidised
alumina flows from the supply chamber through a discharge
conduit.
5. A method as claimed in claim 4 characterised in that

15 the pressure, the time period of injection, and the time
between successive injections of gas are controlled in
relation to the frequency and amount of predetermined
alumina supply to the supply chamber so as to maintain the
desired continuity 'of alumina flow from the discharge
20 conduit.
6. A method as claimed in claim 4 or claim 5

characterised in that gas is injected into the plenum at a
pressure and gas flow rate such as to avoid
25 over-fluidisation of the alumina.
APPENDIX 8
A3 U.S.A. PATENT
i
Appendix 8 A3 U.S.A. Patent Pagel
The Commissioner of Patents

Q
Oh£~ and Trademarks
IMted Has received an application for a patent

for a new and useful invention. The title
and description of the invention are en-
States closed. The requirements of law have
been complied with, and it has been de-
~£ termined that a patent on the invention
/ of
) shall be granted under the law
ne m
>!AnwriC(\^ ^ »•*•
1 United States Patent
Grants to the person or persons having
title to this patent therightto exclude
others from making, using or selling the
WW " - ••
Fyy:- •.••:• i invention throughout the United States
. ^ . ; - ^ , fi of America for the term of seventeen
^ > , & ' . \ ; f> years from the date of this patent, sub-
i ject to the payment of main tenance fees
as provided by law.
ip^y
M
u
ik l Qimrmssioner of Pattara and Traderntuks
Attest
Appendix 8 A3 U.S.A. Patent Page 2
Mill
USO05423968A
United States Patent m [ll] Patent Number: 5,423,968
Kissane [45] Date of Patent: Jim. 13,1995
[54] A L U M I N A S U P P L Y A P P A R A T U S F O R 4.431.491 2/1984 Bonny « al 204/67

