12th International Mineral

Processing Conference

editor
Claudia Velásquez
12th International Mineral
Processing Conference

editor
Claudia Velásquez

oct0ber 26-28
Santiago, Chile
Copyright
© Gecamin, Chile. All rights reserved.
No part of this publication may be reproduced,
stored or transmitted in any form or by any
means; electronic, mechanical, by photocopying,
recording or otherwise, without the prior
written permission from Gecamin.

Cover and book design by Estudio Vicencio.

Printed and bound in Santiago, Chile.

Author’s disclaimer
Any views and opinions presented in the articles
published in these proceedings are solely those of
the authors and do not necessarily represent those
of Gecamin. The authors take full and exclusive
responsibility for technical content, style, language
and accuracy of the information published herein.
This information is not intended nor implied to
be a substitute for professional advice. The editors
are not responsible for any damage to property
or persons that may occur as a result of use of
the information contained in this volume.

I.S.B.N. 978-956-9393-64-8

Gecamin
Paseo Bulnes 197, Piso 6
Santiago, Chile
Postcode: 833 0336
Telephone: +56 2 2652 1500
www.gecamin.com
www.gecamin.com
CONTENTS

Organizers
Committees
Foreword
Preface
Acknowledgments
proceedings sponsor

PLENARies
The Next Big Thing
Peter Amelunxen

Mine-To-Plant Strategies within the Cyclic Copper Business

Jorge Menacho

chapter 1
Comminution: Crushing, Grinding and HPGR

Upgrade of Hydraulic Adjustment System of Symons 7’ Crusher

Sebastián Villalobos, Nelson Mella, Alejandra Pereda, Iván Araya, María Luan and Alexis Guerra

More than an Alternative? Shell supported Ball Mills in Mega Comminution Plants
Egbert Burchardt , Christian Nieland and Alexander Baumbach

Mitigating the Impact of Sea Water on Copper Flotation Response through Optimizing
Grinding Chemistry
Jessica Kinal, Simon Jacques and Christopher Greet

A Public Database of Tumbling Mill Grindability Measurements and their Relationships

Alex Doll

Estimation of Load Level in Industrial SAG Mills Using the Onmill System
Hernán Robles, Tania Ledezma, Luis Vinnett and Jorge Pontt

The Development of SGI Test at El Soldado

Miguel Becerra and Francisca Jorquera

Single Stage SAG Circuit at El Soldado

Miguel Becerra and Francisco Vicuña

Advanced Crusher Control Strategy Using Model Based Predictive Controller

Tomás Carricajo, Juan Morales, Freddy Gómez, César Garrido and Marcelo Jara

Energy and Throughput Optimization for Fine Grinding in Stirred Media Mills
Ilesh Shah

Debottlenecking of Secondary Grinding Circuit: An Hydraulic Analyze over Process Results

Claudio Muñoz

The Effect of Key Operating Conditions on Grinding Media Consumption in Semiautogenous Mills
Levi Guzmán and Jaime Sepúlveda
The Role of HPGR in Dry Minerals Processing Circuits
Frank van der Meer and Javier Valdeavellano

Characterization of Particulate Matter Emitted during the Primary Crushing Simulation and their
Relationship with Geological Units
Javiera Gerding and Óscar Jerez

Mine-to-Plant t Commissioning in Salobo Mine

Jair Alarcón, Ronaldo Fonseca, Grazielle Gontijo and Stephen Jeric

chapter 2
Flotation: Fundamentals and Practical Applications

Adsorption and Flotation Response of the Rhodococcus Erythropolis Bacteria to Hematite

Carlos Castañeda, Antonio Merma, Jhonatan Soto and Mauricio Torem

Bioflotation of Dolomite and Calcite Using Rhodococcus Opacus Strain

Vanessa Coelho, Ronald Rojas, Antonio Merma and Mauricio Torem

Investigation of Uncertainties as an Indicator for Process Evaluation in the Recovery Stages of the
Flotation Circuits
Marcelo Montenegro, Yousef Ghorbani, Glenn Araya and Manuel Leiva

Typical Reproducibility of Metal Balances in Flotation Plants

Luc Lachance and Donald Leroux

Fundamental Aspects of Hematite Fines Electroflotation Using the Bacterial Strain Rhodococcus Opacus
as Bioreagent
Ronald Rojas, Mauricio Torem, Vanessa Coelho and Antonio Merma

Continuos, Real Time Pulp Chemistry Measurements and what They Tell Us about
Metallurgical Performance
Christopher Greet and Kyle Selga

Improving Image Analysis of Online Bubble Size Measurements with Enhanced Algorithms
Claudio Acuña, Luis Vinnett and Seng How Kuan

Enhancing the Performance of Cu-Mo Circuits Using the Jameson Cell

Virginia Lawson

Determination of Collector Affinity for Flotation of Sulfide Minerals

Jocelyn Quinteros, Robinson Constanzo, Gabriela Gioia, Eduardo Enríquez and Nefer Valenzuela

Slag Flotation at El Soldado

Miguel Becerra and Jaime Zúñiga

Advances in Flotation Chemistry: Xanthate Alternative Products

Nefer Valenzuela, Juan Román and Jacques Collin

Humic Substances for the Flotation of Sulfide Minerals: Chronology and Recent Developments
César Sáez and Rosanna Ginocchio
Interactions between Kaolinite and Molybdenite in Fresh and Seawater
Gastón Avilés, Andrés Ramírez, Leopoldo Gutiérrez

Use of Blending to Optimize the Flotation Process of High Clay Ores

Jaime Álvarez, Leopoldo Gutiérrez, Sonia Helle, Úrsula Kelm, Marcos Pincheira and Óscar Jerez

Program of Continous Standard Tests in Mini-Pilot Flotation Plant

Pablo León, Mario Basaure and Juan Reyes

Hematite Flotation Using a Crude Biosurfactant Extracted from a Gram Positive Bacteria
Jhonatan Puelles, Antonio Merma and Mauricio Torem

Improved Flotation Performance through Modernization with Outotec Forced Air Technology
Chris Cruickshank, Koen Aberkrom, Miika Tirkkonen and Alejandro Yáñez

In Situ Gas Holdup Sensor for Industrial Flotation Machines

Miguel Maldonado, Pablo Olivares and César Gómez

Use of Depressant BK-511 in Selective Flotation of Molibdenite, Andina Division, Codelco

Mauricio Bustamante and Héctor Tobar

Modeling H2S Gas Generation in Copper/Molybdenum Flotation Cells

Humberto Estay, Karina González and Rodrigo Kong

On the Interaction Effect between Xanthate Collector, Frothers and Solids: A Laboratory
Characterization Of Copper Sulfide Flotation
Álvaro Videla, Ignacio Molina and Paulo Martins

Inferring Gangue Entrainment from Conductivity Measurements at the Top of the Froth
Rodrigo Araya, Dalton Major, Brigitte Lacouture, César Gómez and Kristian Waters

chapter 3
Concentration Plant Management and Operation

A Process Mineralogy Approach to Gravity Concentration of Tantalum Bearing Minerals

Yousef Ghorbani, Rob Fitzpatrick, Melanie Kinchington and Gavyn Rollinson

Metal Accounting: The Core Responsibility of Process Engineers

Manuel González, Stephane Brochot, Lorenzo Cappai and Pierre Vix

A Review of the Options in Concentrator Layout

Greg Lane, Bianca Foggiatto, Phil Dakin and Andrew Tew

Improving Advanced Control for Better Mineral Throughput, Size Classification and Cu Recovery
Daniel Silva, Bill Gough, Tomás Carricajo and Andrés Chacc

Measuring and Utilizing Ore Quality Variability Using Real Time On-Belt Elemental Analysis to
Improve Mill Feed Quality
Henry Kurth and Óscar Vásquez
chapter 4
Water Resources Optimization and Tailings Management

Effect of Seawater on Solid-Liquid Separation and Transport Properties of Magnetite Tailings

Matías Álvarez, Christian Ihle, Willy Kracht and Cristian Reyes

Use of Seawater in Copper Mines in Chile: Challenges and Opportunities

Macarena Terroba, Rubén Muñoz, Pablo Mardones and José Adriasola

chapter 5
Lab-Scale Metallurgical Tests and Pilot Plant for
Process Control

On the Capacity to Predict Metallurgical Indexes from Mineralogy, Particle Size and Operating
Conditions at Batch Scale
Felipe Contreras, Juan Yianatos, Luis Bergh, Luis Vinnett and Isabel Panire

A Comparison from Pilot Test to Industrial Result: Implementation of Pneumatic Technology in

Molybdenum Selective Flotation Plant
Patricia Véliz, Claudio Muñoz, Egidio Catalán and Heinrich Sprenger

Predicting the Energy Requirements of High Pressure Grinding Rolls Using Piston Press
Test Procedures
Zorigtkhuu Davaanyam, Bern Klein, Stefan Nadolski and Amit Kumar

Jameson Cell Project Evaluation in the Cleaner Circuit at Codelco Andina

Eduardo Morín and Virginia Lawson

Characterization of Laboratory Crushing Tests with Chuquicamata Underground Mine Samples

Bruno Merello, Álvaro Videla, Edisson Pizarro, and Jaime Sepúlveda

Bubble Surface Area Flux Measurement in Laboratory Flotation Cells: Two- and Three-Phase Operation
César Gómez and Jaime Álvarez

chapter 6
Modeling, Simulation and Design of Mineral Processes

How to Avoid Errors in HPGR Product Characterization

Felix Heinicke, Harald Günter and Andre Kamptner

Simulation of the Flotation Process in a Hybrid Cell with the OpenFOAM CFD Framework
Gerhard Holzinger, Stefan Pirker and Lukas Petzold

Comparative Analysis of Grinding Efficiency and Classifications Curves using JKSimmet

Miguel Becerra

Reduction and Fracture Analysis of a Tungsten Ore and its Use for Fundamental Liberation Modeling
Lorena Guldris, Magnus Bengtsson and Magnus Evertsson

Geometallurgical Models, their Interaction with Plant Models and the Importance of Final Reconciliation
Francisco Soto, Javier Pizarro, Romke Kuyvenhoven and Orlando Rojas
Analysis of Autoregressive Model with Exogenous Variables to Predict the C.E.E. SAG Mill as a
Geometallurgical Answer of Porphyry Copper Deposits
Nicolás Miranda and Antonio Candia

Process Development for the Mina Justa Project

Adam Johnston, Ian Ritchie and Daniel Yataco

chapter 7
New Processing Technologies

Current Developments in Preconcentration Using Dense Media Separation

Mauritz Lundt and Ish Grewal

Beneficiation of Industrial Minerals Using a Tribo-Electric Belt Separator

Frank Hrach, Kyle Flynn and Paul Miranda

Multiple Concentrator Survey of Real-Time Individual Cyclone Particle Size Tracking

Christian O’Keefe, Robert Maron, Dylan Cirulis and Juan Medina

Smart Wireless Solutions: Process Control and Variables Monitoring in a Leach Pad Operation
Hugo Barrientos

editor
ORGANIZERS

The Procemin 2016 Conference was organized by Universidad de Concepción, McGill University
and Gecamin, Chile.

Universidad de Concepción, The Department of Metallurgical Engineering

Universidad de Concepción, from 1961 sealed his commitment to mining and the country,
when it began the career of Civil Metallurgical Engineering, until today, more than 800
engineers have graduated from our classrooms to join the productive world. Currently, about
35 professionals graduate each year after an 11-semester training, where the emphasis is on
extractive metallurgy, human resources and economic training; English in a framework of
security and the environment. In 2013, and in order to reinforce our commitment to mining,
the career of Civil Engineering of Mines was created, with an offer of 70 seats and a strong
demand, a career that has had an excellent response positioning ourselves as the second
highest scores nationwide, it also contemplates an 11-semester training and with the same
features and emphasis previously mentioned.
The department also teaches graduate Master and PhD programs in Metallurgical
Engineering, enabling our best undergraduate students can complement their training earlier,
whether on Metallurgy or Mine engineering.
Since its beginnings, the Department has maintained a close relationship with the
productive world through technical assistance programs, memories plan, internships, study
tours and scholarships, which promotes continuous feedback from the industry requirements.
Today we are the national counterpart to the International Centre of Excellence (CEI-Chile-
CORFO) who heads the University of Queesland-Australia.
Learn more by visiting www.eng.metalurgia.udec.cl

McGill University, The Department of Mining and Materials Engineering

Founded in 1821, McGill has carved out a spot among the world's greatest universities growing
to 12 Nobel Prize winners, 40,000 students, 250,000 living alumni, and a reputation for
excellence that reaches around the globe. McGill is the most international of North American
universities located in the heart of Canada’s most culturally diverse city – Montreal.
McGill University has been home to leading research in mining and materials engineering
for 135 years. The Department of Mining and Materials Engineering boasts one of McGill’s
most vibrant research programs, with highly-respected professors and students from around
the globe. Materials Engineering research draws upon many fields to build working prototypes
and processes in mineral/metal processing and advanced materials (aerospace coatings,
biomaterials, nanomaterials, and electronic/energy materials). Mining engineering research
strengths include mine design, planning, and optimization, orebody modeling, stochastic
models and risk analysis in mine valuation, rock mechanics, and numerical modelling.
Mineral processing is of paramount importance to the way we live our lives. As the world’s
economies grow, the demand for metals and other mined products increases. However, while
demand is increasing, the amount of mineral in ores is decreasing placing greater emphasis
on efficient mineral extraction and recovery. The role of mineral processing is to utilize
current knowledge to improve recovery processes, and to develop new processes which
meet strict environmental standards, with specific research interests including:
• Air dispersion measurements in flotation machines
• Novel solvent extraction methods
• Frother roles and analysis
• Froth and foam characterization
• Flotation cell and circuit optimization
• Physicochemical properties of minerals
• Gravity separation
• Self-heating characteristics of sulfide minerals
• Particle tracking in mineral processing
Learn more by visiting www.mcgill.ca/minmat

Gecamin

Powering professional development for sustainable mining

Gecamin is a company with 18 years of experience organizing technical and international

conferences for the mining industry. Our conferences aim to inform and inspire
professionals from all over the world, fostering the exchange of best practices and
innovative experiences.
Over 23,000 professionals have attended our events and have been trained in areas
fundamental to the mining industry. These areas include Geology and Mining, Mineral
Processing, Hydrometallurgy, Sustainability and Environment, Water and Energy,
Maintenance and Automation, and Human Capital.
Gecamin seeks to contribute to the sustainable development of the mining industry
by openly addressing its most pressing concerns and by offering a platform for knowledge
exchange that aims at identifying the most sustainable solutions.
In 2015, Gecamin organized 9 conferences and 9 courses, with a total of 696 technical
presentations, gathering 2,416 delegates from 57 countries. Additionally, there were a total
of 623 representatives of mining sites , and a book with the proceedings was published
for each conference.
Learn more about Gecamin conferences by visiting
www.gecamin.com
COMMITTEES

Organizer
Minerals, Chile • Xavier Boreaux, Magotteaux Andino, Chile •
Executive Committee
Michel Brissette, Freeport-McMoRan Mining Company, USA •
Guillermo Coloma, Sibelco, Chile • Trevor Hines, UK Trade
Chair
and Investment, Chile • Carlos Jara, Cía. Minera Doña Inés
Jaime Sepúlveda, PhD, Executive Director,
de Collahuasi, Chile • Miguel Jiménez, Mineral Processing
J-Consultores, Chile
Consultant, Chile • Romke Kuyvenhoven, Empírica

Past-President Consultores, Chile • Suzanne Lynch-Watson, Multotec,

Australia • Ekkhart Matthies, Weir Minerals, The Netherlands •
Dalibor Dragicevic, Vice-President Processes, Compañía
Iván Menacho, Minera Antamina, Peru • Alfonso Muñoz,
Minera Doña Inés de Collahuasi, Chile
Sierra Gorda SCM, Chile • Claudio Muñoz, Cía. Minera Doña
Co-Organizer Inés de Collahuasi, Chile • Rodrigo Muranda, Moly-Cop,
Froilán Vergara, Director of Department of Metallurgical Chile • Francisco Reyes, FLSmidth, Chile • René Rojas, Teck,
Engineering, Universidad de Concepción, Chile Chile • André Carlos Silva, Universidade Federal de Goiás,
Brazil • Jorge Sumalavia, Radomiro Tomic Division, Codelco,
Executive Director
Chile • Gustavo Tapia, Antofagasta Minerals, Chile • Jannie
Carlos Barahona, General Manager, Gecamin, Chile
van Deventer, Zeobond Group, Australia • Juan Yianatos,

Technical Coordinator Universidad Técnica Federico Santa María, Chile.