ELECTROLYTIC SMELTER 4.437,964 3/1984 Gerphagnon et al 204/245
[75] Inventor: James P. Klaaane, Portland. 4463.253 1/1986 Hrinzmarm et al _.. 204/245 X
Australia 5.045.168 9/1991 Dalen ei al 204/245
FOREIGN PATENT D O C U M E N T S
[73] Assignee: Portland Smelter Senrices Pry. Ltd.,
92/06230 4/1992 WTPO .
Melbourne, Australia
[21] Appl. No.: 1M&6 OTHER PUBLICATIONS
[22] P C T Filed: Jul. 8,1993 Metallurgy, S U 1560-636-A, p. 23, Apr. 1990.
Metallurgy, S U 985-154-A, p. 18 Dec. 1982.
[86] P C T No.: PCT/AU93/D0332
Primary Examiner—Donald R. Valentine
§ 371 Date: Jim. 21,1994
[57] ABSTRACT
§ 102(e) Date: Jam. 21, 1994
A feeder assembly for an alumina electrolysis cell in-
[87] P C T Pub. No.: WO94/01fi01 cludes an alumina dose holder (5) defined between inner
P C T Pub. Date: Jan. 20,1994 and outer walls (6,7, 7") with an inlet port in the outer
wall above an outlet port in the inner wall. T h e inlet and
[30] Foreign Application Priority Data outlet ports (8, 11) are closable and openable by valve
Jul. 14. 1992 [AU] Australia PC3496 means formed by relative movement between the outer
[51] lot. CL' - C25C 3A4 wall (7, T) and a valve seat (10,12) cooperating with a
[52] U.S. C L 204/245; 204/279 seating edge of the outer wall (7, 7"). T h e valve means
[58] Field of Search 204/245, 279 is moved by drive means (13) including a pneumatically
operated piston (14) movable within a cylinder (15)
[56] References Cited concentric with the shaft (2) of an electrolyte crust-
U.S. PATENT D O C U M E N T S brealcing plunger (1). T h e plunger shaft (2) is axially
3.371,026 2/1968 Kitey et ai 204/245 slidable within an annular sleeve (18) of the piston (1*)
3.901.787 8/1975 Niizcki et al 204/245 which is connected to at least one movable component
4,049,529 9/1977 Golla 204/246 of the valve means.
4.321.115 3/1982 Rebmann et al. 204/67 8 Claims, 2 Drawiag Sheets
4,328,085 5/1982 Friedli et al -.... 204/245
U.S. Patent June 13, 1995 Sheet 1 of 2 5,423,968
FIG1
U.S. Patent June 13, 1995 Sheet 2 of 2 5,423,968
FIG2
Appendix 8 A 3 U.S.A. Patent Page 5
968
i 2
ALUMINA SUPPLY APPARATUS FOR pneumatic mechanism and the control valve, this neces-
sary cushioning action affects addition accuracy.
ELECTROLYTIC SMELTER
In another procedure, two separate pneumatic sys-
This invention relates to apparatus for the controlledtems 5 ate employed, one operating the crust breaking
supply of alumina or other solid materials to an electro- mechanism and the other operating the alumina feeding
lytic cell in which the alumina is converted to alumi- system. In this procedure, it b possible for the mecha-
num nism operating the crust breakmg mechanism to form
In the electrolysis of alumina, solid alumina is dis- the necessary hole in the electrolyte crust and retract
solved in a tank or pot containing molten electrolyte 1° the crust breaker so that the feeding system can then be
such as cryolite and it is desirable to maintain the alu- operated to place a charge of alumina into the hole
mina concentration hi the eJectrolyte within a predeter- formed by the crust breaker. There is less air usage as
mined range. In current practice for the electrolysis of the dosing cylinder of the alumina feeding system is of
alumina, the alumina is fed in successive doses of prede- smaller capacity and m a y operate more often than the
termined size into one or more holes which are made in crust breaker as it is independent of the crust breaker
the electrolyte crust so that the alumina can be admitted mechanism. However, this k n o w n procedure requires
w h e n required. A s the electrolysis of the alumina pro- separate housings for the crust breaking mechanism and
ceeds continuously, it would be desirable if the alumina the alumina feeding system. These separate housings
consumed in the electrolysis process could be continu- not only reduce the space available for operating above
ously replaced so as to maintain the optimum alumina the electrolysis tank but also complicate the construe -
concentration in the electrolyte. However, the optimum i tion of the whole assembly. Hence this design is more
operating conditions are such that the electrolyte crust expensive than the single pneumatic mechanism design.
continuously reforms on the surface of the electrolyte, In our prior Australian patent application no. P K
making it difficult to supply alumina continuously to the 2658/90 (which forms the basis for International Appli-
molten electrolyte beneath the crust. cation PCT/AU91/00169) w e proposed a feeder assem-
For this reason, k n o w n alumina feeding procedures ; bly in which a valve mechanism concentric with the
involve the use of a crust breaker which is operated shift of the crust breaking plunger is operable in re-
mtertnittently to break the electrolyte crust and form a sponse to the initial part only of the downward move-
hole through which the solid alumina can be fed. H o w - ment of the plunger but achieves the advantage of hav-
ever, the action of the crust breaker is necessarily such ing the plunger out of the alumina flow. However, like
that the crust breaking mechanism, such as a pneumati- t the first design, it is affected by cushioning, as the crust
cally operated shaft with an appropriate chisel means breaker cylinder is much larger than necessary for dos-
(hereinafter referred to as a plunger) at its free end, will ing of the alumina. This design is still affected by high
be m o v e d in and out of the hole formed by the plunger. air usage but its accuracy of dosing is less affected by
In one k n o w n feeding procedure, a single pneumatic speed.
mechanism is used to operate the crust breaking mecha- i It is therefore an object of the present invention to
nism, and the discharge of alumina from a storage de- provide an improved alumina feeder assembly which
vice is co-ordinated with the downward movement of enables direct feed of alumina into the hole (as is the
the crust breaker. In this procedure, the alumina charge case with the two separate pneumatic systems design
is thus released w h e n the crust breaker is through the and our Australian patent application P K 2658/90) and
crust so that the alumina is notfreeto enter the hole in > significant reduction in air usage compared to a design
the crust until the crust breaker is retracted. While this using a single pneumatic drive cylinder.
procedure has the advantage of a single pneumatic sys- Accordingly, the present invention provides a feeder
tem, it is obvious that not all the alumina will be able to assembly for an alumina electrolysis tank including a
pass through the hole into the electrolyte immediately crust breaking mechanism operable to break a hole in
w h e n the crust breaker is retracted. > crust formed on the surface of molten electrolyte, the
It will also be apparent that with this form of mecha- crust breaking mechanism including a plunger with a
nism, the plunger must travel through the crust each cutting edge mounted on a reciprocable plunger shaft
time a charge of alumina is to be introduced into the and an alumina storage container adapted to release
electrolyte. This not only involves use of sufficient air alumina as required into a dose holder, characterised m
to drive the plunger but also m a y involve dipping the ) that the dose holder is defined between inner and outer
plunger into the electrolyte with each stroke. It is desir- walls, an inlet port is formed in the outer wall above an
able to reduce the number of times the plunger contacts outlet in the inner wall whereby alumina can flow
the electrolyte as far as possible so that wear of the through the dose holder from inlet port to outlet port
plunger can also be reduced. under the influence of gravity, the inlet and outlet ports
Accurate alumina flow control is m a d e difficult by 5 being cVosable and openabfe by valve means formed by
the required relationship between the plunger move- relative movement between the outer wall of the dose
ment and movement of the flow control valve control- holder and a valve seat which cooperates with a sealing
ling alumina discharge. It will be understood that the edge of the outer wall, the valve means being movable
force available and speed necessary for plunger move- by drive means including a pneumatically operated
ment must be sufficient to achieve crust breakage. This 3 piston movable within a cylinder concentric with the
force, speed and the amount of air needed to produce plunger shaft, the piston having an annular sleeve axi-
them are far greater than that needed to m o v e the alu- ally slidable within the cylinder and the plunger shaft
mina flow control valve and it is therefore necessary to being axially slidable within the sleeve which is con-
cushion the final stages of the pneumatic mechanism nected to an extension sleeve in turn connected to at
against the considerable force available for plunger 5 least one movable component or the valve means.
movement. Because of the interaction between the T h e feeder assembly of the present invention includes
a crust breaking mechanism which is preferably pneu-
matically operated. T h e crust breaking mechanism m-
Appendix 8 A 3 U.S.A. Patent Page 6
5 423 968
3 ' '
eludes a plunger with a cutting edge for breaking the
4
crust, mounted on a reciprocable plunger shaft. savings in air usage compared with a single pneumatic
T h e feeder assembly m a y be associated in use with at mechanism.
least one storage container comprising a hopper or The apparatus of the invention preferably includes a
similar vessel forfinallydivided alumina- Other storage 5 housing which surrounds the plunger shaft and the
containers for other additives to the electrolysis tank. generally concentric feed mechanism components de-
such as aluminium fluoride, calcium fluoride, crushed scribed above. T h e lower end portion of the housing
bath, soda ash. or cryolite m a y be associated with simi- preferably tapers inwardly to facilitate direction of the
lar feeder assemblies. T h e other storage containers m a y alumina falling towards the electrolysis tank after leav-
thus be adapted to feed their contents into the tank in a 10 ing the outlet port of the feeder assembly. T h e lower
similar manner to that described below for the alumina. edge of the housing thus defines the periphery of a
T h e storage container or containers are adapted to release port through which altnnina leaves the housing
feed their contents as required into a dose holder. Pref- as it fails towards the tank
erably the storage container contents are first fed into a Alternatively, the lower end portion of the housing
supply chamber which has an exit port communicating 15 m a y be formed to provide two or more downwardly
with an inlet port of the dose holder. T h e dose holder is tapering outlets directed towards the hole formed in the
defined between inner and outer walls which are prefer- electrolyte crhst by the plunger.
ably concentric with each other and with the plunger T o assist a further utivderstandtrig of the invention.
shaft. T h e inlet port of the dose holder is formed in its reference is n o w made to the accompanying drawings
outer wall. T h e dose holder farther includes an outlet 20 which illustrate one preferred embodiment of the pres-
port located below the inlet port so that alumina or ent Invention, It is to be appreciated that this embodi-
other material can flow through the dose bolder from ment is given by way of illustration only and that the
inlet port to outlet port under the influence of gravity. invention is not to be limited by h.
Both the inlet port and the outlet port are closable BRIEF DESCRIPTION OF DRAWINGS
and openable by valve means formed by relative move- 25 The drawings show, somewhat diagrammatically.
ment between the outer wall of the dose holder and a sectional views of a preferred form of feeder assembly.
valve seat which co-operates with a seating edge of the In F I G . 1, the dose holder is open to the entry of
outer wall. Preferably the inlet port is closed by a mov- alumina.
able upper wall seatable in a fixed upper seat and the FIG. 2 shows the dose holder at the opposite extreme
outlet port is closed by movement cf a movable lower 30 of the valve movement, closed to the entry of alumina
seat into abutment with a fixed lower wall. but open to discharge alumina.
In accordance with the invention, w e provide for DESCRIPTION OF PREFERRED EMBODIMENT
movement of the valve means by a drive means includ- In the drawings, a crust breaking plunger 1 is carried
ing a pneumatically operated piston movable within a by plunger shaft 2. A storage container (not shown)
cylinder concentric with the plunger shaft. This piston 35 feeds alumina or other tank additive as indicated by
includes an annular sleeve axiaily slidable within the arrows A into an annular supply chamber 3. Supply
cylinder and the plunger shaft is axiaily slidable within chamber 3 has an exh port or ports 4 which remain open
the sleeve. T h e piston sleeve is connected to a prefera- allowing the alumina to fall towards annular dose
bly annular extension sleeve which is in turn connected holder 5. Dose holder 5 is defined between an inner wall
to at least one movable component of the valve means. 40 6 and an outer wall 7, T, both of which are concentric
In one preferred construction, the drive means is with plunger shaft 2.
connected to m o v e the upper wail of the dose holder T h e outer wail of the dose holder is formed in two
and the movable lower seal simultaneously. Thus the parts. 7 and T. In the illustrated embodiment, the upper
outlet port is closed as the inlet port is opened. Con- part 7 of the outer wall is movable and in its raised
versely, the inlet port is closed as the outlet port is 45 position provides an inlet port 8 to the dose holder S
opened. between its lower seating edge 9 and valve seat 10. T h e
Alternatively, the construction m a y provide for inde- movement of outer wall part 7 and its association with
pendent operation of the inlet port valve means and of valve seat 10 thus provide a valve means for inlet port
the outlet port valve means. For example, the piston 8.
m a y be connected to m o v e the inner wall and the valve 50 A valve means for outlet port 11 of the dose holder is
seat associated with the outlet port of the dose holder, provided by the relative movement between movable
and the outer wall of the dose holder is connected to the valve seat 12 and thefixedlower wail part T of the dose
cylinder wall which is movable concentrically with the holder outer wall.
piston. T h e movable cylinder and outer wall control the Movement of the respective valve means is con-
operation of the valve means associated with the inlet 55 trolled by drive means 13 which includes a pneumati-
port. cally operated piston 14 movable within cylinder 15.
Whatever particular construction is used, the use of Air supply lines 16 and 17 are used to activate the up-
the plunger shaft concentrically slidable within the ward and d o w n w a r d movement of piston 14 within
piston provides valve means operable independently of cylinder 15. T h e lower end of piston 14 is connected to
the plunger. Accordingly, the valve means can be oper- 60extension sleeve 18 which in turn carries the upper part
ated as often as required to add alumma to the electroly- 7 of the dose holder outer wall and the lower wall 15 of
sis tank while the plunger needs to be operated only the dose holder which includes movable valve seat 12.
w h e n necessary to break the crust and allow access of A s will be appreciated, downward movement of pis-
the alumina to the electrolyte mix. T h e cylinder driving ton 14 from the position shown in FIG. 1 brings the
the valve means needs only to be of considerably 65 sealing edge 9 of part wall 7 into contact with valve seat
smaller stroke than that driving the crust breaker. T h e 10, thus closing dose holder inlet port 8. A t the same
cylinder driving the valve means can also be of smaller time the d o w n w a r d movement of piston 14 separates
piston area. There are thus made possible significant
5,423,968
5 6
movable valve seat 12 from the lower edge 21 of dose a sealing edge of the outer wall, the valve means being
holder part wall T , opening outlet port 11 of dose movable by drive means including a pneumatically
holder S and allowing the alumina within the dose operated piston movable within a cylinder concentric
holder to flow out under the influence of gravity. This with the plunger shaft, the piston having an annular
downwardly flowing alumma ts directed by the in- 5 sleeve axiaily slidable within the cylinder and the
wardly tapermg lower part of hotaing 2C'towards the p h l n g e r shaft ^ MiMy sljdable w i t h b m e d
plunger artd plunger^ft^tnd thus towards the hole ,n w h i c h « connoted to an extension sleeve in turn con-
the crust which has been broken by the plunger ^ ^ to at least O M m o V i W e C Q
A s drive means 13 for the valve means of dose holder means.
5 is operable mder^nderitly of the drive means (not 10 x A fceder m a ^ „ ^ ^ m ^
I^*S. ^g,.,\v iT^lî ^ W ^ m **the inner "** outer wal,s of the dose holder
finer control than that achievable if the plunger move- afe ^ ^ ^ ^ ^ other ^^
ment and dose holder valve means movement are inter- sh^n
dependent and is of smaller area and stroke to the crust •» Afi.«^»,™,.„*~.i»i , „i • ~a • « . i_
w
. , . s. 3. A feeder assembly as claimed m claim 1 character-
breaker and hence saves air usage. 15 - . - ... ,1. • . . _ • , ,. >-i«a«»tw
lset
T» - c _.u J flT^ i_ * w1that 'he inlet port is closed by a movable unper
It is a further advantage of the present invention that „,„„ „„,_ u i„ - = \, . * . , uFt«;i
wa
., , ., s„^5._j s J _. -s - . -. « seatable in afixedupper seat and the outlet port is
the m e c M n i s m can be fitted relatively easily mta exist- . ., _ . » 1.1 , . y.
• s_- u 1 1: J - I J closed by movement of a movable lower seat into abut-
i»g plants which use a single pneumatic drive cylinder, « . „ - „ ft. as J 1 n «»*-«. amu owi,
i , _ . r .. e • - , . - . , - ment with a fixed lower wall.
w h e n replacement of the original drive and alumma . . ,„ .„ _., ,. ,. „ ,
delivery mechanism is considered inxessary or desir- 20 * A ^."T*** " d a u n e d m d a ™ 3 characte_r-
able. This can lead to savings in space and in costs asso- "** *•» * « ' * * ? " ^ B "ff ^ W ^ **
ciated with the structure requiredtosupport the feeder " " " I w f * «**, dose h o W e r ««• ** m o v a b l e l o w M
assembly. Costs can also be reduced by the use of this « s J n " lltan « ous, y-
invention to reduce the number of plunger movements . * A <«der assembly as claimed in clam 3 character-
and plunger wear. T h e addition of alttmina indepen- „ « « . » • - « the drive means ts connected for indepen-
dw
dently of plunger movement also allows the possibility " ™vaaeat <*** respective inlet port valve means
of more frequent ahrmina additions, approaching con- ""* " " j " P ° n va,vc, racanf .
dnuous addition of alumina. T h e abilitytooperate the . 6- A feeder "»«--bt*, as claimed m claim 5 character-
respective valve means of the inlet and outlet of the wed m that the piston is connected to m o v e the inner
dose holder independently further allows greater accu- 30 w a l 1 * nd lhe v a I v e f 3 1 "weiiied with the outlet port
a n d the u t e r
racy of the alumina additions by preventing alumina ° " ^ o fthe ^^ ^lder » connected to the
from entering the dose holder as the dose holder con- cylinder wall which e> movable concentrically with the
tents are released. piston to control the operation of the valve means asso-
I claim: ciated with the inlet port.
1. A feeder assembly for an alumina electrolysis cell 35 7 - A feeder assembly as claim in claim 1 chiiracterised
m that the d o s e
including a crust breaking mechanism operable to break holder is concentric with the plunger
a hole in crust formed on the surface of molten electro- **"&•and a housing surrounds the plunger shaft and
dose
lyte, the crust breaking mechanism including a plunger holder, the lower end portion of the bowing taper-
with a cutting edge mounted on a reciprocable plunger ing inwardly with its lower edge defining the periphery
shaft, and an alumma storage container adapted to re- 40 of a release port though which alumina leaves the hous-
lease alumina as- required into a dose holder, character- rng and falls towards the tank.
ised in that ihe dose holder is defined between inner and 8. A feeder assembly as claimed in claim 1 character-
outer walls, an inlet port is formed in the outer wall ised in that the dose holder is concentric with the
above an outlet in the inner wall whereby alumina can plunger shaft and a housing surrounds the plunger shaft
flow through the dose holder from inlet port to outlet 45 and dose holder, the lower end portion of the housing
port under the influence of gravity, the inlet and outlet 55 being formed to provide two or more downwardly
ports being closable and openable by valve means tapering outlets directed towards the hole formed in the
formed by relative movement between the outer wall of electrolyte crust by the plunger.
the dose holder and a valve seat which cooperates with * * * * *
50
60
Appendix 9 Diary of Changes Page 1
APPENDIX 9
DIARY OF CHANGES
Line 2 Start Up: 15.10.86 - 01.07.87 all bar 6 pots 27.02.88

Line 1 Start Up: 07.03.88 - 23.09.88
Fail Safe dates: Start - Terry 01.03.87
-Parker 01.07.88
Retrofit -Parker01.06.89
Dec 88 All Terry piston rods being replaced by K e m p e rods.
04.05.90 Commenced measuring plunger O D at overhaul.
05.05.90 Commenced tieing up feeders after overhaul for ease of installation.
06.05.90 Initial design A 2 feeders.
28.05.90 Commenced using code on plungers (25E90) and tagging failed feeders.
14.06.90 Huck visit for bolting feeders - unsuccessful.
15.06.90 Commenced discussions for A 2 patent.
25.06.90 Commenced C E T E C R & D on plungers - examine failed plungers and investigate
ceramics.
28.06.90 Visit Parker and commissioned Unisearch to assist in R & D and advise on cushioning
analysis of cylinders.
27.07.90 Commenced R & D on spring design at K e m p e (Geelong).
13.08.90 Commenced using reinforced spools and not using old pipe in spools.
22.08.90 Sourced marinate and Unasco 808 for insulation band and SG-200 insulator for bolts.
Oven tests commenced.
31.08.90 Commenced 3 5 m m steel washers for feeder bolts.
13.09.90 High temperature die steel springs tested O K . Ordered 500.
15.09.90 Commenced using cast plungers with pin area increased (to stop pin being caught on
super chute) and straight sides (to stop dragging on lip halfway up).
25.09.90 Commenced using Bakelaque washers on bolts of supers.
04.10.90 Applied for provisional patent for A 2 patent.
22.10.90 C o m m e n c e d changing piston seals only on failure.
05.11.90 Started using straight plungers.
19.11.90 Tested A 2 in off-line pot - chute hit by anode. Redesign.
20-21.11.90 Qishioiiing test at Parker - high deceleration rate 5 0 G (500m/sec) versus 25G target.
27.11.90 Cushion tests at Parker - confirmatory test show same results on old and new cylinders.
15.12.90 Terry found incorrect piston seal width - changed from 25 to 50 thou.
15.12.90 Started usingflatbottomed cast plungers to make bigger hole.
01.02.91 Started using trial 200 Unasco 808 insulation bands.
04.02.91 A 2 feeders (double chute dribble feed) installed in 4070 and single chute in 3092
( A E D D feeder).
05.02.91 Repair failure correct on history for all parts after this date.
07.02.91 Traced "not feeding" to seized spool 1 4 6 m m O D better than 1 5 2 m m O D .
12.02.91 Installedfirstpot (1011) with trial plungers.
15.02.91 Started changing out 1 5 2 m m O D spools to 1 4 6 m m O D spools.
22.02.91 Silicon carbide plungers removed due to breakage.
05.03.91 3092 and 4070 changed to 50° double chutes (Dribble feed).
07 03 91 4033 changed to 5 inconel 6 5 m m O D plungers.
22.03.91 12 and 1 5 m m nozzle placed in chutes of 4070 and 3092 ( 5 0 % feed rate).
24.03.91 3024 changed to 5 long cast plungers.
25.03.91 Started tests on cylinder cushioning at Kempe.
27.03.91 Trialled belville washers on Parker cylinder to improve cushioning. Not successtui.
03.04.91 2061 #5 chute changed to dual outlets (nozzles installed 10.04.91).
17.04.91 Meeting IXL on new specification of cast iron plungers.