Claudia Velásquez, Extractive Metallurgy Consultant, Chile

Technical Committee
Conference Coordinator
Alejandro Astudillo, Ministro Hales Division, Codelco,
Oliver Allendes, Gecamin, Chile
Chile • Felipe Bello, Minera Centinela, Chile • Fernando
Betancourt, Universidad de Concepción, Chile • Alexis
Directing Members Campos, Cytec Solvay Group, Chile • Juan Collao, Pucobre,
César Gómez, Academic, McGill University, Canada • Leopoldo Chile • Fernando Concha, Universidad de Concepción,
Gutiérrez, Assistant Professor, Metallurgical Engineering Chile • Carlos Contreras, Minexx, Chile • Ángel Dávila,
Department, Universidad de Concepción, Chile • Gabriel Díaz & Dávila Ingenieros, Chile • Gustavo de la Torre,
Berkowitz, Metallurgical Processes Director, IM2, Codelco, Minera Milpo, Peru • Boris De Vries, Minterra, Brazil •
Chile • Bert Huls, Principal, Huls Consulting, USA • Claudio Alex Doll, Alex G. Doll Consulting • Miguel Ángel Durán,
García Bernal, Plant Manager, Sierra Gorda SCM, Chile • Universidad Central, Chile • Ricardo Esteban, BBA, Canada •
Jorge Luis Soto, Mineral Processing Manager, Anglo American Bianca Foggiatto, Ausenco, Australia • Dante García, Glencore,
Sur, Chile • Edgardo Tapia, Plant Manager, Teck Carmen de Peru • Cristian Garrido, Mitsubishi, Chile • Julio Giampaoli, El
Andacollo, Chile • Carlos Ahumada, Operations Manager, Teniente Division, Codelco, Chile • René Gómez, Universidad de
Minera Centinela, Chile. Concepción, Chile • Cristian González, IntelliSense.io, Chile •
Christopher Greet, Magotteaux, Australia • Harald Günter,
Koppern, Germany • Levi Guzmán, Moly-Cop, Peru • David
Advisory Committee
Hatton, SGS Minerals, Canada • León Hennen, Independent
Carlos Aguilera, Jiahua Chemicals, Chile • Mario Basaure, SGS
Consultant, Chile • Rodrigo Hurtado, Teck, Chile •
Minerals, Chile • José Antonio Beas, Rockwell Automation,
Antti Jaatinen, Metso, Finland • Simon Jacques,
Chile • Miguel Becerra, El Soldado, Anglo American, Chile •
Magotteaux Andino, Chile • Óscar Jerez, Universidad de
Walter Bergholz, Minera El Morro, Chile • Ramiro Blanch, ALS
Concepción, Chile • Javier Jofré, Independent Consultant,
Chile • Adam Johnston, Transmin, Peru • Bob Maron, Esteban, BBA, Canada • Gonzalo Montes, Universidad de Chile •
CiDRA Minerals Processing, USA • Ron McGregor, Lundin Álvaro Videla, Pontificia Universidad Católica de Chile •
Mining, UK • Nicolás Miranda, Andina Division, Codelco, Christopher Greet, Magotteaux, Australia • Jorge Rosas,
Chile • Gonzalo Montes, Universidad de Chile • Jéssica Hatch, Chile • Bianca Foggiatto, Ausenco, Australia • Antti
Olate, Pampa Norte, BHP Billiton, Chile • Myriam Olguín, Jaatinen, Metso, Finland • Cyril O’Connor, University of Cape
Mathiesen, Chile • Sergio Olmedo, Minera Escondida, Chile • Town, South Africa • Simon Jacques, Magotteaux Andino,
Cyril O’Connor, University of Cape Town, South Africa • Chile • Jorge Venegas, El Teniente Division, Codelco, Chile •
Humberto Pastén, Oxiquim, Chile • Javier Pizarro, Empírica Andreia Rosa, Gerdau Acominas, Brazil • David Robinson,
Consultores, Chile • Juan José Ponce, Minera Chinalco, Peru • CSIRO, Australia • Nicolás Miranda, Andina Division, Codelco,
Julio César Porras, LACC Minerals, Colombia • Jocelyn Chile • Felipe Bello, Minera Centinela, Chile • Miguel Ángel
Quinteros, Pontifica Universidad Católica de Valparaíso, Durán, Universidad Central, Chile • Bob Maron, CiDR A
Chile • Lorenzo Reyes, Universidad Central, Chile • Minerals Processing, USA • Levi Guzmán, Moly-Cop, Peru •
David Robinson, CSIRO, Australia • Patricio Rodríguez, Alex Doll, SAG Milling, Canada • Lorenzo Reyes, Universidad
Outotec, Chile • Andreia Rosa, Gerdau Acominas, Brazil • Central, Chile • Mauricio Torem, Universidade Católica do
Jorge Rosas, Hatch, Chile • Cristian Sáez, Verne, Chile • Rio de Janeiro, Brazil • Juan Collao, Pucobre, Chile • Romke
Héctor Sáez, Asmin, Chile • William Skinner, University of Kuyvenhoven, Empírica Consultores, Chile.
South Australia • Francisco Soto, Empírica Consultores, Chile •
Mauricio Torem, Universidade Católica do Rio de Janeiro, Brazil •
Héctor Toro, Moly-Cop, Chile • Luis Torres, Anglo American,
Chile • César Urcuhuaranga, Minera Antamina, Peru • Jorge
Venegas, El Teniente Division, Codelco, Chile • Álvaro Videla,
Pontificia Universidad Católica de Chile • Sebastián Villalobos,
Metso, Chile • Mauricio Villanueva, Minera Centinella, Chile •
Víctor Zumarán, Minera Escondida, Chile • Pablo Zúñiga,
Andina Division, Codelco, Chile • José Manuel Valenzuela,
Radomiro Tomic Division, Codelco, Chile.

Editorial Committee

EDITOR
Claudia Velásquez, Extractive Metallurgy Consultant, Chile

COPY EDITOR
Oliver Allendes, International Events Coordinator, Gecamin

REVIEWERS
Leopoldo Gutiérrez, Universidad de Concepción, Chile • Javier
Jofré, Chile • Leonardo Soto, Antofagasta Minerals, Chile •
David Hatton, SGS Minerals, Canada • Hector Sáez, Asmin,
Chile • Javier Pizarro, Empírica Consultores, Chile • Ricardo
FOREWORD

Once again, this 12th version of our now traditional International Mineral Processing
Conference, Procemin 2016, has gathered an impressive collection of technical
contributions covering a broad range of relevant aspects of mineral processing
technologies. There are many examples of best practices for concentrator management and
equipment selection, testing and sizing methodologies through mathematical modeling
and simulation for both comminution and flotation (concentration) applications.
As Chair of Procemin 2016, I am pleased to confirm that the technical program so
compiled is well aligned with our initially stated charter: “Innovation: a path, a challenge,
a necessity”. My personal gratitude goes to all those who made it possible with their
high-level contributions.
Conferences of this nature, Procemin 2016 in particular, provide effective
communication platforms to share our accumulated knowledge and most importantly,
to learn from others about various fields of specific interest. Sharing and learning should
be our first step towards the needed innovation in the mineral processing discipline
and the entire mining industry. Then comes the challenge of overcoming our natural
risk adversity and move forward to try and further develop new technological options.
Because the technologies we have today will not be adequate for future generations: true
innovation is a must.
This Procemin 2016 Proceedings files are full of stories, experiences, new proposals
and innovative ideas, coming from various authors from as many as 16 different countries,
including Peru, Brazil, Canada, Germany, United Kingdom and others. I kindly invite you
to explore the contents of these proceedings and enjoy reading those closer to your field of
personal interest.
Sincerely,

Jaime Sepúlveda
Chair
Procemin 2016 • 12th International Mineral Processing Conference
PREFACE

The International Mineral Processing Conference has been a platform for mining
professionals, technicians, operators, researchers and suppliers to present their
accomplishments since the beginning of this conference in 2002. The 2016 slogan,
“Innovation: a path, a challenge, a necessity” has established the primary focus of
discussion for the three-day conference. The organizing committee, comprised of
representatives from Universidad de Concepción, McGill University and Gecamin, had
the challenge of providing the opportunity for mineral process professionals to open
the path to new ideas, accelerate evaluation processes and technological developments.
With the coming of concentration and flotation at the beginning of the 20th century,
came the birth of large-volume mining. Largely due to the introduction of technology
capable of extracting more valuable mineral to the mining value chain, and with that the
ability for better extraction processes in general. Today, and due to the drop in ore grade,
this chemical-physical phase has become all the more important and strategic.
We have seen mineral deposits like Minera Escondida, discovered in 1981 and began
processing minerals in 1990, become the first foreign investment megaproject in the
country. Chuquicamata Division started its activities in 1915 and reached an annual
average of 100 million tons of fine copper during the decade of the 20’s, which led to a
total of 1.6 billion tons extracted from the largest open pit in the world in 85 years and a
total fine production of 33,045,198 ktmf (thousands of metric ton kilos).
The 90’s were the beginning of the mining mineral deposit boom, considering Cerro
Colorado and Candelaria in 1994, Zaldívar in 1995 and Los Pelambres in 1999. In the 80’s
large flotation cells were about 50 m3, while today they are over 500 m3, which shows that
despite the fact that flotation is over 100 years old, it has only been in the last 30 years
that the magnitudes have grown tenfold. Which other technological developments do we
have in store for 2020? We look forward to seeing the future and its advances.
On behalf of the Procemin 2016 organizing committee, I invite you to take part in the
Procemin conference series and present your experiences and applications of technology,
methodology and advances in mineral processing.

Claudia Velásquez
Technical Coordinator
Procemin 2016 • 12th International Mineral Processing Conference
ACKNOWLEDGEMENTS

The Organizing Committee acknowledges with gratitude the efforts of all the authors for
contributing a large variety of high quality, detailed and innovative papers to the technical
program. We also would like to thank the reviewers, Metallurgical Department of the
Universidad de Concepción, McGill University, Canada, the employees from Gecamin, and
all those involved in the creation of these proceedings for their assistance. The support
of the Organizing, Advisory and Technical Committees has been greatly appreciated, as
has been the support of the Procemin 2016 Chair, and the Chairs of technical sessions.
The Organizing Committee also wishes to thank the following sponsors (as of
September 29, 2016) for their generous support:
Platinum: Outotec
Gold: SGS Minerals, Cytec Solvay Group, Emerson, thyssenkrupp
Industrial Solutions
Silver: CiDR A-TIAR, Carl Zeiss Group, Transmin Metallurgical Consultants, Aggreko,
Aminpro, Asmin, Scantech, CRC ORE, Thermo Fisher Scientific
Social: Weir Minerals, Clariant
Lanyard: Emerson
Proceedings: Moly-Cop
Students: Magotteaux
Institutional Partners: Sociedad Nacional de Minería (SONAMI), Chile; Consejo
Minero, Chile; Servicio Nacional de Geología y Minería (SERNAGEOMIN), Chile;
Brazilian Mining Association (IBRAM), Brazil; UK Trade & Investment; Instituto de
Ingenieros de Minas del Perú (IIMP); Valor Minero, Chile; IMPC 2016, Cámara de
Comercio Peruano-Chilena; Embassy of Canada; Minnovex A.G., Chile; Asociación de
Industriales Antofagasta, Chile; Exponor Chile 2017
Official Media: Nueva Minería y Energía, Chile
Official Media TV: Programa Rumbo Minero, Peru
Media Partners: Direcmin, Chile; El Industrial, Chile; Panorama Minero, Argentina;
Mining Turkey; International Mining; E&MJ, USA; Noticias de Mineracao, Brazil;
African Mining Brief; Revista Rumbo Minero, Peru.
Finally, we would like to thank all the delegates who attended the conference and
exchanged their valuable knowledge and expertise, thus contributing to the great success
of this 12th edition of the International Mineral Processing Conference, Procemin 2016.
We are looking forward to seeing you all again during the 13th edition of Procemin, in
the year 2017.

Executive Organizing Committee

Procemin 2016 • 12th International Mineral Processing Conference
PROCEEDINGS SP ONSOR

We proudly acknowledge Moly-Cop as the Proceedings Sponsor of the 12th International

Mineral Processing Conference.
PLENARIES
The Next Big Thing
Peter Amelunxen
Aminpro, Peru

ABSTRACT
We all know the story of froth flotation. The first patents date to 1885, when Carrie Everson
discovered a method of concentrating sulfide minerals from a powdered ore using a froth phase.
Everson’s process was successful in small plants, but not on a larger scale, and because she did not
have the financial resources to continue her work, she returned to being a teacher.
By the turn of the century, technological developments in the U.S. and Europe where creating
higher demand for base metals. This coincided with the depletion of high-grade, secondary ores,
which could be shipped directly to the smelter; the subsequent period of declining ore grades and
higher energy requirements (sound familiar?) created a challenging economic climate for mining
operations. This motivated the development of alternative processes, and by 1906, bulk flotation of
zinc concentrates was working well at the Broken Hill lead and zinc mine in New South Wales.
In 1911, froth flotation was implemented for lead and zinc in the U.S. for the first time—in Basin,
Montana. It was implemented for copper at the Inspiration Mine in Arizona during the mid-1910s.
By 1915, most of the principal copper and lead mines were using, or experimenting with, froth
flotation processes.
Froth flotation took twenty years to progress from development to implementation, but it wasn’t
for ten more years that it was widespread and accepted. When compared to other industries (say,
computer technology, or aerospace), thirty years is a very long time. It is tempting to regard this
time lag as a singular incident, but it is actually a pattern that is readily apparent from the history of
other technical milestones in our field, including, for example, the development of autogenous and
semi-autogenous grinding in the 1930s and 1940s, the introduction of SX/EW during the 1960s, and
the implementation of High Pressure Grinding Rolls for hard rock processing in the late 20 th
century.
In this presentation, we examine some of the reasons for the delays in the adoption of new
technology. We also review some the exciting new developments in our field that may one day
(hopefully in less than 30 years) become the next big thing. Lastly, we discuss some of the
challenges and possible solutions that may help to accelerate the adoption curve of future
innovations.

There is no full article associated with this abstract.

1
Mine-To-Plant Strategies within the Cyclic Copper
Business
Jorge Menacho
De Re Metallica Ingeniería, Chile

ABSTRACT
A macro-economic analysis of the Copper business evolution is presented. A regular cyclic price
pattern along the last fifty years is consistently demonstrated and the probable situation for the next
twenty years is anticipated by means of a novel mathematical model especially developed for this
purpose.
As the high and low price-waves are considered into the long-term business, several existing
paradigms are set into question.
Valorization of the block model is strongly affected when using a cyclic price pattern instead of a
fixed value for the middle and long term. This means the cutoff grade has to be considered as a
cyclic function along the time.
Flowsheet and operational strategies at the Mine and the Plant need to be flexible according to the
macro economical wave change. New strategies arise under this framework. Selectivity at the mine
is highlighted either in the proper mine operations or by handling stocks to be processed at the
plant according to the price cycle evolution.
Flexibility at the plants means to adopt process configurations with lower efficiency but lower cost.
New intermediate products are generated to be processed when the economic conditions pay for it.
Several examples are given to illustrate impacts of this viewpoint which provides a different way to
successfully face the cost and productivity issues in our changing Copper business.

There is no full article associated with this abstract.

1
CHAPTER 1

Comminution: Crushing,
Grinding and HPGR
Upgrade of Hydraulic Adjustment System of Symons
7’ Crusher
Sebastián Villalobos*1, Nelson Mella2, Alejandra Pereda2, Iván Araya2, Maria Luan2,
Alexis Guerra2.
1. Process Optimization, Metso, Chile
2. Mining Crushers and Screens, Metso, Chile

ABSTRACT
Upgrade of equipment represents a significant challenge in order to save on operational
expenditure (OPEX) and avoid large capital expenditure (CAPEX) by adopting new technologies.
This is in addition to other challenges that mining operations are already facing, including
decreasing grades and increasing metallurgical complexity, highest safety standards, adverse
climatic situations and hard-to-access areas.
In 2014 Metso introduced a retrofit for the Symons 7’ crusher to improve the control of its bowl
adjustment making it equivalent to Metso’s current MP and HP crusher models, eliminating the
need for a bowl locking arrangement. The package is composed of new structural and hydraulic
components that are fitted to existing crusher components with some modifications. It also
includes a new PLC driven hydraulic power unit with Ethernet communication capabilities to
allow remote operation and connectivity to the most advanced automation systems. By
implementing the upgrade, Symons 7’ crusher has gained full rotational control throughout the
entire bowl removal and installation process.
This paper presents a case study in which close coordination between the project team and the
crusher plant maintenance team was crucial. The project scope included mechanical modifications
and process optimisation to enable better utilisation and operation of the crusher, particularly in
terms of final product size and throughput. Cooperative teamwork during the project, followed up
by close monitoring of the key variables required to maintain successful equipment operation,
have resulted in significant operational benefits to this operation in terms of equipment
availability, a better utilization of available assets according to power consumption, and ease of
operation and maintenance with a focus on safety.

1
INTRODUCTION
Since 2010 until now, Metso has completed three large crusher upgrade projects in Chile with great
success in terms of return on investment and process improvement. These upgrades include
conversion kits for various applications, ranging from primary crushing to the pebble crushers.
Such investments generate great value, just at a time when productivity of invested capital is a key
issue for the global mining and metals sector as falling commodity prices and a rising supply are
ushering in a period of restraint in capital project investment. These kinds of investment projects
consider much less resources and greatly reduce the risk of budget overrun for the operation.
In worldwide context, despite an estimated US$20b–US$55b being spent globally in up-front
engineering and design, average budget overruns were 62% and 50% of projects were reporting
delays, as shown the Figure 1.

100% 88%
73%
80% 62% 60%
60% 40%
40% 29%
27%
20%
0%
Overall Europe North Africa Asia Latin Oceania
America America

Source: EY study fo projects in the mining sector and metals

sector (October 2014).

Figure 1 Average cost overruns by region

This case study for an upgrade of a Symons 7’ crusher equivalent to Metso’s current MP and HP
crusher models demonstrates the benefits of upgrading to maximize asset utilization and the
associated reduced risk of budget overrun.

METHODOLOGY

Overview of Upgrade
The latest generation of Metso crusher upgrades breathe new life into the existing crushers, making
operations more productive, cost efficient and safer overall with an advanced bowl adjustment
control, hydraulic motor bowl adjustment system and hydraulic tramp release and clearing system.
The following summarises the components of the retrofit kits and their benefits.
Advanced bowl adjustment control

A good challenge is to find affordable technological solutions to improve the existing equipment
when the market is not favourable to large CAPEX investments. A new generation hydraulic
clamping system improves the functionality of crusher, allowing the 7’ Symons crusher to have the

2
same level of bowl adjustment automation capabilities as those provided on current HP and MP
series crushers. By implementing this upgrade, the Symons crusher will have full rotational control
throughout the entire bowl removal and installation process. This allows faster bowl adjustment,
removal and installation times as well as minimizing labor requirements for all operations
involving bowl rotation activities.
Hydraulic motor bowl adjustment system

The new generation bowl clamping system provides faster and safer bowl adjustment than the
manual or hydraulic ramstyle systems. It also eliminates the mechanical bowl locking arrangement,
replacing it with a hydraulic break system. Heavy-duty hydraulic motors turn the ring gear up to
three full turns in one minute, speeding both bowl adjustment and bowl removal during liner
change-out.
Hydraulic tramp release and clearing system

A crusher equipped with the new generation tramp release permits a quick and convenient clearing
of a plugged cavity and allows for the inspection of crusher liners and the repair of mainframe
liners in “clear mode”. The retrofit minimizes the potential for stalling, allowing the crusher to
operate with a smaller closed side setting for finer feed to the downstream circuit. This results in
more crushing and less grinding. In addition, it also maintains a constant adjustment ring “hold
down” force, minimizing ring bounce. Along with the above, the Symons crusher 7’ is transforming
into a machine with the latest technology through the addition of new components and
modification of existing components, including:
a) New or modified bin.
b) New adjustment cap.
c) New dust seal.
d) New ring gear.
e) New clamping ring
f) Set ring modified.
g) Modified bowl.
Figure 2 shows the addition and modification of existing components.