26.04.91 60 cast iron plungers made as Stage 1 of plunger trials - gauge capabilities and location
of hardness point for measurement.
29.04.91 International patent application for A 2 feeder and airslide chute.
01.05.91 Visit ITC Wollongong re angles of chutes for reacted ore.
03.05.91 Nozzles taken off 3092 - blocking.
09.05.91 Started testing Terry cylinders on cushion rig.at workshop
14.05.91 Feeder survey complete - all locations have cylinder number identified -
18.05.91 Rest of nozzles out of 4070.
23.05.91 Started tests on slowing feeders on 8 pots.
01.06.91 Started testing breaking hole at zero speed
03.06.91 Started fracture tests on piston/shaft connection at Unisearch.
12.06.91 Tested new Parker cylinder insulation 3 0 M O H M to 1.5G O H M . N o insulation on old
cylinders
20.06.91 Tested speed of cylinders with video.
22.07.91 Completed piston rod/shaft fracture tests - Terry better than Parker.
25.07.91 Training started on F R S for tradesmen at Kempe.
25.07.91 Started using high temperature springs.
02.08.91 1001-1051 dwell time 2.5 to 2.0 sees.
Sep 91 Stopped spigotted flanges.as not making a difference
01.09.91 Report "Pot Alumina Feeder Design Review" published.
28.09.91 Visit Brenko Thermospray for plunger coatings
02.09.91 Started using studs on Parker broken piston rods.
17.09.91 All pots 2.0 sec dwell time.
17.09.91 Changed Terry rod and piston seals to blue teflon in white cartridge.
18.09.91 2065, 2066, 2067 and 2068 dual dwell time trials.
07.10.91 3026 superstructure modified with holes for better feeder venting.
15.10.91 Reduced grease on seals on overhaul.
16.10.91 20 Terry prototype teflon rod seal in service.
23 10.91 Test insulation on feeder in workshop - 0.4G new and zero old. Insulation breaks down
in service.
28.10.91 At Unisearch (noise reduction for silencers, fatigue tests, orifice design).
Nov,91 Parker rod seals changed from 530 (Parker) to 548 (Terry). Terry changed from 432 to
442.
12.11.91 Line 1 n o w slow speed.
12.11.91 Started Terry F E C in Parker.
13.11.91 Started 103E slow speed 3.0 sec dwell time.
19.11.91 Future Parker piston rods 4140 5 thou (60 micron) chrome from 10L45, 2 thou (25
micron).
19.11.91 Purchased cast double outlet chutes.
22.11.91 Purchased 12 new feeder racks with safer access.
25.11.91 All pots (both lines) at slow speed.
02.12.91 Line 1 4 sec to 3 sec dwell time and Line 2 @ 4 sec dwell time.
05.12.91 N e w pot 1075 is A 2 .
28.01.92 First Portland Aluminium spec, plungers in service - still using H R for 1 month.
Feb 92 Changed to blue F E C casings and blue bushes for Terry.
01.02.92 Started using 4140 Parker piston rods.
02.02.92 Started using bronze inserts instead of hardened inserts.
26.02.92 R & D Parker teflon rod seals arrived.
02.03.92 3049 cast changed to A 2 feeders and IXL cast chutes.
27.03.92 Cylinder trial start at Parker.
01.04.92 Report "Optimising Pot Feeders" published.
08.04.92 At Rockdale and Wenatchee to give presentation on feeders to Feed Alumina NetworK.
27.05.92 Insert on all feeders ( 8 1 m m stroke) ex workshop.
08 06.92 1064 A 2 changed to A E D D - dags causing anode effects.
12.06.92 3095 1,3,5 A 2 feeders changed to A E D D - dags causing anode effects.
24.06.92 At Parker for autopsies on cylinder trial.
Page 3
01.07.92 Parker piston back up washers no longer used.