3
 Figure 2 Detailed components drawing showing both the original and modified parts

CASE STUDY

Upgrade of the hydraulic adjustment system of a Symons 7’ crusher

Owned by Codelco Chile, Andina is a copper and molybdenum deposit located in the province of
Andes, Chile. In 2015-2016, Metso assisted Andina in developing a road map for the success of this
campaign that was to update the Symons N° 5 crusher of the quaternary stage as shown in Figure 3:

Figure 3 Flow-sheet showing the tertiary and quaternary crushing stages in Andina

4
Objectives

In order to avoid deviations and overruns in project implementation, the project team, composed of
Andina personnel, Metso MCS (Mining Crusher Solution) and Metso PRO (Process Optimization)
team members, defined the objectives of the work to be:
a) Increase the availability and upgrade the Symons 7’ N°5 crusher to enable advanced bowl
adjustment control automation capabilities as per those provided on current HP and MP
series crushers; and
b) Provide greater security and minimize risk exposure with the replacement of the old
system of springs with new hydraulic cylinders.

While these were the initial objectives, on-site support and the close communication with Andina
staff led to important recommendations to:
c) Propose and install a new crusher liner which increases the useful crushing area; and
d) Evaluate the results from the perspective of process analysis.

Project Start up

The development of the project initiation strategies began at the end of November 2015 and
involved co-ordination meetings prior, strategic contracts and the addition and modification of
existing components. The people training strategies included visits, team building and assistance
for supervisors and operators. Finally, crusher N°5 was commissioned on 27 April 2016 following
an extensive test phase and a thorough check of the entire set of parameters, including:
a) Ring breaks
b) Oil leaks
c) CSS (close side setting)
d) Tramp release pressure within the normal operating ranges (2150-2300 psi)
e) Clamping pressure within the normal operating ranges (2200-2600 psi)
f) Difference in temperature between oil outlet and return in the temperature controller.
In parallel with these activities, the performance of the recommended liners was evaluated to meet
customer requirements for throughput and size reduction, as shown Figure 4.

Figure 4 Old liners (left) and recommended genuine liners (right)

5
The recommended genuine liners, with CSS 8 mm have been installed and have produced a 6%
increase in volume from the longest parallel zone, allowing an increase in throughput of between 6
and 10%.

RESULTS AND DISCUSSION

Analysis of historical process data

Metso PRO analysed the historical data from the PI system, covering the following periods: from
09-01-2015 to 03-15-2016 (6.5 months) before the upgrade, and from 04-23-2016 to 05-17-2016 (24
days) after the upgrade. As shown Figure 5, availability increased from 65% before to 88% after the
upgrade. Each marked area shows the equipment downtime.

Figure 5 Availability before and after upgrade

Figure 6 Tonnage processed by crusher before and after upgrade

6
As shown the Figure 6, the average throughput before and after the upgrade increased by around
10%, from 301 to 336 tph with a closed side setting of 9 mm. In the same way, the current was
analysed to compare the power consumption. Figure 7 shows how current consumption increased
from 14.8 before to 20.4 A after the upgrade, which is approximately 60% of the nominal.
It is important to note that crushers operate better if they are using 60% or more of the nominal
current (and by analogy, available power) since they are better utilising the installed mechanical
value for optimal process performance.

Figure 7 Current consumption of crusher before and after upgrade

CONCLUSIONS
Upgrading a crusher with a retrofit kit can be a cost-effective alternative to replacing the entire
piece of equipment since it results in better asset utilisation, and the upgraded crusher incorporates
the latest mechanical safety and operational features for improved overall performance.
This case study demonstrates the benefits of a crusher upgrade on equipment availability, safety
and process performance. According to the historical data from the PI system a 23% increase in
availability, from 65% to 88%, resulted after the crusher upgrade.
Metso recommended setting the availability to 85% and recovering the remaining 3% for
conducting equipment inspections (clampings leaks, ring jumps) as well as adjusting the CSS at
least once every 12 hours.
The process data analysis also shows an increase in tonnage processed from 301 to 336 tph (~10%).
These results demonstrate valuable operational improvements. However, further benefits in overall
plant production could be achieved if a full evaluation of the feed ore characteristics and blending
practices are conducted to achieve the optimal product size from the primary crushing circuit.
The analysis also demonstrates that the upgrade resulted in a better utilization of available assets,
in particular the installed motor power. This is shown by the increase in current from 14.8 A before

7
to 20.4 A after the upgrade, bringing the current reading into the desired operational range of 60%
or more of the nominal.

REFERENCES
EYGM Limited, Business risks facing mining and metals 2015–2016, 2015.
Metso Corporation, New generation crusher upgrades, Leaflet number 3251-04-16-MSE/Ersmark -
ENGLISH ©2016.
Metso Corporation, Advanced bowl adjustment control, Leaflet number 3252-04-16-MSE/Ersmark -
ENGLISH ©2016.
Metso Corporation, Hydraulic motor bowl adjustment system, Leaflet number 3253-04-16-
MSE/Ersmark - ENGLISH ©2016.
Metso Corporation, Hydraulic tramp release and clearing system, Leaflet number 3254-04-16-
MSE/Ersmark - ENGLISH ©2016.

8
More than an Alternative? Shell Supported Ball Mills
in Mega Comminution Plants
Egbert Burchardt 1, Christian Nieland1 and Alexander Baumbach3
1. Thyssenkrupp Resource Technology, Germany
2. Thyssenkrupp Industrial Solutions, Chile

ABSTRACT
Ball Mills have been the state of the art comminution system for fine grinding over centuries.
“Everybody” in the Mining Industry understands the design and operation of these mills. In the
middle of the 19th century, the need for fine grinding mills had become evident. At that time,
stamps driven by steam engines were used, which were very limited in grinding fines. The first ball
mill in the mineral industry, built by Krupp Grusonswerk Germany, was delivered in the year 1905
to an Australian gold mine. Since then, a continuous design development and growth of mill sizes
commenced.
For the last decades, trunnion supported mills were the default choice in the mining industry, with
few exceptions. In 2004, Phelps Dodge was the first mining company that decided to employ shell
supported mills for a new mega concentrator - Cerro Verde. Few years later, the new 110,000 tpd
plant of Sierra Gorda SCM introduced this particular ball mill design also into the Chilean Mining
Industry. All the shells of these mills received final welding on-site. Although most of the suppliers
in the mining industry follow the trunnion supported design, shell supported ball mills have
received more and more recognition and acceptance in the mining industry - especially if large ball
mill sizes are required. The perception is growing, that both concepts are technically reliable and
proven.
This paper summarizes particular design features and manufacturing strategies of shell supported
mills and their specific advantages. It summarizes the reasons why Mining Companies selected this
design for their mega concentrators.

1
Mitigating the Impact of Sea Water on Copper
Flotation Response through Optimizing Grinding
Chemistry
Jessica Kinal1, Simon Jacques1 and Christopher Greet2
1. Magotteaux Andino , Chile
2. Magotteaux, Australia

ABSTRACT
As the supply of fresh water is scarce in Chile, particularly in the north region, mining operations
are increasingly moving towards the use of sea water in their processing plants. In fact, it has been
predicted by the Chilean Copper Commission that the use of sea water (either direct or desalinated)
will rise from 16% in 2015 to 50% by 2026.
The most common issue with the use of sea water in flotation is the buffering effect whereby a
greater amount of lime is required for adequate pH adjustment. Additionally, the precipitation of
colloidal particles at high pH not only causes issues with scaling, but these colloids are hydrophilic
and as such act as a depressant for flotation. On the other hand, with the high level of dissolved
salts in sea water, these electrolytes can have a positive impact on the formation of a stable froth
and enhance the flotation of naturally hydrophobic minerals.
In order to mitigate the negative impacts of sea water whilst capitalising on the positive benefits,
the pulp chemical conditions of the system need to be optimised. In particular reference to pyrite
depression in copper systems, where the use of lime with sea water can be problematic, it becomes
important to consider innovative solutions. Changing the grinding media alloy type in the mill
from traditional forged steel to more inert high chrome media modifies the pulp chemistry and
leads to enhanced copper flotation response, along with improved selectivity against pyrite and
lower lime consumption, regardless of the water type used.
This paper outlines the results from a series of laboratory grinding and flotation tests on a number
of copper ores whereby a range of water qualities are utilised, comparing the impact of grinding
with forged steel and high chromium media at various pH levels.

1
INTRODUCTION
Given that the majority of mineral deposits in the world are located in remote arid areas, the
consideration of water supply has a direct impact on the development of new mining operations,
expansion of existing plants, or even the continuing operation of brownfields deposits. The current
practice is to recycle water as much as possible, however due to losses through drainage,
evaporation or contamination, a percentage of top up water is still required (typically 20-40%).
Gleik (1996) estimates that the main source of water on earth comes from the ocean, at 96.5%, with
fresh water from groundwater, lakes and rivers comprising only 0.768%. Thus it is quite clear,
especially given the inherent importance of protecting fresh water sources for human use, that it
makes good sense to look towards saline sources for water supply to industrial operations. The first
thought is often to desalinate the water, with the most widely used process being reverse osmosis
(Greenlee 2009). However, this is a high cost process and the side effect of concentrated brine
solution production brings its own environmental issues. Therefore operations are looking more
and more towards the use of seawater without desalination (Castro 2012).
Chile is the world’s largest copper producing country, with the majority of these mines situated in
an extremely dry desert area. According to Hiam-Galvez et al (2011), the current total Chilean
mining water consumption is estimated to be 11.5 cubic metres per second. Given the issues and
restrictions with sourcing fresh or groundwater for these mines, it has been predicted by the
Chilean Copper Commission (COCHILCO) that the use of sea water (either direct or desalinated)
will rise from 16% in 2015 to 50% by 2026. Some mines in Chile already use saline water; an
example being the Minera Esperanza plant pumping water some 145km from the Pacific Ocean to
the mine site, with the only pre-treatment of the water being removal of any bio-solids.
Additionally, Sierra Gorda and Las Luces in Chile, Batu Hijau in Indonesia and Çayeli Bakir in
Turkey use seawater in their flotation processes (Wang and Peng, 2014).
Proportionally, most of the water used in mining is in the flotation process (Dunne 2012) with only
small amounts consumed across the rest of the operation for activities such as dust suppression and
human consumption. Therefore, the water quality requirement should be determined based on the
ability to obtain cost effective valuable mineral recovery through flotation; a surface chemistry
separation process. It must be noted that almost all flotation chemicals are designed for use in fresh
water (Peng, Zhao and Bradshaw, 2012) so this should also be a consideration when designing a
new operating plant that is to use saline water.
Liu, Moran and Vink (2013) provide a comprehensive list of the negative consequences of water
quality on flotation which includes: reduction in particle surface hydrophobicity; change in particle
surface charge; inadvertent activation of unwanted minerals; slime coatings on mineral surfaces;
interactions of ions in the water with flotation reagents and oxygen consuming microbial growths.
There are also positive consequences that can come from compromised water quality which include
compression of the electrical double layer, formation of smaller bubbles, and residual reagent
return resulting in a potential cost saving.
Sea water has a high ionic strength. Certain ions, such as Cs+ and I- are considered “structure
breaking” and others such as Li+, Mg2+, F- and Cl- “structure making” (Wang and Peng 2014). The
structure of the interfacial water is important, as if the structure of water is strongly bonded due to
structure making ions, it will be hard for the collector molecules to reach the surface and be
adsorbed (Hancer et al. 2001). Along with the retardation of collector attachment, these ions are
often hydrophilic, further hindering flotation. With up to 80% of water being recycled in processing

2
plants, the quantity of ions can build to high levels which has an important impact on the recovery
of sulphide minerals (Sinche et al. 2013).
Flotation of copper porphyry ores involves the need to depress pyrite in order to upgrade the final
concentrate. Lime is the commonly used reagent of choice for this; however it is well known that
pH buffers in saline water, with the amount of lime required to raise the pH increasing
exponentially, particularly after pH 10.5. Parraguez et al. (2009) have studied Pourbaix diagrams to
understand the mechanisms behind this buffer effect, finding a variety of reasons. Firstly, the
reaction of calcium ions with mineral species formed by carbonate and magnesium precipitate out
in the slurry above pH 10.5. Given calcium is a pyrite surface modifier, if these ions are being
consumed by chemical reactions, there will be a loss of effectiveness of the depression of the iron
sulphide. Additionally, above pH 10.5, magnesium hydroxide is formed which consumes large
quantities of hydroxyl ions, explaining the plateau zone, preventing further increase in pH along
with the retardation of oxidation of pyrite. To add to these issues, magnesium hydroxide deposits
on the surface of the minerals affecting their floatability and recovery (Hirajima et al. 2016).
The impact of grinding media on pulp chemistry and subsequent flotation response is well
understood (Gonςalves et al., 2003, Huang and Grano, 2006, Bruckhard et al., 2011), however not so
much literature has been produced when considering flotation in high ionic strength water. This
paper attempts to increase the knowledge in this area of flotation.

EXPERIMENTAL

Sample preparation
Bulk samples of copper porphyry ores were collected from site, one from Northern Chile and the
other from Eastern Australia. The ores were initially crushed to 100% passing 2 mm before being
thoroughly mixed, split, bagged into sub-charges and kept in a freezer prior to testing. Samples
were submitted for head assay, as well as mineralogical analysis by QEMSCAN, and are presented
in Table 1.

Table 1 Elemental and calculated mineralogical head assays for the two ore samples tested

Sample Cu, % Fe, % Mo, ppm As, ppm S, % IS, % NSG, %

Australia 0.9 4.0 - 125 0.6 0.2 98.1

Chile 1.7 3.1 69 - 3.2 4.6 92.7

Water
A variety of water sources were used. Tap water, site process water, seawater and 50:50 blends
were used. Each water type was analysed for cation and anion content (Table 2).

3
Table 2 Analysis of the various water types used during testing

Water, mg/L Tap Water Mine Process Tap/Sea Process/Sea Seawater

TDS 330 5,200 20,000 23,000 41,000

Cl- 110 770 11,000 11,000 20,000

NH4+ <0.01 0.28 0.02 <0.01 0.02
NO3- 0.11 0.08 0.07 0.06 0.02
SO42- 15 980 510 980 1000

Ca2+ 23 210 210 370 430

Mg2+ 14 92 650 820 1,300
K+ 3 130 180 290 390
Na + 58 1,200 5,000 5,800 10,000

HCO3- 92 44 120 98 160

CO32- <10 11 <10 <10 <10

Grinding and flotation tests

A series of grinding and flotation tests using a variety of water qualities, pH levels and grinding
media alloy types were completed for each sample. The test procedure was the same for all samples
tested, within each ore type. Grinding was completed using the Magotteaux Mill® which allows the
control and monitoring of the pulp chemistry during grinding (Greet et al., 2004). A sample of ore
was ground with 2.5 litres of water and 15 kg of grinding media to achieve the desired P80. The mill
speed was adjusted for the two alloys tested to keep the particle distribution the same for the tests
within each ore type. The Magotteaux Mill® discharge pulp chemistry was calibrated to match plant
conditions and the grinding procedure then kept constant for all media and water types tested.
After grinding, the pulp was transferred into a 5 L Magotteaux Flotation Machine and
homogenised during two minutes. The pulps were conditioned as per the operating plant, with
regards to addition of lime for pH adjustment and dosing of specific reagents. The Australian ore
was processed at natural pH as per plant conditions, and the Chilean ore processed at standard pH
(10 during grinding and 11 for flotation) as well as reduced pH (9 during grinding and 10 for
flotation).
Following conditioning, four timed copper rougher concentrates were collected for 0.5, 1.5, 3, and 5
minutes, for a total flotation time of 10 minutes. The froth was hand scraped every 10 seconds. The
same water quality that was used during grinding was also used to maintain the pulp level in the
flotation cell. The pulp chemistry (Eh, pH, DO and Temperature) was recorded using a 90 MLF TPS
meter during grinding, and before and after flotation. All tests were completed in triplicate to
obtain confidence in the data produced. The reproducibility was examined for each condition and
tests repeated if necessary.

4
Chemical analysis
The liquid and solid samples for the Australian tests were assayed at Bureau Veritas chemical
laboratory in Adelaide, Australia. The assays for the Chilean tests were completed by ALS,
Santiago. Iron sulphide (IS) and non-sulphide gangue (NSG) were deducted from the elemental
assays using the following equations:
%Cp (Chalcopyrite) = %Cu x (183.3/63.5) (1)
%IS (Iron Sulphide) = (%S – (2 x 32.06/183.3) x %Cp) x (119.85/2 x 32.06) (2)
%NSG (Non-sulphide Gangue) = 100 - %IS - %Cp (3)

RESULTS AND DISCUSSION

Impact of water on grinding chemistry

The buffering effect is demonstrated in Figure 1 where the requirement for lime during grinding
and flotation to achieve the same pH increases as the water quality moves from process water
through to seawater. As expected, the higher the target pH, the greater the impact of water quality
on lime consumption.

12000

10000
Lime Addition, g/t

8000

6000

4000

2000

High pH Low pH

Figure 1 Lime requirement to target pH for various water types

5
 200 3,0

150 2,5

Eh, mV (SHE) 2,0

DO, ppm
100
1,5
50
1,0
0 0,5

-50 0,0

Eh DO

Figure 2 Grinding pulp chemistry as a function of water quality, forged steel media, natural pH

Keeping the pH the same (natural) and turning the focus to look at the impact of water type on
pulp potential, Figure 2 shows a very clear trend with the use of forged steel media whereby as the
ionic strength of the water increases, the pulp potential decreases, moving from oxidative (positive
SHE readings) for clean tap water to reducing conditions (negative SHE readings) for seawater. The
same trend was seen with dissolved oxygen levels.
Impact of water quality on copper flotation response
The flotation results of tests using the two water types of most differing ionic strength after
grinding with high chrome media are shown in Figure 3. It is interesting to note the flotation
kinetics are faster for the seawater tests, but with low copper concentrate grade due to the poorer
selectivity against gangue minerals. For the remainder of the paper, the results will be discussed as
a function of recovery at a particular grade (as demonstrated below) or grade at a particular
recovery.