28.07.92 Trial single outlet chute 2009 # 1.
07.08.92 Changed dual dwell time short circle from 1.7 to 1.9 sec (100 pots).
17.08.92 Started routine shotsize testing of n e w pot feeders.
02.09.92 Started 5 3 m m stroke spool inserts on all feeders ex workshop.
03.09.92 Started stroke testing of feeders in pots - pressures and stroke times.
30.09.92 A2's changed to 2 sec short and 4.5 sec long every 3rd cycle.
Sep 92 Started flange insulation and boomerangs.
06.10.92 20 Norton seals arrived.
21.10.92 2009 chutes changed to single outlet.
25.10.92 20 Dover seals arrived.
N o v 92 Changed A 2 pots to single outlet chutes.
21.12.92 All pots 4.5/2.2/2 dwell time.
24.12.92 Changed or cut off #3 feeder plus others on all A 2 pots as dag getting caught and
chutes - excess anode effects.
29.12.92 5.0/2.0/2 dwell time.
30.12.92 3095 changed from A 2 to A E D D - dags.
06.01.93 Started installing Parker piston back up washers again.
06.01.93 High feeder changeouts due to pointed feeders - 20 feeders had only plungers changed
(no cylinder overhaul).
15.01.93 Started quick overhauls based on air leaks.
18.01.93 All pots dwell 4.5/2.0/2 -> 4.5/2.2/2.
02.02.93 100 trial rubber shirts on assemblies to prevent ore leaks.
16.02.93 Accepting air leaks of under 30 1/min
16.02.93 Don't overhaul cylinders if plunger worn only.
18.02.93 Dual dwell time back on all pots.
22.02.93 Trial mobile plunger coolers operating.
22.02.93 Started trial of seals/plungers/bushes on 9 n e w pots.
26.02.93 Increased grease.
26.02.93 First Bimetal plungers installed.
01.03.93 Report "Point Feeding Literature Survey" published.
05.03.93 4011 changed to double outlet chutes - not dribble feed.
05.03.93 Purchased 40 air leak testers.to measure air leaks in pots.
14.03.93 Started work on plunger penetration investigation.
19.03.93 Started 8 day trial of plungers in liquid bath.
19.03.93 Started using studs for broken piston rods again - 360 days extra life.
27.03.93 Started first 7 pots using plunger coolers to stop dagging on start up.
29.03.93 Traced 24 broken plungers - 7 0 m m shorter and 1 0 0 % better wear rate than feeders in
same pot.
29.03.93 C o m m e n c e d national phase P C T patent applications for A 2 feeders.
01.04.93 N o more overhauled feeders installed on n e w pots.
06.04.93 10 feeders installed with 5 0 m m shorter plunger.
08.04.93 Purchased 50 A E D D and 35 A 2 feeder chutes.
10.04.93 Started cone lock nuts on boomerang brackets.
20.04.93 Visit from Huck re better feeder bolting - rej ected.
22.04.93 First superstructure modified with feeders not removed by operators (1066).
23.04.93 Purchased 35 non magnetic slings.
26.04.93 Alternating plungers (PA, H R , Incoloy, 310SS, 2 1 % C R ,
2 5 % C R , 2 7 % C R , Ni-resist) for all overhauls until used.
28.04.93 Installed short plungers on all feeders in 1031 and 4023.
30.04.93 Engaged C E T E C , Shell and Optimol for grease investigation.
05.05.93 Started al. bronze for hardened inserts.
07.05.93 All pots 5 sec dwell time after set for 2 hours.
10.05.93 Started stepped washer welded to bolt on top flange.
10.05.93 Started installing swivelfittings in centre port.
12.05.93 Started changeout of A 2 chutes to double oulet (2069).
26.05.93 Started installing short plungers in another 7 pots.
Page 4
02.06.93 Started installing double outlet chutes on 10 AEDD pots
03.06.93 7 pots with short plungers completed.
10.06.93 1096 installed with split air header.
28.06.93 All west half of pot lines covering shift after set - 8 0 % reduction in blocked feeders.
July 92 Air leak and plunger survey completed.
01.07.93 C o m m e n c e d feeder removal for inserts and west end ownership change. High turnover
of feeders started.
01.07.93 Commissioned and trained operators on F R S .
01.07.93 All plungers n o w 5 0 m m shorter ex K e m p e .
01.07.93 All magnalube finished. N o w using T R A N Z 414.
07.07.93 Patent applicationsfiledfor A 3 feeders (PCT plus 7 others).
12.07.93 All A 2 n o w converted to double outlet chutes.
13.07.93 A 2 feeders changed 4.0/2.2/3 to 4.5/2.2/3 dwell time.
15.07.93 Purchased 2600 new feeder ID tags and 2 extra Terry hollow cylinders for A3.
15.07.93 3002-16 even numbers changed to break every 3rd cycle (not 2nd cycle).
15.07.93 First batch of 310SS cast production model ex DCL.
16.07.93 Stopped reusing feeders in n e w pots.
19.07.93 All slings changed to non magnetic.
21.07.93 2064 converted to A 2 feeders.
21.07.93 Finished installing double outlet chutes on 10 A E D D pots.
22.07.93 Plunger shaft acceptance changed from 4 7 m m to 4 5 m m .
26.07.93 Increased grease quantity to about lOOg/overhaul.
26.07.93 Introduced "alumina in assembly" fault.
28.07.93 Engaged Unisearch to fatigue test A 3 .
August,93 Trimming guide to prevent partial vacuum in spring chamber.
03.08.93 Started cutting guides for less suction in spring chamber.
04.08.93 Restrictors out of 2018-34 odd numbered pots.
09.08.93 103 odd 4/2.7/2; even 5/2.2/2 dwell time trials.
13.08.93 2073#3 first A 3 installed (with kidney plate turned off) -Ortman.
13.08.93 All slings new design.
13.08.93 102 all 4.5/2.4/2
18.08.93 First A 3 removed due to incorrect air set up - 8 second dwell and large dags. Crust
break rod seal failed and minor air leak on Ortman feed cylinder.
23.08.93 103 all 4.5/2.4/2 dwell time.
27.08.93 101, 103, 104 all on 4.5/2.4/2.
Sept,93 T A 2 R 2 , G P S O , A 2 M grease trials started (alternating feeders).
Sept,93 Started drilling holes in assembly to prevent partial vacuum in spring chamber.
08.09.93 Started drilling holes in assemblies to depressurise spring chamber and to assist
cleaning.
14.09.93 40 stainless spanners for feeder changeouts arrived.
20.09.93 Ordered Kennett ladder for feeder changeouts.
18.10.93 2073 #3 A 3 feeder removed for inspection and replaced with Terry feed cylinder.
Crust break seals, bad leak but minor on feed cylinder. Bushes show only pitting from
gas attack - 60,000 cycles equivalent to 80 days operation.
28.10.93 20 Atlas Copco cylinders arrived.
28.10.93 4011 (A2 pot) taken off line.
Nov,93 Parker rod seals changed from 548 to 548s (short reversed with teflon wiper). Terry
changed from 442 to 4 3 0 Y to same design.
08.11.93 Changed to wider walled F E C .
10.11.93 1094 on-line as split header.
12.11.93 1022 on split header.
01.12.93 2070 on A 3 indep feed.
07.12.93 2070 on end stroke sensing.
10.12.93 Started installing Nicrofer plungers.
30.12.93 # 1 and 3 removed from 2070 - leak and siezed.
01.01.94 Normal bolts on boomerangs.
25.01.94 Trial Kennet ladder arrived. Works but heavy to move.
01.02.94 2046 #1 and 2047 #3 changed to 1 5 5 m m long plungers ( 5 0 m m short). N o problems