6
 40
35
30

Cu Grade, %
25
20
15
10
5
0
0 20 40 60 80 100
Cu Recovery, %
Tap Water Seawater

Figure 3 Copper grade/recovery curves comparing flotation response after grinding with high chrome media
in tap and seawater demonstrating the selection of data at a particular copper grade

The impact that water quality has on grinding pulp chemistry follows through to flotation
response, with the copper recovery at 15% copper grade for the forged steel tests (Figure 4)
deteriorating markedly as the ionic strength of the water increases, and the pulp potential and
dissolved oxygen levels decrease. Fixing the copper recovery at 82% demonstrates that as the water
quality deteriorates, copper concentrate grades decrease due to poorer selectivity against non-
sulphide gangue (Figure 5). Given non-sulphide gangue is generally hydrophilic and not amenable
to flotation, this is a strong indication that the froth structure has become more stable with the
higher ionic strength water, inhibiting free flowing drainage and thus recovering more gangue
through entrainment.

88
86
84
Cu recovery, %

82
80
78
76
74
72
70

Forged

Figure 4 Effect of water quality on copper recovery at 15% copper grade, forged steel grinding media

7
 14 14
12 12

NSG Recovery, %
10 10
Cu grade, % 8 8
6 6
4 4
2 2
0 0

Copper NSG

Figure 5 Effect of water quality and grinding media type on copper grade and non-sulphide gangue recovery
at 82% copper recovery, forged steel grinding media

By choosing two water types to focus on the impact of high and low pH on flotation, it can be seen
that as expected, increasing pH depresses iron sulphide recoveries regardless of water quality
(Table 3). The requirement for lime to depress pyrite is significantly greater for seawater, as the
poorer water quality affects selective copper flotation. However with the addition of 5 times more
lime to reach target pH than that needed with the use of process water (Figure 1), copper flotation is
even further retarded despite the depression of pyrite. It is expected that the precipitation of
various hydroxyl species as well as calcium ions on the particle surfaces in high ionic strength
water has negatively affected flotation response.

Table 3 Cu recovery at 14% Cu grade, and IS and NSG recoveries at 85 percent Cu recovery
comparing water type and pH, forged steel media

Recovery, %
Water Type pH level Cu @ 14% Cu IS @ 85% Cu NSG @ 85% Cu
grade recovery recovery

Process Low 83.4 63.1 12.0

Water High 83.6 59.3 6.6

Low 81.6 74.3 7.4

Seawater
High 79.6 57.9 16.9

Countering water quality through optimised grinding chemistry

Given that the use of seawater is becoming more and more inevitable in today’s economic and
environmental climate, innovative solutions should be considered. Optimising grinding chemistry

8
is one clear option. Tests were completed using a high chromium media alloy to determine whether
the negative impacts that the use of highly saline water can be negated through the adjustment of
grinding pulp chemistry.

300 12
11
250 10
9
Eh, mV (SHE)

DO, ppm
200 8
7
150 6
5
100 4
3
50 2

Forged Eh HiCr Eh
Forged DO HiCr DO

Figure 6 Grinding pulp chemistry as a function of water quality, comparing the reactivity of forged steel and
high chrome media

As seen earlier, water quality has a substantial impact on grinding pulp chemistry when using
traditional forged steel media. However the introduction of the more electrochemically inert high
chrome media almost completely negates this reactivity (Figure 6). Firstly, the pulp potentials and
dissolved oxygen levels measured during grinding (at the same pH) are consistently higher for the
high chrome media, which should have a positive effect on copper flotation response and the
depression of pyrite through selective oxidation. Also, very interestingly, as the ionic strength of
the water increases, the reactivity of the system with the use of high chrome media decreases, with
the pulp potentials and dissolved oxygen levels even increasing slightly.
Correlating the impact of grinding media type on pulp chemistry with flotation performance, here
it is very clear that by using high chromium grinding media the flotation response of copper is able
to be maintained at a superior level compared with the more reactive forged steel media, regardless
of the water quality being used (Figure 7). Measuring concentrate quality through the recovery of
gangue minerals, at any given copper recovery (Figure 8), the more reactive forged steel media
results in higher recovery of both iron sulphides and non-sulphide gangue compared with the more
inert high chrome alloy. Interestingly, as the water quality worsens, the difference between both
media decreases, probably due to the deterioration of pulp chemistry leading to retardation of
flotation of all minerals (i.e. lower mass recovery overall). For the seawater tests, the selectivity
against non-sulphide gangue remains superior for high chrome media, resulting in a higher
concentrate grade along with the significantly higher copper recovery.
It is clear that the best flotation performance is achieved with the cleanest system; that is, with high
chrome grinding media and tap water. However as the water quality in operating plants is
inevitably of lower quality (at the very least process water and moving more towards the use of

9
seawater), it is obvious that grinding chemistry plays an even bigger role in maintaining optimal
selective copper flotation.

90

85
Cu recovery, %

80

75

70

Forged HiCr

Figure 7 Effect of water quality and grinding media type on copper recovery at 15% copper grade, comparing
forged steel and high chromium grinding media

90 18
80 16
70 14
NSG Recovery, %
IS Recovery, %

60 12
50 10
40 8
30 6
20 4
10 2
0 0

HiCr IS Forged IS HiCr NSG Forged NSG

Figure 8 Effect of water quality and grinding media type on iron sulphide and non-sulphide gangue recovery
at 82% copper recovery

10
CONCLUSIONS
The requirement for lime to achieve target pH increases exponentially with the increasing ionic
strength of water, with this trend more obvious at higher pH targets than low. Increasing lime
retards copper flotation, along with pyrite, although this is somewhat alleviated with the use of
high chrome grinding media.
When grinding with the electrochemically reactive forged steel media, increasing the ionic strength
of the water leads to a strong decrease in electrochemical potentials, resulting in worsening copper
flotation performance with lower recoveries at any given copper grade, and higher recoveries of
iron sulphide and non-sulphide gangue minerals for any given copper recovery. The use of high
chrome media counters the impact of water ionic strength on electrochemical potentials, and thus
flotation response, with a much reduced negative impact of sea water on copper flotation and
selectivity against iron sulphide and non-sulphide gangue.
This study highlights the importance of considering grinding chemistry to negate the impacts of
sweater on flotation of copper ores. Using a more electrochemically inert media such as high
chromium results in better copper grades and recoveries even with a high ionic strength water, and
as such should be considered as the grinding media of choice in Chilean mining operations who use
seawater or saline groundwater in flotation.

ACKNOWLEDGMENTS
The authors wish to give thanks to the mines in Chile and Australia who provided the ore samples
and flotation test targets, along with the Magotteaux metallurgical staff in the Chilean and
Australian laboratories for undertaking this test work.

REFERENCES
Bruckard W.J., Sparrow G.J., and Woodcock J.T. (2011) A review of the effects of the grinding
environment on the flotation of copper sulphides. International Journal of Mineral Processing 100,
pp 1–13
Castro S. (2012) Challenges in flotation of Cu-Mo sulfide ores in sea water. In: Water in Mineral
Processing. Edited by J. Drelich. Colorado: SME, pp 29-40.
Dunne R. (2012) Water water everywhere and not a drop to drink. In: Water in Mineral Processing.
Edited by J. Drelich. Colorado: SME, pp 1-15.
Gleick P.H. (1996) Water Resources. In: Encyclopedia of Climate and Weather. Edited by S.H.
Schneider. New York: Oxford University Press, 2, pp 817-823.
Gonςalves K.L.C., Andrade V.L.L., and Peres A.E.C. (2003) The effect of grinding conditions on the
flotation of a sulphide copper ore. Minerals Engineering 16, pp 1213–1216.
Greenlee L.F., Lawler D.F., Freeman B.D., Marrot B., and Moulin P. (2009) Reverse osmosis
desalination: Water source, technology, and today’s challenges. Water Research 43 pp 2317-2348.
Greet, C.J., Small, G.L., Steinier, P., and Grano, S.R. (2004) The Magotteaux Mill®: investigating the
effect of grinding media on pulp chemistry and flotation performance. Minerals Engineering 17, pp.
891–896.

11
Hancer, M., Celik, M.S. and Miller, J.D. (2001) The significance of interfacial water structure in
soluble salt flotation systems. J. Colloid Interface Sci. 235 (1), 150–161.
Hiam-Galvez D., Fleming H., Ghobeity A., and Osores O. (2011) Water supply in Peru & Chile —
the challenges and the solutions. In: Proceedings of the 8th International Mineral Processing
Conference, GECAMIN, Santiago Chile, pp 395-407.
Hirajima, T., Suyantara, G.P.W., Ichikawa O., Elmahdy A.M., Miki H., and Sasaki K., (2016) Effect of
Mg2+ and Ca2+ as divalent seawater cations on the floatability of molybdenite and chalcopyrite.
Article in Press Miner. Eng.
Huang, G. and Grano, S. (2006) Improving the flotation behavior of a sulfide ore by controlling
electrochemical interactions during grinding, Electrochemical Society Transactions, Volume 2,
Electrochemistry in Mineral and Metal Processing VII, pp 9 - 20.
Liu W., Moran C.J., and Vink S. (2013) A review of the effect of water quality on flotation, Minerals
Engineering, 53, pp 91-100.
Parraguez, L., Bernal, L. and Cartagena, G. (2009) Chemical study for selectivity and recovery of
metals sulphides by flotation using sea water, In: Proceedings of the 6th International Mineral
Processing Conference, GECAMIN, Santiago Chile, pp 323-333
Peng Y., Zhao S., and Bradshaw D. (2012) Role of saline water in the selective flotation of fine
particles. In: Water in Mineral Processing. Edited by J. Drelich. Colorado: SME, pp 1-15.
Rumball J. A. and Richmond G. D. (1996) Measurement of oxidation in a base metal flotation circuit
by selective leaching with EDTA, International Journal of Mineral Processing, 48, pp 1 to 20.
Sinche M., Levay G., and Zanin M. (2013) A case study on the effect of recycled process water on
Cu-Mo flotation. In: Proceedings of the 10th International Mineral Processing Conference,
GECAMIN, Santiago Chile, pp 361-369.
Wang B., and Peng Y. (2014) The effect of saline water on mineral flotation – A critical review,
Minerals Engineering, 66-68, pp 13-24.

12
A Public Database of Tumbling Mill Grindability
Measurements and their Relationships
Alex Doll
Alex G. Doll Consulting. Canada

ABSTRACT
This work presents a public database of over 800 grindability measurements and a set of equations
for converting between different grindability tests based on this database. Several laboratory
grindability measurements commonly used in the mining industry; each is generally applicable to a
particular grindability model and is incompatible with other models. Conversion between different
test types is possible using a series of empirical relationships between those tests conducted at
similar size classes.

The commonly used grindability tests included in the database are the Bond work indices for ball
milling, rod milling and crushing; the drop weight test results A, b, A×b, DWi, Mia, Mic, Mih and
ta; SAG grindability index, SGI or SPI™; and other values such as Mib and point load index.

Some examples of power-based model specific energy predictions will be compared to published
mill surveys to observe how well the different models predict the specific energy of an industrial
mill.

1
INTRODUCTION
Grindability measurements are a key input to the design and optimization of mineral process
plants. As such, grindability parameters are often reported in published documents such as
technical papers (here at Procemin) and in project evaluation reports such as the NI43-101 reports
issued by companies listed on Canadian stock exchanges. Collecting and comparing these
published grindability measurements provides a basis for basic research, such as calibration of
specific energy models, and for benchmarking new projects.
Several specific energy consumption models have been published that require empirical laboratory
measurements; often models use grindability measurements that are distinct from other models
and incompatible with certain laboratory tests. Another benefit of collecting a database of
grindability measurements is to provide relationships for comparing and, possibly, converting
between the different measurement types.

METHOD
Grindability results have been collected from a large number of published documents. These have
been entered into a database containing discrete tables for each class of test and where the test
results are indexed using a `sample name` to link multiple tests performed on the same sample. If
the report offers geological or other differentiating characteristics of a sample, these are captured in
a `lithology` table. The database includes a summary `view` that consolidates the sample name, a
unique ID number for the sample, the originating project or mine, and some of the most common
grindability measurements. All tests on the same sample are identified with the same ID number,
so the relationship between tests on the same sample can be tracked across the different database
tables.
The database includes fields for optional details of the various tests. These optional data are
entered in the database if they are published, and are left blank otherwise. Few authors provide a
complete tabulation of the test details, so many of the detail fields are blank.
Example data from the Lithology table is given in Table where the sample's name, ID number, the
sample lithology (where it is known), and reference information is provided. The summary view
for these samples is given in Table. Note that the ID numbers in this table match the ID numbers in
Table (and all the other database tables).

2
Table 1 Example Lithology database table

Id Name Litho Litho comment

1529 Lik composite 7 Zazu metals NI43-101 March 3, 2014

1600 Gamsberg Pyrite van Drunick Gerold & Palm, IMPC2010

1601 Gamsberg Pyrrhotite van Drunick Gerold & Palm, IMPC2010

1602 Gamsberg Magnetite van Drunick Gerold & Palm, IMPC2010

1603 New Gold Hypogene New Gold NI43-101, Dec 2009

1604 New Gold Mesogene New Gold NI43-101, Dec 2009

1605 Huckleberry SAG feed 2012 Wang et al, CMP 2013

1606 Huckleberry HPGR product Wang et al, CMP 2013

1607 Cadia Hill Englehardt et al, SAG 2011 speaker notes

1608 Ridgeway Englehardt et al, SAG 2011 speaker notes

1609 Cadia East Englehardt et al, SAG 2011 speaker notes

1643 53392-2 Metasediments Alacer Gold NI43-101 July 29, 2014

Table 2 Summary view of major grindability results for example samples

Id Name Program WiBM WiRM WiC Density Axb SGI Ai

1529 Lik composite 7 Lik 13.9 13.6 6.8 67.3 0.17

1600 Gamsberg Pyrite Other 13.25 3.52 79.65

1601 Gamsberg Pyrrhotite Other 12.9 3.42 58.6

1602 Gamsberg Magnetite Other 13.9 3.61 67.9

1603 New Gold Hypogene Other 21.8 18.5 8 2.7 0.16

1604 New Gold Mesogene Other 19.8 18.3 8 2.77 0.07

1605 Huckleberry SAG feed 2012 Huckleberry 18 2.76 31.1

1606 Huckleberry HPGR product Huckleberry 15.4

1607 Cadia Hill Cadia 17.5 20 30 35 0.33

1608 Ridgeway Cadia 18.7 21 30 42 0.43

1609 Cadia East Cadia 20.3 29 30 29.7 0.2

1643 53392-2 Çöpler 13 2.58 84.6 76.7 0.26

3
Table 3 Bond ball mill work index database table for example samples

WiBM WiBM WiBM WiBM

Id Name µclosing f80 p80 gpr WiBM Synthetic Laboratory Comment

1529 Lik composite 7 13.9 ALS Kamloops Zazu metals NI43-101 March 3, 2014

1600 Gamsberg Pyrite 106 14.1 van Drunick Gerold & Palm, IMPC2010

1600 Gamsberg Pyrite 150 12.4 van Drunick Gerold & Palm, IMPC2010

1601 Gamsberg Pyrrhotite 106 13.5 van Drunick Gerold & Palm, IMPC2010

1601 Gamsberg Pyrrhotite 150 12.3 van Drunick Gerold & Palm, IMPC2010

1602 Gamsberg Magnetite 106 14.6 van Drunick Gerold & Palm, IMPC2010

1602 Gamsberg Magnetite 150 13.2 van Drunick Gerold & Palm, IMPC2010

1603 New Gold Hypogene 21.8 SGS Lakefield New Gold NI43-101, Dec 2009

1604 New Gold Mesogene 19.8 SGS Lakefield New Gold NI43-101, Dec 2009

Huckleberry SAG feed Wang, Nadolski, Mejia, Drozdiak & Klein,

1605 106 2578 77 0.99 18 SGS Lakefield
2012 CMP 2013

Huckleberry HPGR Wang, Nadolski, Mejia, Drozdiak & Klein,

1606 106 2302 76 1.2 15.4 SGS Lakefield
product CMP 2013

1607 Cadia Hill 17.5 Englehardt et al, SAG 2011 speaker notes

1608 Ridgeway 18.7 Englehardt et al, SAG 2011 speaker notes

1609 Cadia East 20.3 Englehardt et al, SAG 2011 speaker notes

1643 53392-2 13 SGS Lakefield Alacer Gold NI43-101 July 29, 2014

4
Database field definitions
Common fields in several of the database tables include:
• Id – The unique index number of this sample.
• Name – The human-readable sample name.
• Synthetic – Is this a `fake` sample, such as a mathematical average of actual test results?
• Laboratory – The laboratory where a particular test determination was performed.
• Comment – The document reference where the data originated or other comments.
The `summary` database table includes the following fields:
• Program – The project or mine this sample belongs to. Some samples belong to a `other`
group as they do not have many related samples.
• WiBM – The Bond ball mill work index, in metric units.
• WiRM – The Bond rod mill work index, in metric units.
• WiC – The Bond impact crushing work index, in metric units.
• Density – The coarse particle density measured in either the Bond impact crushing work
index test or the drop weight test, t/m³.
• Axb – The product of the `A` and `b` parameters from a drop weight test, unitless.
• Mia – The coarse tumbling particle coefficient for a Morrell power equation, kWh/t.
• Mib – The fine tumbling particle coefficient for a Morrell power equation, kWh/t.
• CI – The Minnovex crushing index determined as part of a SPI™ determination, unitless.
• SGI – The SAG Grindability Index or SAG Power Index™, minutes.
• Ai – The Bond abrasion index, unitless.
The `litho` database table includes the following fields:
• Drillhole – Identifier of the drill hole a sample originated from.
• Dist from – Downhole position where a sample originated from, m.
• Dist to – Downhole position where a sample originated from, m.
• Litho – Lithology identifier for a sample.
• Alteration – Alternation regime identifier for a sample.
• Zone – Zone identifier for a sample.
• Length – Downhole contiguous length of a sample, m
The `ai` database table includes the following field:
• Ai – Bond abrasion index, unitless.
The `dwt` (drop weight test) table includes the following fields:
• A – The fitted coefficient of a “t10 versus Ecs” curve in a drop weight test.
• b – The fitted exponent of a “t10 versus Ecs” curve in a drop weight test.
• Axb – The product of the fitted A and b parameters in a drop weight test.
• ta – The abrasion resistance measurement of a JK DWT™.

5
 • DWT density – The coarse particle density determined in a drop weight test, t/m³.
• SMC – Boolean field indicating of this is a SMC Test™ result (value = 1 for SMC).
• DWI – The Drop Weight Index determination for a sample, kWh/m³
• Mia – The coarse tumbling particle coefficient for a Morrell power equation, kWh/t.
• Mih – The high pressure grinding roll coefficient for a Morrell power equation, kWh/t.
• Mic – The crushing coefficient for a Morrell power equation, kWh/t.
The `pli` (point load index) table includes the following fields:
• # Specimens – The quantity of specimens tested for a particular sample.
• PLI – The mean IS50 value of a set of specimens, MPa.
• PLI Min – The minimum IS50 value of a set of specimens, MPa.
• PLI Max – The maximum IS50 value of a set of specimens, MPa.
• Std Dev – The standard deviation of IS50 values in a set of specimens, MPa.
The `sgi` (SAG grindability index) table includes the following fields:
• CI – The Minnovex crushing index determined as part of a SPI™ determination, unitless.
• SGI – The SAG Grindability Index or SAG Power Index™, minutes.
The `ucs` (unconfined or uniaxial compressive strength) table includes the following fields:
• # Specimens – The quantity of specimens tested for a particular sample.
• UCS – The mean unconfined pressure of sample failure of a set of specimens, MPa.
• UCS Min – The minimum pressure of sample failure of a set of specimens, MPa.
• UCS Max – The maximum pressure of sample failure of a set of specimens, MPa.
• Std Dev – The standard deviation of pressure of sample failure in a set of specimens, MPa.
The `wibm` (ball mill work index) table includes the following fields:
• WiBM µclosing – The closing screen size used in the test, µm.
• WiBM f80 – The test feed 80% passing particle size, µm.
• WiBM p80 – The test product 80% passing particle size, µm.
• WiBM gpr – The test average grams per revolution of the final cycles, g/rev.
• WiBM – The Bond ball mill work index determination, metric units.
• Mod BWI – Boolean field indicating if this is a non-standard test, such as an open-cycle
“modified BWI” test or a SAGDesign test with non-standard size distribution of the feed.
The `wic` (crushing work index or low energy impact work index) table includes the following
fields:
• # Specimens – The quantity of specimens tested for a particular sample.
• WiC – The mean crushing work index of a set of specimens, metric units.
• WiC Min – The minimum crushing work index of a set of specimens, metric units.
• WiC Max – The maximum crushing work index of a set of specimens, metric units.
• Std Dev – The standard deviation of crushing work index of a set of specimens, metric
units.

6
 • WiC density – The coarse particle density measured in a crushing work index test, t/m³.
The `wirm` (rod mill work index) table includes the following fields:
• WiRM µclosing – The closing screen size used in the test, µm.
• WiRM f80 – The test feed 80% passing particle size, µm.
• WiRM p80 – The test product 80% passing particle size, µm.
• WiRM gpr – The test average grams per revolution of the final cycles, g/rev.
• WiRM – The Bond rod mill work index determination, metric units.
The rod mill work index table includes results from laboratories whose apparatus does not conform
to Bond's original specification. The most significant deviation is several laboratories in Australia
use a mill without a wave liner – the liner specified by Bond.

RESULTS AND DISCUSSION

The entire database is too large to tabulate in this document (it would be over 200 pages), so instead
it is freely available for download as an OpenDocument spreadsheet on the author's website at this
link: https://www.sagmilling.com/articles/28/view/?s=1 . The database will be updated
periodically and the latest revision will always be available at the web link.
Comparisons and regression equations between different grindability metrics appear in Figures 1
through 9. The comparisons only consider tests done in the same size classes as defined by Doll &
Barratt, 2009. Linear, logarithmic, exponential and power regression relationships are attempted on
all plotted pairs and the relationship with the highest R² value is displayed.
Fine size class: 30

Morrell Mib versus Bond

25 f(x) = 1.49 x^0.78
ball mill work index. R² = 0.91
Variation expected due to 20
WiBM

different exponents in the

15
two equations.
10

0
0 5 10 15 20 25 30 35 40
Mib
Figure 1 Bond ball mill work index v. Morrell Mib

7
No relationship between 25

Bond abrasion index and

ball mill work index. 20

WiBM
15

f(x) = 3.88x + 13.8

10
R² = 0.06

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6
Ai
Figure 2 Bond ball mill work index v. Bond abrasion index

Medium size class: 100

f(x) = 83.41 x^-0.42 f(x) = 108 x^-0.55
Best relationship requires R² = 0.46 R² = 0.66

separating the Bond-type

WiRM

mills with wave liners from

10
the non-standard mills.
Nonstandard
Power (Nonstandard)
Bond-type
Power (Bond-type)

1
10 100 1000
Axb
Figure 3 Bond rod mill work index v. Drop weight A×b

Generally a good 10000

relationship between the two f(x) = 8882.72 x^-1.24
R² = 0.83
parameters. 1000
SGI

100

10

1
10 100 1000
Axb
Figure 4 SAG Grindability Index v. Drop weight A×b

8
Compares the Mia 100
parameter determined from f(x) = 445.76 x^-0.85
R² = 0.99
the SMC test™ used in
Morrell power equations to

Mia
the A×b parameter reported
10
from any drop weight test
(both SMC test™ and the JK
DWT are drop weight tests).

1
10 100 1000
Axb
Figure 5 Morrell Mia v. Drop weight A×b

DWI is a volumetric 10

parameter (kWh/m³) f(x) = 91.21 x^-0.97

R² = 0.99
determined from the SMC
DWI/density

test™. Dividing DWI by the

sample density gives a mass 1
parameter (kWh/t) suitable
for comparing to A×b.

0.1
10 100 1000
Axb
Figure 6 (Morrell DWI ÷ density) v. Drop weight A×b

Only regression for Bond- 100

type apparatus shown. f(x) = 1.41 x^0.5

R² = 0.89

Too few examples of

WiRM

samples tested on both the

10
non-standard rod mill
apparatus and the SAG
Nonstandard
Grindability Index. Bond-type
Power (Bond-type)

1
1 10 100 1000
SGI
Figure 7 Rod mill work index v. SAG Grindability Index

9
Morrell's crushing parameter 100
Mic is (perfectly?) related to f(x) = 576.99 x^-1
R² = 1
the drop weight test A×b
determination.

Mic
10

1
10 100
Axb
Figure 8 Morrell Mic index v. Drop weight test A×b

Coarse size class: 100

Noisy data with poor overall f(x) = 2.9 x^0.31

R² = 0.38
relationship.
WiC

10

1
0.1 1 10 100 1000
UCS
Figure 9 Crushing work index v. Unconfined compressive strength

Specific energy predictions

Some of the samples that include all the parameters for Bond work indices, SMC tests and SGI
values were run against different power-based specific energy models using SAGMILLING.COM
software. The samples were run in a circuit consisting of mills based on Los Bronces Confluencia
operating in an SABC-B configuration grinding from feed F80 of 150 mm to a cyclone overflow
product P80 of 180 µm. No attempt to optimize any of the simulations was done.

10
Table 4 Example specific energy predictions by three models

Test results or predictions Model Etotal, kWh/t

Optimized Morrell
Name WiBM WiRM WiC A×b SGI Mic Mia Mib SGI
Bond/Barratt Mi
Los Bronces
16.5 16.7 10.0 30.0 130.0 19.2 24.8 20.5 15.3 19.1 15.2
median
Antapaccay
17.6 13.5 5.6 47.1 79.4 12.3 16.9 12.3 15.0 12.6 14.5
UGM 1
Antapaccay
16.7 13.5 5.7 54.8 74.5 10.5 14.8 10.5 14.2 11.0 13.8
UGM 2
Antapaccay
15.5 11.6 5.2 44.1 72.7 13.1 17.8 13.1 12.5 13.4 13.0
UGM 3
Antapaccay
14.7 12.1 7.9 53.1 72.4 10.9 15.2 10.9 12.1 11.3 12.5
UGM 4
Antapaccay
10.0 10.8 7.3 64.9 61.0 8.9 12.8 8.9 8.1 9.4 10.0
UGM 5
Antapaccay
14.6 14.3 8.2 47.5 90.8 12.2 16.8 12.2 12.6 12.6 12.9
UGM 6
Boddington Apr
14.4 20.0 27.7 30.0 130.9 19.2 24.7 17.7 16.2 18.2 15.2
2010 survey
Volta Grande
16.5 16.0 20.0 30.6 133.2 18.9 24.3 18.9 15.9 18.5 15.3
comp 2-3
Huckleberry
18.0 16.3 7.0 31.1 125.2 18.6 24.0 25.6 16.5 20.2 15.8
SAG feed 2012
Malartic Sep 21
16.3 16.3 18.5 24.1 115.3 23.9 29.8 23.9 13.4 23.0 14.5
survey
Yanacocha 17.5 13.8 10.0 72.9 43.5 7.9 12.3 25.3 15.4 14.0 12.9
Meadowbank
13.9 15.9 10.0 40.9 86.4 14.1 19.0 14.1 13.2 14.3 12.3
Vault
Corani avg 15.1 10.2 6.3 111.0 35.6 5.2 8.1 5.2 11.8 5.8 11.2
Italics indicate test parameters based on interpolations (Figures Figure through Figure) or assumed.

This particular sub-set of the overall database shows the Optimized Bond/Barratt model and the
Amelunxen SGI model are generally very close in their specific energy consumption predictions
(average absolute difference is 6.4%), but the Morrell Mi model is substantially different from the
other two models (average absolute difference of 20.4% versus Bond/Barratt and 20.6% versus SGI).

11
The Author's experience is that any ore sample can potentially confuse any grindability test, so the
observation that the Morrell Mi model is substantially different from the other two is likely an
artifact of the sub-set of results that were chosen for Table. One might observe a different pattern
had a different set of samples been chosen for the specific energy consumption calculations.

25
Etotal, kWh/t to 180 µm

20

15

10

Optimized Bond/Barrat t Morrell Mi SGI

Figure 10 Specific energy consumption predictions for three models

CONCLUSIONS
A large quantity of grindability test results have been published in conference proceedings, NI43—
101 reports and other works. The author has collected and collated over 800 examples of such
published grindability results and generated a public database of test results suitable for
benchmarking other projects or performing research such as extracting relationships between the
different test parameters.
The database is freely available for download as an OpenDocument spreadsheet on the author's
website at this link: https://www.sagmilling.com/articles/28/view/?s=1

12
ACKNOWLEDGEMENTS

SPI™ and SAG Power Index™ are trademarks of SGS Mineral Services.
SMC Test™ is a trademark of SMC Testing Pty Ltd.

REFERENCES

Alruiz, O., Camacho, J., Suazo, C., Rojas, O. and Hofmann, A. (2008) A Treatment Capacity Model:
the Approach Developed by Compañía Minera Doña Inés de Collahuasi SCM. In A. Casali, C.
Gómez and R. Kuyvenhoven (Eds.) Proceedings of the V International Mineral Processing Seminar
(paper № C5-A6, pp. 277 – 299) Santiago, Chile.
Angove, J.E. and Dunne, R.C. (1997) A Review of Standard Physical Ore Property Determinations.
Proceedings of World Gold '97 Conference (pp. 139 – 144) Singapore.
Ashanti website, http://www.anglogoldashanti.com . Retrieved 2012-12-01.
Ballantyne, G., Powell, M., Clarke, N., Di Trento, M., Kock, F. and Putland, B. (2016) Introduction to
energy curves and AngloGold Ashanti case studies. Presentation to the 2016 Annual Meeting of the
Society of Mining Engineers, Phoenix, USA.
Barns, K., Lane, G., Osten, K. and Scagliotta, N. (2004) Benchmarking Energy Efficiency — A Case
Study at Macraes Gold Mine. Proceeding of the AusIMM MetPlant 2004 Conference (pp. 221 – 259)
Perth, Australia.
Barratt, D.J. (1992) Design of Pebble Crushing Circuit. Proceedings of the SME Annual Meeting and
Exhibit (preprint № 92-241) Phoenix, USA.
Barratt, D.J. and Allan, M.J. (1986) Testing for autogenous and semiautogenous grinding: A
designer's point of view. Minerals and Metallurgical Processing, (paper № 86-608, pp. 65 – 74)
Society of Mining Engineers.
Becerra, M. and Amelunxen, P. (2012) A comparative analysis of grinding circuit design
methodologies. In A. Doll, W. Kracht and R. Kuyvenhoven (Eds.) Proceedings of the 9th
International Conference on Mineral Processing (pp. 468 – 476) Santiago, Chile.
Bennett, D., Tordoir, A., Walker, P., La Rosa D., Valery W. and Duffy K. (2014) Throughput
Forecasting and Optimisation at the Phu Kham Copper-Gold Operation. Proceedings of the 12th
AusIMM Mill Operators' Conference (pp. 381 – 390) Townsville, Australia.
Bigg, A.C.T., MacPherson, A.R. and Colombo, A.F. (1992) The development of a fully autogenous
grinding circuit for the Curragh Resources – Stronsay project. In Comminution, Theory and
Practice, S.K. Kawatra (Ed.), Society of Mining Engineers (p. 319) ISBN: 0-87335-112-6.
Burger, B., McCaffery, K., McGaffin, I., Jankovic, A., Valery, W. and La Rosa, D. (2006) Batu Hijau
Model for Throughput Forecast, Mining and Milling Optimization, and Expansion Studies. In
Advances in Comminution, (ed. S. Komar Kawatra) Society for Mining, Metallurgy and
Exploration, Inc. (pp. 461 – 479) ISBN: *978-0-87335-246-8.
Burger, B., Vargas, L., Arevalo, H., Vicuña, S., Seidel, J., Valery, W., Jankovic, A., Valley, R. and
Nozawa, E. (2011) Yanacocha Gold Single Stage SAG Mill Design, Operation & Optimization. In K.
Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.) Proceedings of an International Conference on

13
Autogenous Grinding, Semiautogenous Grinding and High Pressure Grinding Roll Technology
(paper № 127) Vancouver, Canada.
Cole, J., Seidel, J. and Orlich, J. (2006) The Design, Start-Up And Operation Of The Phoenix Project
SAG Mill. In M. J. Allan, K. Major, B. C. Flintoff, B. Klein and A. L. Mular (Eds.) Proceedings of an
International Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-79 – I-
87) Vancouver, Canada.
Doll, A.G. and Amelunxen, P. (2013) Power-based grinding models. Short course presented at the
10th International Conference on Mineral Processing, Santiago, Chile.
Doll, A.G. and Barratt, D.J. (2009) Case studies on the effect of sample dimensions on comminution
testwork results. In P. Amelunxen, W. Kracht & R. Kuyvenhoven (Eds.), Proceedings of the VI
International Mineral Processing Seminar (pp. 99–110) Santiago, Chile.
van Drunick, W., Gerold, C. and Palm, N. (2010) Implementation Of An Energy Efficient Dry
Grinding Technology Into An Anglo American Zinc Beneficiation Process. Proceedings of the XXV
International Mineral Processing Congress (IMPC) (pp. 1333 – 1341) Brisbane, Australia.
Dunne, R., Chittenden, R., Lane, G. and Morrell, S. (1999) The Cadia Gold Copper Project
Exploration to Start up. Proceedings of the SME Annual Meeting (Preprint № 99-160) Denver,
USA.
Dunne, R., Morrell, S. and Lane, G. (2000) Design of the Biggest Mill in the World. Proceedings of
the AusIMM After 2000 – The Future of Mining (pp. 179 – 185) Sydney, Australia.
Dunne, R., Morrell, S., Lane, G., Valery, W. and Hart, S. (2001) Design of the 40 Foot SAG Mill
Installed at the Cadia Gold Copper Mine. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.),
Proceedings of an International Conference on Autogenous and Semiautogenous Grinding
Technology (pp. I-43 – I-58) Vancouver, Canada.
Duval, L. and Wood, K. (1989) Testing, Design and Operation of SAG Circuit at Les Mines Selbaie.
In A. Mular and G. Agar (Eds.) Proceedings of Advances In Autogenous and Semiautogenous
Grinding Technology (pp. 345 – 356) Vancouver, Canada.
Ehm, G. and Hill, G. (1992) The Design, Construction and Operation of the Boddington Gold Mine
Supergene-Basement Plant. Extractive Metallurgy of Gold and Base Metals: Proceedings of
the International Conference on Extractive Metallurgy of Gold and Base Metals (pp. 79 – 94)
Kalgoorlie, Australia.
Engelhardt, D., Robertson, J., Lane, G., Powell, M.S. and Griffin, P. (2011) Cadia Expansion – From
Open Pit to Block Cave and Beyond. In K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.)
Proceedings of an International Conference on Autogenous Grinding, Semiautogenous Grinding
and High Pressure Grinding Roll Technology (paper № 121) Vancouver, Canada.
Errey, J. (2009) Process Selection and Design for the Palmarejo Silver Mine. Proceedings of the
AusIMM Tenth Mill Operators' Conference (pp. 33 – 41) Adelaide, Australia.
Faria, E. and Latchireddi, S. (2011) Commissioning and Operation of Milling Circuit at Santa Rita
Nickel Operation. In K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.) Proceedings of an
International Conference on Autogenous Grinding, Semiautogenous Grinding and High Pressure
Grinding Roll Technology (paper № 137) Vancouver, Canada.