after 6 weeks operation before pot cut out.
08.02.94 Inspection of Ortman removed ex 2070 #3 on 31.12.93 - lock lite 567 siezed piston rod.
08.02.94 2070 #1 Terry out - ore leak due to non-viton 'O' ring on FEC.
08.02.94 2073 end stroke sensing installed.
11.02.94 2070 #1 seized (Ortman) no fault found at room temperature. Non-conforming K e m p e
put in.
11.02.94 2070 charged to feed every time.
15.02.94 2070 #1 too slow, orifice removed.
21.02.94 2070 #2 not feeding - anode effects.
28.02.94 2074 end stroke sensing commissioned.
01.03.94 Shot size testing on 2070 showed 2.5s dwell time caused variable shot size. Changed to
4.0s. Anode effects dropped from 2/day to 0.1/day.
07.03.94 2075 sequential feed installed but not commissioned.
18.03.94 Reinstalled all restrictors. All pots running at same speed. Result of trials from 8/93 to
3/94 of alternating sections of restrictors in and out. Statistical evidence of less anode
effects at slower speed.
June,94 Using T A 2 R 7 grease.
APPENDIX 10
CONFIDENTIAL TO ALCOA ***
PLANT CODES
INTEGRATED SODERBERG
T1 Portland (Alcoa) S1 Lista (Elkem)