14
Fragomeni, D., Boyd, L., Charland, A., Kormos, L., Lotter, N.O. and Potts, G. (2005) The Use of End
Members for Grind – Recovery Modeling, Tonnage Prediction and Flowsheet Development at
Raglan. Proceedings of the 37th Annual Meeting of the Canadian Mineral Processors (pp. 75 – 98)
Ottawa, Canada.
Frostiak, J., Gonzales, I., Thomas, K. and Manning, K. (2000) The Pierina Project: From Acquisition
to Production. Proceedings of the 32nd Annual Meeting of the Canadian Mineral Processors (pp.
279 – 294) Ottawa, Canada.
García, D. and Villanueva, J. (2013) Antapaccay: Tintaya expansion. In M. Álvarez, A. Doll, W.
Kracht and R. Kuyvenhoven (Eds.) Proceedings of the 10th International Mineral Processing
Conference (paper № 102) Santiago, Chile.
Hadaway, J.B. and Bennett, D.W. (2011) An Overview of the Design, Construction Commissioning
and Early Years of the SAG/Ball Grinding Circuit at Phu Kam Copper Gold Operations in Laos. In
K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.) Proceedings of an International Conference
on Autogenous Grinding, Semiautogenous Grinding and High Pressure Grinding Roll Technology
(paper № 144) Vancouver, Canada.
Hall, D. and Toscano, P. (2016) Development of the Rainy River Gold Project and Processing Plant.
Proceedings of the 48th Annual Canadian Mineral Processors Operators Conference (pp. 67 – 81)
Ottawa, Canada.
Hart, S., Parker, B., Rees, T., Manesh, A. and McGaffin, I. (2011) Commissioning and Ramp Up of
the HPGR Circuit at Newmont Boddington Gold. In K. Major, B.C. Flintoff, B. Klein and K.
McLeod (Eds.) Proceedings of an International Conference on Autogenous Grinding,
Semiautogenous Grinding and High Pressure Grinding Roll Technology (paper № 41) Vancouver,
Canada.
Hart, S., Valery, W., Clements, B., Reed, M., Song, M. and Dunne, R. (2001) Optimisation of the
Cadia Hill SAG Mill Circuit. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an
International Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-11 – I-
30) Vancouver, Canada.
Kock, F., Siddall, L., Lovatt, I.A., Giddy, M. and DiTrento, M (2015) Rapid Ramp-up of the
Tropicana HPGR Circuit. In B. Klein, K. McLeod, R. Roufail and F. Wang (Eds.) Proceedings of an
International Conference on Autogenous Grinding, Semi-autogenous Grinding and High Pressure
Grinding Roll Technology (paper № 72) Vancouver, Canada.
Lane, G. (2015) The Comminution Circuit Design for the Constancia Project. In B. Klein, K.
McLeod, R. Roufail and F. Wang (Eds.) Proceedings of an International Conference on Autogenous
Grinding, Semi-autogenous Grinding and High Pressure Grinding Roll Technology (paper № 39)
Vancouver, Canada.
Lane, G., Bennett, C. and Villanueva, C. (2001) Cerro Corona Grinding Circuit Design. In D.J.
Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an International Conference on Autogenous
and Semiautogenous Grinding Technology (pp. III-137 – III-150) Vancouver, Canada.
Lane, G., Foggiatto, B. and Bueno, M. (2013) Power-based comminution calculations using
Ausgrind. In M. Álvarez, A. Doll, W. Kracht and R. Kuyvenhoven (Eds.) Proceedings of the 10th
International Mineral Processing Conference (paper № 202) Santiago, Chile.

15
Larsen, C., Cooper, M. and Trusiak, A. (2001) Design and Operation of Brunswick’s Ag/SAG
Circuit. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an International Conference
on Autogenous and Semiautogenous Grinding Technology (pp. IV-350 – IV-367) Vancouver,
Canada.
Lunt, D.J., Thompson, A. and Ritchie, I. (1996) The design and operation of the Kanowna Belle
milling circuit. In A. Mular, D.J. Barratt & D.A. Knight (Eds.), Proceedings of an International
Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-81 – I-96) Vancouver,
Canada.
Magnusson, R., Hollow, J., Mosher, J. and Major, K. (2001) The Fort Knox mill: Design,
commissioning and operation. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an
International Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-159 – I-
173) Vancouver, Canada.
McGhee, S., Mosher, J., Richardson, M., David, D. and Morrison, R. (2001) SAG Feed Pre-Crushing
at ASARCO’s Ray Concentrator: Development, Implementation and Evaluation. In D.J. Barratt,
M.J. Allan & A.L Mular (Eds.), Proceedings of an International Conference on Autogenous and
Semiautogenous Grinding Technology (pp. I-234 – I-247) Vancouver, Canada.
McLaren, D., Mitchell, J., Seidel, J. and Lansdown, G. (2001) The design, startup and operation of
the Batu Hijau concentrator. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an
International Conference on Autogenous and Semiautogenous Grinding Technology (pp. IV-316 –
IV-335) Vancouver, Canada.
McLean, E. and Watt, J. (2009) Processing Strategies for Hidden Valley Operations. Proceedings of
the AusIMM Tenth Mill Operators' Conference (pp. 43 – 52) Adelaide, Australia.
Millard, M. (2002) The Use of Comminution Testwork Results in SAG Mill Design. Proceedings of
AusIMM Metallurgical Plant Design and Operating Strategies Conference (pp. 56 – 71) Sydney,
Australia.
Muteb, P.N. and Fortin, M. (2015) Meadowbank Mine Process Plant Throughput Increase. In B.
Klein, K. McLeod, R. Roufail and F. Wang (Eds.) Proceedings of an International Conference on
Autogenous Grinding, Semi-autogenous Grinding and High Pressure Grinding Roll Technology
(paper № 72) Vancouver, Canada.
Nelson, M., Valery, W. and Morrell, S. (1996) Performance characteristics and optimisation of the
Fimiston (KCGM) SAG mill circuit. In A. Mular, D.J. Barratt & D.A. Knight (Eds.), Proceedings of
an International Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-233 –
I-248) Vancouver, Canada.
Oestreicher, C. and Spollen, C. (2006) HPGR versus SAG Mill Selection For The Los Bronces
Grinding Circuit Expansion. In M. J. Allan, K. Major, B. C. Flintoff, B. Klein and A. L. Mular (Eds.)
Proceedings of an International Conference on Autogenous and Semiautogenous Grinding
Technology (pp. IV-110 – IV-123) Vancouver, Canada.
Parker, B., Rowe, P., Lane, G. and Morrell, S. (2001) The Decision to Opt for High Pressure Grinding
Rolls for the Boddington Expansion. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of
an International Conference on Autogenous and Semiautogenous Grinding Technology (pp. III-93 –
III-106) Vancouver, Canada.

16
Rasmussen, G. (1998) Huckleberry Mines Ltd. Process Design and Project Startup. Proceedings of
the 30th Annual Operators Conference of the Canadian Mineral Processors (pp. 323 – 336) Ottawa,
Canada.
Segura, J. and Tavares, L.M. (2014) Comparing comminution routes for a Brazilian iron ore using
sustainability principles. In J. Yianatos, A. Doll, C. Gomez and R. Kuyvenhoven (Eds.) Proceedings
of XXVII International Mineral Processing Congress (paper № 9-12, pp. 109 – 121) Santiago, Chile.
Seidel, J.F., Idzal, D., Colombo, A.F. and Albert, T.E. (1996) Grinding circuit for Bibiani Gold Project
in Ghana. In A. Mular, D.J. Barratt & D.A. Knight (Eds.), Proceedings of an International
Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-313 – I-328)
Vancouver, Canada.
Starkey, J. and Dobby, G. (1996) Application of the Minnovex SAG Power Index at five Canadian
SAG plants. In A. Mular, D.J. Barratt & D.A. Knight (Eds.) Proceedings of an International
Conference on Autogenous and Semiautogenous Grinding Technology (pp. I-345 – I-360)
Vancouver, Canada.
Starkey, J. and Holmes, G. (2001) Grinding Circuit Design at Kubaka using SPI and Bond Testing.
Proceedings of the 33rd Annual Operators Conference of the Canadian Mineral Processors (pp. 119 –
134) Ottawa, Canada.
Starkey, J. and Meadows, D. (2007) Comparison of Ore Hardness Measurements for Grinding Mill
Design for the Tenke Project. Proceedings of the 39th Annual Operators Conference of the Canadian
Mineral Processors (pp. 19 – 31) Ottawa, Canada.
Starkey, J., Robitaille, J., Cousin, P., Jordan, J. and Kosick, G. (2001) Design of the Agnico-Eagle
Laronde Division SAG mill. In D.J. Barratt, M.J. Allan & A.L Mular (Eds.), Proceedings of an
International Conference on Autogenous and Semiautogenous Grinding Technology (pp. III-165 –
III-178) Vancouver, Canada.
Strickland, D., Hall, D. and Kresin, P. (1999) Gold Ore Processing at the Musselwhite Mine -
Feasibility, Design, and Operation. Proceedings of the 31st Annual Operators Conference of the
Canadian Mineral Processors (pp. 311 – 330) Ottawa, Canada.
Strohmayr, S. and Valery, W. (2001) SAG Mill Circuit Optimisation at Ernest Henry Mining. In D.J.
Barratt, M.J. Allan & A.L Mular (Eds.) Proceedings of an International Conference on Autogenous
and Semiautogenous Grinding Technology (pp. III-11 – III-42) Vancouver, Canada.
Tew, A., Harvey, R., Marun, J., Anderson, R. and Strohmayr, S. (2003) Evolution of SAG Milling
Performance — A Comparison of Alumbrera and Ernest Henry. Proceedings of the AusIMM
Eighth Mill Operators' Conference (pp. 27 – 37) Townsville, Australia.
Torrealba-Vargas, J., Dupont, J.-F., McMullen, J., Allaire, A. and Welyhorsky, R. (2015) The
Successful Development of the Detour Lake Grinding Circuit: From Testwork to Production. In B.
Klein, K. McLeod, R. Roufail and F. Wang (Eds.) Proceedings of an International Conference on
Autogenous Grinding, Semi-autogenous Grinding and High Pressure Grinding Roll Technology
(paper № 38) Vancouver, Canada.
Wang, C., Nadolski, S., Mejia, O., Drozdiak, J. and Klein, B. (2013) Energy and Cost Comparisons of
HPGR Based Circuits with the SABC Circuit Installed at the Huckleberry Mine. Proceedings of the
45th Annual Operators Conference of the Canadian Mineral Processors (pp. 121 – 136) Ottawa,
Canada.

17
Weidenbach, M. (2009) Angas Zinc Mine – Start-Up and the First Twelve Months of Operation in
the Adelaide Hills. Proceedings of the AusIMM Tenth Mill Operators’ Conference (pp. 133 – 140)
Adelaide, Australia.
Weidenbach, M., Triffett, B. and Treloar, C. (2011) Optimisation of the Prominent Hill SAG Mill. In
K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.) Proceedings of an International Conference
on Autogenous Grinding, Semiautogenous Grinding and High Pressure Grinding Roll Technology
(paper № 33) Vancouver, Canada.
Weller, K.R., Campbell, J.J., Wilkie, G.J., Thornber, M.R., Bateman, R. and Ellis, S. (1998) The Ores
of the Golden Mile, Kalgoorlie: Coherent Metallurgical Testwork from Comminution, through
Flotation, Roasting and Leaching, to Mineralogical Determination of Key Mineral Deportments
using QEM*SEM. Proceedings of the AusIMM Annual General Meeting (pp. 339 – 347) Mount Isa,
Australia.
Wickens, J., Deal, M., Spicher, M., Tittes, P. and McDaniel, S. (2013) Haile Gold Mine Metallurgy
and Flowsheet Review. Proceedings of the 45th Annual Operators Conference of the Canadian
Mineral Processors (pp. 251 – 266) Ottawa, Canada.
Wirfiyata, F. and Mccaffery K. (2011) Applied Geo-Metallurgical Characterisation For Life Of Mine
Throughput Prediction At Batu Hijau. In K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.)
Proceedings of an International Conference on Autogenous Grinding, Semiautogenous Grinding
and High Pressure Grinding Roll Technology (paper № 32) Vancouver, Canada.
Wolmarans, E., Morgan, P.J. and Smit, D.S. (2011) Selection and Commissioning of ABC Grinding
Circuit For Nkomati Nickel. In K. Major, B.C. Flintoff, B. Klein and K. McLeod (Eds.) Proceedings
of an International Conference on Autogenous Grinding, Semiautogenous Grinding and High
Pressure Grinding Roll Technology (paper № 9) Vancouver, Canada.

Project reports
The following are project reports that are downloaded from the Canadian Securities Administrators
SEDAR filing system (http://www.sedar.com) unless otherwise noted. The bold text indicates the
property name where the grindability data is indexed in the database.
NI43-101 report on the Ajax project, prepared for KGHM Polska Miedź S.A. and Abacus Mining
and Exploration Corporation 2016-02-19 by M3 Engineering & Technology Corporation, Tucson,
USA.
NI43-101 report on the Albany Graphite project, prepared for Zenyatta Ventures Ltd. 2015-07-09 by
RPA Inc., Toronto, Canada.
NI43-101 report on the Rose deposit, Mills Lake Deposit on the Kamistiatusset Iron Ore property,
parepared for Alderon Iron Ore Corp. 2012-12-17 by BBA Inc., Montréal, Canada.
NI43-101 report on the Asmara project, prepared for Sunridge Gold Corp 2014-03-27 by SENET
(Pty) Ltd., Modderfontein, South Africa.
NI43-101 report on the Aurora Gold project, prepared for Guyana Goldfields, Inc. 2013-01-29 by
Tetra Tech Inc. (undisclosed origin).
NI43-101 report for the Back River gold property, prepared for Sabina Gold & Silver Corp. 2014-03-
04 by Tetra Tech, Vancouver, Canada.

18
NI43-101 report on the Bau project, prepared for Besra Gold Inc. 2013-12-15 by Besra Gold Inc.,
Sarawak, Malaysia.
NI43-101 report on the Blackwater Gold Project, prepared for New Gold Inc. 2014-01-14 by Amec
Americas Inc., Vancouver, Canada.
NI43-101 report on the Brucejack Project, prepared for Pretium Resources Inc. 2013-06-21 by Tetra
Tech, Vancouver, Canada.
NI43-101 report on the Buriticá gold project, prepared for Continental Gold Inc. 2015-08-07 by
Mining Associates Pty Ltd., Spring Hill, Australia.
NI43-101 report on the Cameron Gold project, prepared for Coventry Resources Ltd. and Crescent
Resources Corp. 2012-07-05 by Datageo Geological Consultants, Mullaloo, Australia.
NI43-101 report on the Canadian Malartic property, prepared for Agnico Eagle and Yamana Gold
2014-06-16 by Canadian Malartic GP, Val-d'Or, Canada.
NI43-101 report on the Capoose Gold-Silver project, prepared for New Gold Inc. 2014-05-23 by
GeoSim Services Inc., Vancouver, Canada.
NI43-101 report on the Çöpler Sulfide Expansion Project, prepared for Alacer Gold Corp. 2014-07-
29 by Jacobs Engineering Group, Denver, USA.
NI43-101 report on the Copper Creek Resource, prepared for Redhawk Copper Inc. 2013-07-25 by
SGS Metcon/KD Engineering, Tucson, USA.
NI43-101 report on the Copper Mountain project, prepared for Copper Mountain Mining Corp
2008-09-11 (also referenced in 2009 reports) by KWM Consulting Inc., Pitt Meadows, Canada.
NI43-101 report on the Corani project, prepared for Bear Creek Mining Corporation 2015-05-30 by
M3 Engineering & Technology Corporation, Tucson, USA.
NI43-101 report on the Cosmo Deeps gold project, prepared for Crocodile Gold Corp. 2013-12-31
by Mining Plus Pty Ltd., Melbourne, Australia.
NI43-101 report on the Detour Lake mine, prepared for Detour Gold Corporation 2012-10-18 by
BBA Inc., Montréal, Canada.
NI43-101 report on the Dumont project, prepared for Royal Nickel Corporation 2012-07-18 by
Ausenco Solutions Canada Inc., Vancouver, Canada.
NI43-101 report on the Duncan Lake Iron property, prepared for Century Iron Mines Corp. and
Augyva Mining Resources Inc. 2013-05-06 by Met-Chem Canada Inc., Montréal, Canada.
NI43-101 report on the Duparquet project, prepared for Clifton Star Resources Inc. 2014-03-26 by
InnovExplo, Val-d'Or, Canada.
NI43-101 report on the Fekola gold project, prepared for B2Gold Corporation 2015-06-30 by
Lycopodium Minerals Pty Ltd., Spring Hill, Australia.
NI43-101 report for the Grassy Mountain project, prepared for Calico Resources Corp. 2015-07-09
by Metal Mining Consultants, Inc., Highlands Ranch, USA.
NI43-101 report for the Harper Creek copper project, prepared for Yellowhead Mining Inc. 2011-03-
31 by Wardrop Engineering Inc., Vancouver, Canada.

19
NI43-101 report for the Harper Creek copper project, prepared for Yellowhead Mining Inc. 2012-03-
29 by Merit Consultants International Inc., Vancouver, Canada.
NI43-101 report on the Main Zone Optimization, Huckleberry Mine, prepared for Huckleberry
Mines Ltd. and Imperial Metals Corporation 2011-09-01 by K. Christensen, G.R. Connaughton and
P. Ogryzlo.
NI43-101 report for the Karma Gold Project, prepared for True Gold Mining Inc. 2014-10-10 by P&E
Mining Consultants Inc., Brampton, Canada.
NI43-101 report for the Krumovgrad project, prepared for Dundee Precious Metals Inc. 2014-03-21
by CSA Global (UK) Ltd. Horsham, UK.
NI43-101 report on the Schefferville Area Report: Schefferville Area Direct Shipping Iron Ore
Projects Resource Update In Western Labrador And North Eastern Quebec, Canada For Labrador
Iron Mines Holdings Limited. 2013-04-12 by SGS Canada Inc.
NI43-101 report on the Pre-Feasibility Study Update, Lac À Paul Apatite Project for Ressources
d'Arianne Inc. 2012-07-13 by Met-Chem Canada Inc., Montréal, Canada.
NI43-101 report on the Lik deposit, prepared for Zazu Metals Corporation 2014-03-03 by JDS
Energy & Mining Inc., Vancouver, Canada.
NI43-101 report on the Los Helados property, prepared for NGEx Resources Inc. 2013-10-31 by
Behre Dolbear International Ltd., Santiago, Chile.
NI43-101 report on the Lucky Friday Mine, prepared for Hecla Mining Company 2014-04-02 by
Hecla Mining Company, Coeur d'Alene, USA.
NI43-101 report for the Magino Project, prepared for Argonaut Gold Inc. 2013-12-17 by JDS Energy
& Mining Inc., Vancouver, Canada.
NI43-101 report on the Maracás Vanadium Project, prepared for Largo Resources Ltd. 2013-03-04
by Runge Pincock Minarco (Canada) Ltd., Toronto, Canada.
NI43-101 report on the Masbate Gold Project, prepared for CGA Mining Limited 2011-10-31 by
M.B. Turner, A.J. Vigar and S.T. Jones.
NI43-101 report on the Mercedes gold-silver project, prepared for Yamana Gold Inc. 2014-02-25 by
RPA Inc., Toronto, Canada.
NI43-101 report on the Metates Gold-Silver Project, prepared for Chesapeake Gold Corp. 2013-03-
13 by M3 Engineering & Technology Corporation, Tucson, USA.
NI43-101 report on the Montagne d'Or gold deposit, Paul Isnard Project, prepared for Nord Gold
N.V. and Columbus Gold Corporation 2015-07-31 by SRK Consulting (U.S.) Inc., Lakewood, USA.
NI43-101 report on the Moose River Consolidated Project, prepared for Atlantic Gold Corporation
2015-08-13 by Ausenco Engineering Canada Inc., Vancouver, Canada.
NI43-101 report on the Mount Milligan project, prepared for Terrane Metals Corp 2009-10-23 by
Wardrop Engineering Inc., Vancouver, Canada.
NI43-101 report on the Namoya Gold project, prepared for Banro Corporation 2013-12-31 by
Venmyn Deloitte (Pty) Ltd., Woodmead, South Africa.