T2 British Alcan (Alcan)
T3 Mosjoen (Elkem/Alcoa)
T4 Le Terriere (Alcan)
T5 Grande Baie (Alcan)
T6 Pt Henry (Alcoa)
T7 Sao Luis (Alcoa) GUILLOTINE
T8 Boyne (Comalco)
T9 Wenatchee (Alcoa) G1 Bell Bay (Comalco)
T10 Kurri Kurri (Alcan) D11 N2AS (Comalco)
T11 Massena (Alcoa)
T12 Mead (Kaiser)
T13 Badin (Alcoa)
T14 Seebree (Alcan)
T15 Rockdale (Alcoa)
T16 Warrick (Alcoa)
T17 Mt Holly (Alumax)
INDEPENDENT
D1 ISAL (Alusuisse)
D2 Karmoy (Hydro Aluminium/Pechiney)
D3 Granges (Reynolds)
D4 Alouette (Pechiney)
D5 Dunkirk (Pechiney)
D6 Beconcour (Pechiney)
D7 Aluar
D8 Lauralco (Pechiney)
D9 Baie Comeau (Pechiney)
D10 Tomago (Pechiney)
D11 Ti Wai (NZAS)
D12 Tennessee (Alcoa)
D13 Ardal (Hydro Aluminium)
D14 St John de Maurienne (Pechiney)
D15 Lockerby (Pechiney)
D16 Alba (Pechiney)
D17 VAW
D18 Venalum (Hydro Aluminium)
D19 Kidricevo (Pechiney)
D20 Sundalsora (Hydro Aluminium)
D21 Zia (Pechiney)
D22 L M G (Alusuisse)
D23 Steg (Alusuisse)
D24 Inalco (Pechiney)
D25 Alusaf (Pechiney)
APPENDIX 10
*** CONFIDENTIAL TO ALCOA ***
PLANT CODES
INTEGRATED SODERBERG
T1 Portland (Alcoa) S1 Lista (Eikem)
T2 British Alcan (Alcan)
T3 Mosjoen (Elkem/Alcoa)
T4 Le Terriere (Alcan)
T5 Grande Baie (Alcan)
T6 Pt Henry (Alcoa)
T7 Sao Luis (Alcoa) B A R B R E A K (GUILLOTINE.
T8 Boyne (Comalco)
T9 Wenatchee (Alcoa) G1 Bell Bay (Comalco)
T10 Kurri Kurri (Alcan) D11 N Z A S (Comalco)
T11 Massena (Alcoa)
T12 Mead (Kaiser)
T13 Badin (Alcoa)
T14 Seebree (Alcan)
T15 Rockdale (Alcoa)
T16 Warrick (Alcoa)
T17 Mt Holly (Alumax)
INDEPENDENT
D1 ISAL (Alusuisse)
D2 Karmoy (Hydro Aluminium/Pechiney)
D3 Granges (Reynolds)
D4 Alouette (Pechiney)
D5 Dunkirk (Pechiney)
D6 Beconcour (Pechiney)
D7 Aluar
D8 Lauralco (Pechiney)
D9 Baie Comeau (Pechiney)
D10 Tomago (Pechiney)
D11 TiWai(NZAS)
D12 Tennessee (Alcoa)
D13 Ardal (Hydro Aluminium)
D14 St John de Maurienne (Pechiney)
D15 Lockerby (Pechiney)
D16 Alba (Pechiney)
D17 VAW
D18 Venalum (Hydro Aluminium)
D19 Kidricevo (Pechiney)
D20 Sundalsora (Hydro Aluminium)
D21 Zia (Pechiney)
D22 L M G (Alusuisse)
D23 Steg (Alusuisse)
D24 Inalco (Pechiney)
D25 Alusaf (Pechiney)