20
NI43-101 report on the New Afton project, prepared for New Gold Inc. 2009-12-31 by Scott Wilson
Roscoe Postle Associates Inc., Toronto, Canada.
NI43-101 report on the New Liberty Gold Project, prepared for Aureus Mining Inc. 2012-10-22 by
AMC Consultants (UK) Limited, Maidenhead, UK.
NI43-101 report on the Nico gold-cobalt-bismuth-copper deposit, prepared for Fortune Minerals
Limited 2012-07-02 by P&E Mining Consultants Inc., Brampton, Canada.
NI43-101 report on Niobec Expansion, prepared for Iamgold Corporation 2013-12-10 by Iamgold
Corp., Longueuil, Canada and Niobec Inc., St-Honoré, Canada.
NI43-101 report on the Nkout Iron Project, prepared for Afferro Mining 2013-05-28 by SRK
Consulting (UK) Limited, Cardiff, UK.
NI43-101 report on the Palmarejo Project, prepared for Coeur Mining 2014-12-31 by Coeur Mining
Inc.
NI43-101 report on the Pebble project, prepared for Northern Dynasty Minerals Ltd. 2014-12-31 by
Hunter Dickinson Inc., Vancouver, Canada.
NI43-101 report on the Peñasquito Project, prepared for Goldcorp Inc. 2007-12-31 by F.H Brown,
R.H. Bryson, R. Rivera and S. Ristorcelli.
NI43-101 report on the Platreef project, prepared for Ivanhoe Mines Ltd. 2014-03 by OreWin Pty
Ltd., Adelaide, Australia.
NI43-101 report on the Quimsacocha Gold Project, prepared for Iamgold Corporation 2009-02 by
Iamgold Technical Services, Longueuil, Canada.
NI43-101 report on the Rainy River project, prepared for Rainy River Resources Ltd. 2012-06-04 by
SRK Consulting (Canada) Inc. Toronto, Canada.
NI43-101 report on the Rainy River project, prepared for New Gold Inc. 2014-01-16 by BBA Inc.,
Montréal, Canada.
NI43-101 report on the Sabodala gold project, prepared for Teranga Gold Corporation 2014-03-13
by AMC Consultants (Canada) Ltd., Toronto, Canada.
NI43-101 report on the San Bartolomé project, prepared for Coeur d’Alene Mines Corporation
2012-01-01 by D.J. Birak and K.R. Blair.
NI43-101 report on the Snelgrove Lake property, prepared for Mamba Minerals Limited 2013-12-20
by King and Bay West Management Corp. and BBA Inc., Montréal, Canada.
NI43-101 report on the Spanish Mountain project, prepared for Spanish Mountain Gold Ltd. 2010-
12-20 by AGP Mining Consultants Inc., Barrie, Canada.
NI43-101 report on the Tepal project, prepared for Geologix Explorations Inc. 2013-04-30 by JDS
Energy & Mining Inc., Vancouver, Canada.
NI43-101 report on the Tulsequah Chief deposit, prepared for Chieftain Metals Inc. 2010-11-08 by
SRK Consulting (Canada) Inc., Vancouver, Canada.
NI43-101 report for the Turnagain project, prepared for Hard Creek Nickel Corporation 2011-12-02
by AMC Mining Consultants (Canada) Ltd., Vancouver, Canada.

21
NI43-101 report on the Twangiza gold mine, prepared for Banro Corporation 2015-07-29 by SRK
Consulting (UK) Limited, Cardiff, UK.
NI43-101 report on the Volta Grande project, prepared for Belo Sun Mining Corp. 2014-03-31 by
AGP Mining Consultants Inc., Barrie, Canada.
NI43-101 report on the White Foil project, prepared for La Mancha Resources Inc. 2009-02 by La
Mancha Resources Inc. Montréal, Canada.
NI43-101 report on the Yaramoko gold project, prepared for Roxgold Inc. 2014-06-04 by SRK
Consulting (Canada) Inc., Toronto, Canada.
NI43-101 report on the Zafranal project, prepared for AQM Copper Inc. 2013-01-16 by Tetra Tech
Wardrop, Vancouver, Canada.

22
Estimation of Load Level in Industrial SAG Mills
Using the onMill System
Hernán Robles1, Tania Ledezma1, Luis Vinnett2 and Jorge Pontt2
1. Netos, Chile
2. Automation and Supervision Centre for Mining Industry, Universidad Técnica Federico Santa
María, Chile

ABSTRACT
In the last years, the variation of metal prices has influenced significantly on the economic balance
of the mining industry. This condition has imposed constraints in terms of the operation and
control of mineral processing plants to improve the overall productivity and profitability. On this
subject, the semi-autogenous grinding (SAG) operation, one of the most energy intensive stage in
mineral concentrators, has been under constantly evaluation to decrease its impact on the total
plant costs.
Total load and ball load have shown to have a significant influence on the SAG mill performance.
However, real-time measurements of these variables at industrial scale are not currently
straightforward or their reliabilities are limited. In this paper, a new approach (onMill system) for
load level estimation in industrial SAG mills is presented. The onMill system uses new
instrumentation to determine the position of the load along with either phenomenological or
empirical models for the bearing pressure, power draw and liner wear. The toe and shoulder
positions were determined by vibration analysis. These vibrations were recorded by a sensor
attached to the outer shell of the SAG mill. Estimations for load level (JC) and ball level (JB) were
obtained in one of the largest industrial SAG mill operations. Results delivered by the onMill
system, including vibration analysis, showed a lower variability regarding the load level estimation
obtained from the combined method based on bearing pressure and power draw. In addition, the
load position measurements (toe and shoulder) were in good agreement with disturbances in
operational variables such as the feed tonnage and mill rotation speed.

1
INTRODUCTION
Comminution contributes to about 40–50% of the total operating costs, and more than 60% of the
total energy consumption in mineral processing (Wenzheng, 1991). Despite the costs vary from
mine to mine, grinding may contribute about 90% of total operating costs of comminution (Wills
and Tim, 2011). Thus, grinding stage generates economic impact and high energy consumption on
the productive process, therefore it is clear the need to implement solutions that enable lower
production costs while maintaining quality and safety standards.
A strategy to deal with this scenario has been the process optimization by means of automatic
control of key variables. Expert, model predictive controllers and fuzzy logic have been some of the
process control systems most commonly used in the mining industry. These systems have delivered
suitable operational performance and benefits for the grinding circuits such as throughput increase
with variability decrease, power consumption reduction, mill load variability decrease, among
others (Bouffard, 2015; Wei and Craig, 2009; Bauer and Craig; 2008).
The required information to implement control strategies is not always available, especially in
complex systems such as grinding. For this reason, significant efforts have been focused on
implementing new instrumentation or soft sensors capable of monitoring the behaviour of the mill
and the load level. In the semi-autogenous grinding (SAG), the following approaches and
developments are highlighted: (i) monitoring of electrical variables along with vector control as the
MONSAG system (Pontt 2004; Pontt et al., 1997); (ii) acoustic patterns analysis such as the
SAGAnalyzer (Magne et al., 2005; Almond and Valderrama, 2004), Electric Ear, AudioMill and
SmartEar (Herbst and Flintoff, 2012; Metso, 2010); (iii) proximity and conductivity sensors as the
Sensomag (Keshav et al., 2011).
This paper describes the onMill system: an instrumentation configuration based on vibration
measurements, which has been installed on a 38’x22’ SAG mill. This system estimates the total load
and ball filling level within SAG mills. The onMill can be used to close control loops for grinding
optimization at industrial scale.

METHODOLOGY
The onMill system corresponds to a multi-sensor configuration that operates together with a
specialized software to estimate the filling level in industrial SAG mills. Fig.1 shows the
arrangement of the instrumentation. A sensor attached to the outer shell of the mill (VIB-SCAN)
registers the vibrations caused by the motion of the load within the mill. Vibration data are sent
wirelessly to an electronic transceiver (VIB-RECEIVER) and then by optical fibre to a preprocessor
(VIPROC), where vibration analyses are carried out to determine the load position. This
information along with additional operating variables obtained from the distributed control system
(DCS) are transmitted to a remote computer, where a dedicated software is run.

2
 Figure 1 onMill system

Hardware
The VIB-SCAN was attached to the mill shell. The device includes two safety-lifting magnets, as
shown in Fig 2. Two redundant sensors allow reliable vibration measurements on the three axes
(normal, tangential and longitudinal) to be obtained.

SAFETY LIFTING MAGNET

Figure 2 VIB-SCAN

The VIB-RECEIVER receives the wireless signal transmitted by the VIB-SCAN. This signal is
converted to optical fibre signal and sent to the pre-processor device (VIPROC). The VIPROC
contains two single-board computers. The first one performs the calculation of the toe and shoulder
positions from the vibration signals, whereas the second one estimates the instantaneous power
draw from the mill cycloconverter. Load position and power draw estimations are sent via Ethernet
to the server, where the software is running.

Software
Filling level estimations are performed in the onMill software by means of an optimization
algorithm that includes phenomenological and empirical models for the power draw, bearing
pressure and toe and shoulder position. Additional models such as ball filling level and liner wear
as a function of time were also incorporated in the software. Table 1 summarizes the models
implemented in the onMill software, which were detailed by Robles et al. (2015).

3
Table 1 SAG mill phenomenological and empirical models

Model Equation
Power draw and PGross =PNL  PNet
load position,
Morrell (1993) PNL  f1  DM , LM , C 
PNet =k  Pcone +Pcylinder   f 2  DM , LM , C ,  L ,  T , S 
 T  f3  J C , C  , S  f 4  J C , C 
Bearing pressure Pres   J B    J C  J B   
Liner wear DM  t   m  D0  2 kL more  t  treline 

Ball wear

DB t , tR , i   m  D B0 
 kD t  t R , i 
 n° Balls  t R , 1  n° Balls t R , 2  
n° Balls t R , n  
B t      
 DB t , t R , i  slot size
 DB  t , t R , 1  
DB t , t R , 2  
DB t , t R , n  
Ball filling
VBalls  t 

 n° Balls  t  D  t , t  , J B t  
n
VBalls  t   3

6 i 1
R, i B R, i
V 1   B 

The total level estimation, JC, was performed by minimizing the weighted-squared error of the
power draw, bearing pressure and load position (toe and shoulder). Geometry information along
with operating conditions (from the DCS) were included in the model structure, such as ball filling
level, liner wear, rotation speed, % solids, among others. The minimization was carried out by
modifying the fill level of coarse ore, JM. Thus, the total fill level was approximated as the sum of JM
and JB.

RESULTS AND DISCUSSION

Ball filling level estimation

The proposed model depends on the wear of balls, ball recharge, and the internal volume of the
mill (ball wear and ball filling equations of Table 1). The ball filling estimation was performed from
recharge data for an industrial SAG mill (38’x22’) operation. The ball wear rate was fitted for two
periods: the first one from day 1 to 635 and the second one from day 636 to 935. The model was
adjusted to sixteen total grind out measurements. The best fits were obtained with apparent ball
wear rates (ball diameter) of 46.5 and 42.0 µm/h for each period, respectively. Results are presented
in Fig. 3.

4
 28
Model
Grind out
23
Jb [%]

18

13
0 234 468 701 935
Time [days]

Figure 3 Ball filling level estimation, onMill

The ball wear rate decrease might be related to changes in the SAG mill operation during the
second period. As shown in Table 2, the run time of the SAG mill and the ball recharge rate
remained approximately constant during the second period (regarding the first one). However, a
decrease in the feed tonnage was observed in the second period.

Table 2 Operational parameters along with 95% confidence intervals, 38’x22’ SAG mill

Period 1 2
Parameter Value Value
Run time, h/day 21.8 ± 0.3 21.0 ± 0.4
Feed tonnage, tpd 125133 ± 2220 116182 ± 2536
Ball recharge rate, gr /t (ball/ore) 186 ± 10 171 ± 7

Eleven of sixteen tests (about 70%) presented less than 1% of difference with regard to the actual
ball filling level measured during the grind out operations, Table 3. Grind out measurements were
carried out by the lifters exposed counting method developed by MolyCop (Moly-Cop Chile, 2000).
The highest differences occurred around the 7th test and might be related to the lack of grind out
measurements among 6th, 7th and 8th tests, which did not allow for the model feedback. In addition,
the uncertainties in the lifters exposed counting method, also contributed to the model deviations.

Table 3 Comparisons between onMill ball filling estimation and grind out measurements

Test Grind out JB, % onMill estimated JB, % Test Grind out JB, % onMill estimated JB, %
1 19.55 19.99 9 22.12 20.29
2 22.12 21.73 10 17.20 15.99
3 18.93 19.89 11 22.12 21.15
4 19.55 20.32 12 20.05 20.03
5 19.55 19.36 13 19.55 17.83
6 18.60 17.96 14 19.55 19.67
7 17.80 15.81 15 19.50 19.01
8 17.13 19.04 16 19.55 19.68

5
Power draw and bearing pressure estimation
Successful monitoring of the power draw for a SAG mill of 38’x22’ was achieved using the onMill
system. Correlation coefficients in the range of 0.74 to 0.98 were obtained between the estimated
power draw and that delivered by the DCS. In addition, correlation coefficients between the
estimated bearing pressure and the actual values were in the range of 0.50 to 0.77. The lower
correlation coefficients for the bearing pressure might be caused by lack of crash stop information.
Power draw model parameters can be estimated from grind out information, with some additional
considerations (total filling equals ball filling, with no interstitial pulp filling). However, because of
its model structure, the bearing pressure model (Table 1) must be calibrated with both crash stop
and grind out data.

Vibrations measurements
Vibration measurements were performed by the VIB-SCAN (Fig. 2) in a SAG mill of 38’x22’. Fig. 4a
shows the vibrations recorded in the outer mill shell. Intense vibrations are observed in the fourth
quadrant, where the toe of the load is located.
(a) (b)
0,16 90°
Normalized vibrations [-]

0,12 S
rm
0,08

180° 0°
ri
0,04

T
0,00
0 90 180 270 360
Angular position [°] 270°

Figure 4 Angular reference. (a) Vibration measurements. (b) Morrell’s model

According to the angular reference of Fig. 4b, a consistent response of the load position variables
(toe - shoulder), Fig. 5, under changes in the feed tonnage (Fig. 6a) and mill rotational speed (Fig.
6b) was observed. Increased filling level related to a simultaneous increase in the solid tonnage and
water flow rate at the minute 80, leads to higher shoulder position (closer to 90 °) and, at the same
time, a decrease in the toe position (closer to 180 °), as shown in Fig. 5. The same effect was
observed under rotational speed increase at the minute 80 (Fig. 6), due to the load expansion
(Napier-Munn et al., 1996). However, this behaviour was not observed after the second increase in
the rotational speed at the minute 220, which may indicate a mill discharge condition (water flow
rate and speed increase for constant tonnage).

6
 240 60

Shoulder position [°]

Toe position [°]
225 45

Toe
Shoulder
210 30
0 70 140 210 280
Time [min]

Figure 5 Load positions measurements

(a) (b)
6000 1900 10,0

Rotational speed [rpm]

Solid tonnage [t/h]

Fresh water [m3/h]

1800 9,5
5500
1700 9,0
tph
Water
5000 1600 8,5
0 70 140 210 280 0 70 140 210 280
Time [min] Time [min]

Figure 6 SAG mill measurements. (a) Fresh solid tonnage and water flow rate. (b) Mill rotational speed

Fig. 7 shows the estimation of the total filling level based on the combined power draw and bearing
pressure models in comparison with that obtained with the onMill system. The analysis was
conducted for the same period in which the load position was measured. The filling level obtained
with the power draw and bearing pressure models was JC = 34.0 ± 2.1%, whereas with the onMill
system was JC = 31.7 ± 1.1%. It should be noted that including the load position information the JC
variability (sample standard deviation) decreased close to 1%.
40
Filling level [%]

30

20
Total, Power-Pressure
Total, onMill
Ball
10
0 70 140 210 280
Time [min]

Figure 7 Filling level estimations

7
A higher filling level was obtained with the power draw and bearing pressure models. Closer
estimation to that obtained with the onMill system may be achieved by including additional crash
stop information (model feedback). Nevertheless, this does not guarantee a variability decrease due
to the variability of the power draw and bearing pressure measurements.

Total filling level estimation

The onMill system was tested with four industrial tests that included more than 250 SAG mill
operating data. These conditions considered continuous operations without maintenance stops.
Table 4 shows that test 3 presented significant higher total filling level than other tests, which was
related to the higher bearing pressure during this test. Power draw was greater in tests 3 and 4,
which was associated to the higher ball filling level (JB ~ 22%) and the higher rotation speed during
these tests. The total filling levels were in good agreement with the typical operating range for SAG
mills (30-35%) (Gupta & Yan, 2006).

Table 4 Total filling level estimation in the SAG mill

Set JB, % JC, % Power draw, kW Bearing pressure, kPa Speed, rpm Feed tonnage, tph
1 17.8 27.4 ± 2.4 14789 ± 1228 5483 ± 77 9.17 ± 0.41 5994 ± 406
2 19.3 30.3 ± 2.9 14455 ± 667 5394 ± 44 9.09 ± 0.28 5496 ± 505
3 22.5 31.7 ± 1.1 15455 ± 741 5523± 21 9.34 ± 0.25 5494 ± 139
4 22.3 29.8 ± 2.8 15634 ± 382 5416 ± 32 9.32 ± 0.14 5450 ± 131

CONCLUSIONS
The onMill system estimates the load level in industrial SAG mills with typical operating
information provided by the DCS and measurements of non-invasive instrumentation.
Ball filling level estimations in an industrial SAG mill (38’x22’) were performed for two different
operating conditions. The best fits were obtained with apparent ball wear rates (ball diameter) of
46.5 and 42.0 µm/h. Seventy percent of the estimations presented less than 1% difference with
regard to the actual ball filling level measured during the grind out operations.
The proposed algorithm was able to successfully forecast the power draw with correlation
coefficients greater than 0.8. Considering the bearing pressure model structure, crash stop
information is recommended to improve the correlation coefficients between the estimated values
and those delivered by the DCS.
The estimated total filling levels were in good agreement with the typical JB and JC values observed
in the SAG operation (grind out and crash stop procedures). The estimated JB and JC values were
also in good agreement with changes in the key variables such as feed tonnage, ball filling level,
rotational speed, bearing pressure and power draw.
The filling levels obtained from the onMill system presented less variability than those obtained
only by the combined power draw and bearing pressure models. This result encourages the
implementation of additional instrumentation to obtain robust total filling estimation in industrial
SAG mills.