Optimising Alumina Feeders in Aluminium Smelting Pots

Uploaded by

Copyright:

Available Formats

Optimising Alumina Feeders in Aluminium Smelting Pots

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

Optimising Alumina Feeders in Aluminium Smelting Pots

Uploaded by

Copyright:

Available Formats

University of Wollongong

Optimising alumina feeders in aluminium smelting

Research Online is the open access institutional repository for the

A THESIS SUBMITTED IN PARTIAL FULFILMENT OF TH

J.P.KISSANE, B.E., MTEAUST, CPENG

DEPARTMENT OF MECHANICAL ENGINEERING

for a degree to any other university or institution.

Crustbreak/feeding systems in prebake aluminium smelting pots are critical to efficient

testament to his patience.

improvement, the hard work in the plant is what makes results.

That's a lot of people, but 1 had a lot of help. Thanks.

NOTE TO THE READER

T integrated feeder plant

Appendix 10 lists the plant names against the codes.

Chapter 1 Introduction and Literature Survey 1

CHAPTER 1 INTRODUCTION AND LITERATURE SURVEY

1.1 W H Y ALUMINA FEEDER TECHNOLOGY is IMPORTANT 1

1.2 LITERATURE SURVEY 7

1.2.1 Aluminium Smelting Technology. 7

1.2.2 Objectives ofAlumina Feed Control 13

1.2.3 Initial Prebake Feeding Systems 17

1.2.4 Integrated Feeders 19

1.2.6 Comparison ofIntegrated and Independent Feeder Designs 27

1.2.8 Continuous Feeders 33

1.2.9 Aeration Feeders 34

1.2.11 Pot Alumina Supply Systems 38

1.2.14 Alumina Dose Design and Accuracy 44

1.2.16 Pneumatic Systems 50

1.2.17 Other Feeder Developments 52

1.3 DEVELOPMENT OF THIS RESEARCH 54

1.3.1 Research Objectives 54

1.3.2 Optimizing Existing Feeder Designs 54

1.3.3 New Feeder Designs 56

1.3.5 Document Layout 59

CHAPTER 2 SHOT SIZE ANALYSIS

2.3.1 Dwell Time 65

2.3.3 Stroke Time 69

2.3.4 Validity of Off-Site Testing.

2.4.2 T6 Integrated Feeder 78

2.4.3 Tl 1 Integrated Feeder 80

2.4.4 Portland A2 Feeder 80

2.4.5 A3 Feeder and Sequential Feed. 84

2.4.6 Decision on Use ofSpool Inserts 85

2.5.1 Bulk Density of Alumina 87

2.5.2 Flow Through Inaccuracies in Dosing Systems 8

2.5.3 Effect of Springs 89

2.5.4 Effects of Cylinder Design on Shotsize 9

2.5.6 Effect of Height ofAlumina in the Hopper 9

2.5.8 Effect ofSpool Manufacture Quality. 100

2.6 O N - S I T E T E S T I N G O F S H O T SIZE 102

2.6.1 Til-ColdPots 102

2.6.2 Portland - Cold Pots. 102

2.6.3 Portland-Operating Pots 103

2.6.4 Portland - A3 Feeders 108

2.7 EFFECTS OF PNEUMATICS ON SHOT SIZE 109

2.8 DIRECT FEED INTO THE HOLE Ill

2.10 MAJOR FINDINGS FROM SHOT SIZE ANALYSIS 117

2.11 RECOMMENDATIONS FROM SHOT SIZE ANALYSIS 118

CHAPTER 3 DOSE DELIVERY

3.1 SUMMARY 119