8
NOMENCLATURE
PGross gross power consumed by a SAG mill
PNL power supplied to the motor when the mill is empty (no-load power)
PNet power supplied to move the load
DM mill diameter
LM mill length
ɸC fraction of the theoretical critical speed
k lumped parameter which accounts for heat losses and inaccuracies associated with
assumptions and measurements of the load shape and motion
PCone cone end theoretical power
PCilynder cylindrical section theoretical power
ρL load density
θS shoulder angular position
θT toe angular position
JC total mill filling
Pres mill bearing pressure
α, β, γ empirical parameters, bearing pressure model
JB ball filling
D0 diameter of the mill with new liners
kL liner wear rate
ṁore solid tonnage (raw feed)
t time
(t - treline) time since the mill reline
DB diameter of balls recharged at time tR, i
DB0 diameter of a new ball
kD apparent ball wear rate
B(t) array of balls where the first row consists of the amount of balls recharged at time tR, i,
whereas the second row consists of the diameter of those balls
i index, corresponds to the columns of matrix B(t)
n°Balls amount of balls recharged at time tR, i
VBall mill internal volume occupied by balls
V mill internal volume
εB porosity of the balls bed

REFERENCES
Almond, D. G. & Valderrama, W. (2004). ‘Performance enhancement tools for grinding mills’,
Proceedings of the first International Platinum Conference – Platinum Adding Value, Sun City-South
Africa, South African Institute of Mining and Metallurgy, Sun City, pp. 103-110.
Gupta, A., & Yan, D. S. (2006). Mineral Processing Design and Operation. Elsevier, First Edition,
Amsterdam, The Netherlands.

9
Herbst, J. A., & Flintoff, B. (2012). Recent advance in modeling, simulation, and control of mineral
processing operations. In: Separation technologies for minerals, coal, and Earth resources. Society for
Mining, Metallurgy, and Exploration, First Edition, Englewood, Colorado, USA, pp. 667–687.
Keshav, P., de Haas, B.; Clermont, B.; Mainza, A.; Moys, M. (2011). ‘Optimisation of the secondary
ball mill using an on-line ball and pulp load sensor - The Sensomag’. Minerals Engineering, Issue 24,
p. 325–334.
Magne, L., Valderrama, W.; Pontt, J.; Perelli, E.; Velásquez, C.; Sepúlveda, G. (2005). System to
determine and analyze the dynamic internal load in revolving mills, for mineral grinding. United States of
America, Patent nº US 6,874,366 B2.
Metso (2010). www.metso.com, PTI products, viewed 23 June 2016, <
http://www.metso.com/in/miningandconstruction/MCT_India.nsf/WebWID/WTB-101221-22577-
4A53D?OpenDocument#.V2wBBLjhDDc>.
Moly-Cop Chile (2000) http://www7.uc.cl/sw_educ/simula2/html/simulacion.html, viewed 6 July 2015,
<http://www7.uc.cl/sw_educ/simula2/html/mediachargelevel.html>.
Morrell, S. (1993) The prediction of power draw in wet tumbling mills, Brisbane: PhD Thesis,
Department of Mining and Metallurgical Engineering, University of Queensland.
Napier-Munn, T. J., Morrell, S., Morrison, R. D., Kojovic, T. (1996) Mineral Comminution Circuits,
Their Operation and Optimisation, Second Edition, JKMRC Mining and Mineral Processing, Brisbane,
Australia.
Pontt, J., Valderrama, W., Magne, L. & Pozo, R. (1997) ‘MONSAG: A system for On-Line charge
monitoring in semiautogenous grinding mills’, Workshop on SAG grinding, Viña del Mar, Chile.
Pontt, J. (2004) ‘MONSAG: A new monitoring system for measuring the load filling of SAG mill’.
Minerals Engineering, Volume 17, pp. 1143-1148.
Robles, H., Ledezma, T., Inostroza, J., Vinnett, L., Valderrama, V., Pontt, J. (2015) ‘New Approach
for Load Level Estimation in SAG Mills Using a Sensor System and Optimization Algorithm’,
Proceedings of the 11th International Mineral Processing Conference – Procemin 2015, Santiago Chile,
Gecamin, Santiago, pp. 292-301.

10
The Development of SGI Test at El Soldado
Miguel Becerra and Francisca Jorquera
El Soldado, Anglo American, Chile

ABSTRACT
SAG milling operation at El Soldado, owned by Anglo American Copper is the only one operating
circuit in Chile which acts as a closed direct circuit, known as Single Stage SAG. With the
geometallurgy development during last decade it was adopted Axb, obtained from SMC test, like
hardness parameter to estimates throughput and feed geometallurgical model. The inherent
complexity of this kind of circuits and the difficult to obtain massive data from drill cores and
improve the estimations, it were necessary search another way to estimate the specific energy
consumption.
Amelunxen (Amelunxen, 2014) described in detail the features to implement the test called SGI (Sag
Grindability Index, a variation of the Starkey test) with a standard feed size distribution, grinding
conditions, mill size and one target: develop a grinding kinetic test to find a specific products size
distribution. El Soldado developed the implementation of this test with an extensive test campaign
to develop the correlationship between SGI with Axb and industrial specific energy consumption.
The results obtained was an own tool to estimate (for different objectives) an indicative parameter
of hardness using less sample with low costs, fast results and reliability. This papers presents the
results obtained from sixty samples which correlating SGI results with Axb and Bond Ball Work
Index. For specific energy consumption estimation the analysis shows four surveys comparing
laboratory and industrial results.

There is no full article associated with this abstract.

1
Single Stage SAG Circuit at El Soldado
Miguel Becerra and Francisco Vicuña
El Soldado, Anglo American, Chile

ABSTRACT
SAG milling operation at El Soldado, owned by Anglo American Copper is the only one operating
circuit in Chile which acts as a closed direct circuit, known as Single Stage SAG. Commonly found
in the literature (Putland, 2005; Mainza et al, 2006) important operational analysis for this circuit
but focused mostly on treating gold ores and differents mill sizes. Copper mining in Chile has its
own characteristics associated with most common circuits SABC-B (or A) type including pebble
crushing stage that were adopted in partial form to this operation without considering their
inherent complexity and conceptual differences.
This paper presents a detailed analysis of operational trends and surveys including the effect of the
circulating load from classification equipment (hydrocyclones) and the influence of the Jb/Jc rate in
the circuit operation including size distribution around. Current instrumentation allows estimation
of this parameters continuously and online, which in turn gives the possibility of rescuing relevant
information to the analysis.
The results indicate a strong dependence on specific energy consumption with circulating load and
evidence the characteristics of the operation of this circuit complexities. In adition, the product size
distribution and the pebble generation is affected by changes in this parameter.
The effect of the ball filling level and the particle size distribution under 3 inch in the fresh feed
play a fundamental role in the generation of pebbles and overall circuit performance.
The benefit of operating balls fill levels under 13 % and circulating loads betwen 250 % ans 300 % to
decrease the specific energy consumption in this kind of circuit and the gravitant effect of the feed
particle size distribution over pebble generation was identifies and demonstrated.

There is no full article associated with this abstract.

1
Advanced Crusher Control Strategy Using Model Based
Predictive Controller
Juan Morales1, Tomás Carricajo1, Freddy Gómez2, César Garrido2 and Marcelo Jara2
1. ANDRITZ, Chile
2. Cía. Minera Doña Inés de Collahuasi, Chile

ABSTRACT
A reliable control system which reduces variability and also minimizes human intervention is desirable
for the effective and optimal operation of any process. The Rosario crusher from Compañía Minera Doña
Inés de Collahuasi (CMDIC) has been operated manually since its startup in the first quarter of 2012. The
challenges have been to design and implement an automated control strategy that stabilizes and
maximizes the ore throughput in automatic mode, while reducing the stress on the main equipment of the
transport system.
The crusher control system encompasses the discharging of trucks in Rosario through to the ore transport
system that feeds the stock pile. The throughput is measured a distance away from the crusher, in
conveyor CV206, which generates a physical dead time in the process. Secondary goals are to maintain the
ore levels in bins to protect damage in the feeder and provide recommendations to operators in order to
manage truck dumping.
An advanced control design consisting of two layers was implemented. On the lower level layer, model-
based predictive control was implemented; chosen due to its stability, convergence and excellent handling
of delays. The second layer is a supervisory control that maximizes throughput, protects main equipment
of the ore transport system to maximize its life time and controls truck dump signals.
Preliminary results during the first days of operation revealed improvement in process variability while
main equipment ran in a safe condition. The results after several months of operation have been even
better showing an increase in throughput up to 10 % and decrease in process variability of up to 30 %.
The crusher has now reached its design capacity of production due to improvements in the ore transport
system, DCS and ACE Controller lead by CMDIC engineers with support of designers, vendors and the
ANDRITZ expert team. These improvements result in increased concentrator daily tonnage by 7 % due to
maximizing stock pile level in Ujina where ACE Controller has been part of the solution to reach this
number.

1
INTRODUCTION
Automatic control encounters many challengers in the mining industry. An Advanced Process Control
(APC) strategy is desirable to optimize many mineral processing operations due to variations in
unmeasured ore properties, nonlinear behaviors of the equipment and transportation delays that
confound traditional control systems. Compañía Minera Doña Inés de Collahuasi has been focused on
stabilizing its processes with risk management as a base to reach reliable and systematic results since 2013.
With that focus CMDIC is working on improving control of production (Jara & Gómez, 2016) of all
processes and, in particular, Rosario’s crusher from CMDIC shows many aspects that a regular control
cannot handle.
Rosario’s crusher is a 63 x 114 inch gyratory crusher (one of the world’s largest for this application)
operating at 4500 amsl with design capacity to process 10500 tph representing nowadays a 97 % average
of CMDIC’s production. This asset has been operated in manual mode by operators since the beginning of
the operation in the first quarter of 2012. CMDIC began to investigate advanced control applications for
the crusher with the main objective of decreasing variability and reducing workload on the operators.
Expert systems are a popular approach to manage the control of these difficult processes in the mining
industry; these kinds of control systems are traditional in grinding and flotation areas. In some cases, the
expert system may control the actuators directly using a rule-based approach, or rely on conventional
proportional-integral-derivative (PID) loop controllers to perform the underlying regulatory control of the
process where possible. However, rule-based expert and PID controller cannot handle dead times,
transportation delays and unmeasured ore properties that mining processes present. In this case model-
based predictive control (MPC) provides an additional tool to improve the control of critical processes
where rule-based expert or PID controller is not well suited to the application.
MPC controller (Clarke, Mohtadi & Tuffs, 1987) is able to reduce process variability and reaches the best
performance that could be obtained with PID or expert system control strategy. MPC is able to control
process and equipment where there are delays in the response of actuator and multiple interactions
between process variables. In particular, MPC is able to optimize the control of processes that exhibit an
integrating type response in combination with transport delays or variable interaction. This type of
response is particularly difficult to control, and it is common in mineral processing for many different
processes including level control of flotation cells, SAG mill weight control and crushers.
MPC has been successfully applied to different mining process at several copper mines. MPC consistently
demonstrates the ability to provide reduced process variability and increased stability compared to PID or
expert system based control. This paper presents the successful control strategies that have been
implemented in Rosario’s crusher from CMDIC in 2015.

ADVANCED CONTROL SYSTEM

ANDRITZ AUTOMATION offers an APC system known as the ANDRITZ Advanced Control Expert
(ACE™ *). This control system has been implemented in different industries, such as Mining, Pulp and
Paper and Oil and Gas. The solution has a control philosophy that divides the control objectives into two
advanced control layers. Thus, each layer solves different problem at the same time.

* Trademark of the ANDRITZ GROUP. For information regarding ownership and countries of registration, visit
www.andritz.com/trademarks.

2
The lower layer objective is to stabilize the process using model predictive control tools. The MPC was
chosen for the first layer due to its good stability, convergence and handling of delays; it generally
replaces conventional PID controllers in classical control systems. ACE has the BrainWave™ * MPC
controller (Gough, 2011), this MPC platform offers stable solutions; also well understood in terms of
gains, time constants, variable interaction, etc. and has short implementation time.
The second and upper layer aims to optimize the process targets, while at the same time collect
information for decision making in order to increase performance and automatically adapt its conditions
to changing ore and/or process conditions.
This layer is a supervisory control, which sends the desired process operation points to the MPC
controllers in order to drive the process towards optimal performance. It is similar to a traditional expert,
but is not only rules; several automatic control modules were developed in this layer to improve the
process and adapt the control system parameters. As a result of ACE Crusher, this layer has specific logic
to maximize throughput, protect the equipment to extend its life cycle and make recommendations about
trucks discharging. Most of the logic has been implemented on the platform of advanced control
IDEAS™  (ANDRITZ, 2016).

Model-based Predictive Control implemented at CMDIC

The advanced process controller used by BrainWave is a model-based predictive controller that uses a
state-space model derived from the Laguerre orthogonal basis functions (Zervos & Dumont, 1988) so that
adaptive control could be achieved without the need to know the process order or the time delay in
advance. Each Laguerre basis function is a polynomial multiplied by a decaying exponential; so these
basis functions make an excellent choice for modelling transient behaviour as they are similar to transient
signals.
Model adaptation may be executed in real time following a recursive algorithm (Salgado, Goodwin &
Middleton, 1988). Particularly, BrainWave includes a special algorithm designed for processes with an
open loop integrating response that outperforms traditional MPC approaches, and also behaves especially
well both in fast and slow changing processes, such as the crusher, grinding circuits and flotation areas.

CRUSHER CONTROL STRATEGY

As a simple explanation of the process, trucks from the mine site are constantly feeding the crusher while
the bin under the crusher collects all the processed ore. Furthermore, there is a variable speed feeder
below the bin that transports the ore to conveyor belt CV206 and the production of the crusher is
measured in this last belt.
The main objective of ACE Crusher is to reduce production variability, as the measurement of weight on
the conveyor belt. As it is shown in Figure 1, BrainWave MPC layer manipulates the rate of speed of the
feeder to minimize variability and maximize production on the conveyor belt. Furthermore, the
supervisor layer sets the lower layer target production to keep the bin level in a range according to ore
flow from trucks. This also avoids emptying the feeder while conveyor motors are protected from high
temperature, torque and vibration. The control system also makes recommendations about truck
dumping based on bin level performance. In some cases conveyor CV206 operates with less than four

* Trademark of the ANDRITZ GROUP. For information regarding ownership and countries of registration, visit
www.andritz.com/trademarks

3
motors (normal condition) and ACE Controller allows it to keep the strategy even in this condition giving
the operators enough flexibility to run the system.

Figure 1 ACE Crusher

RESULTS
Before ACE was implemented, the crusher was controlled in manual mode as we already mentioned.
Figure 2 shows typical crusher and bin behavior under both manual control and ACE control. These
graphics and the results shown below were calculated with data during October 1, 2015 and December 31,
2015.

Figure 2 Main Result Trends; Manual Control v/s ACE

4
The MPC control is able to maintain production on target with low variability while the supervisor
prevents motor problems in the conveyor belt. This performance improvement also increases the
operational availability for process.
Manual control was defined as base case to compare with ACE control. Thus, after three months of
continuous operation, the percentage improvements for using ACE are:
 A decrease of 25.1 % in production variability. This number means a significant increase in
stability and controllability and therefore an improvement in availability.
 An increase in throughput of up to 10.8 %. This result is due to increased availability, the ore is
transported faster with ACE than manual control while conveyor motors are protected.
 A decrease of 25.6 % in bin level variability and a decrease in feeder speed variability of 20 %.
The reduction in variability is seen in all relevant process variables.
Another important result is that the system has been used 99 % of the time since it was installed. This is
because operators were involved in the design from the beginning; thus, operators have understood and
trusted the ACE control actions. This confirms that an adequate Management of Change is necessary for
successful implementations.
As a part of the vision for continuous improvement to do with reducing variability with risk
management, in 2015 CMDIC invested in re-designing its ore transport systems (CV206 and CV207).
Considerable changes to the brake system, tension control in CV206, Control Logics in DCS, Alarm
Management and ACE Controller increased utilization of the transport system by 4 %, reduced variability
of availability by 8 % and increased the concentrator daily tonnage by 7 %. This was due to maximizing
the stock pile level in Ujina where ACE Controller has been part of the solution to reach this successful
result.

NEXT STEPS: IMPROVEMENTS AND OPPORTUNITIES

As a result of these successful experiences CMDIC is evaluating some other APC applications in their
secondary grinding, classification, flotation, thickeners and filtering processes. Regarding the Rosario’s
Crusher and the ore transport system some improvements could be implemented in ACE Controller:
 Link Dispatch System via ACE Controller to predict in a better way the real ore flow from
trucks
 Improve Adaptive capacity of ACE Controller adding scenarios considering variability of
mineral from open pit mine
 Implement BrainWave in others feeder in order to operate the whole system in automatic mode
 Improve feeder performance during a system start up with external control actions
 Integrate particle size distribution to ACE Controller in order to achieve quality product with
less variability and maximizing throughput
The motivation for these changes is to get a fully automated process. With these steps, a Fault Tolerant
Control System could be implemented in Rosario’s Crusher in short term.

5
CONCLUSIONS
The two-layer control strategy proved to be a successful solution for controlling the Rosario’s Crusher bin
level, feeder speed and truck dumping. It has also demonstrated that choosing adequate and intelligent
process variables and analyzing their behavior, can create more optimization opportunities. Thus, the
mining industry can achieve economical results by including more and better practices of advanced
controlling, such as the one shown and discarding outdated technologies that deliver fewer benefits and
are less reliable.
Advanced control systems improve the process, not only in terms of production, but also in safety,
availability, stability and production cost. Reducing variability on the operation improves stability and
opens the door for optimization. The APC strategy has been presented to demonstrate its potential to
increase the mining business value. Technology is the engine of productivity; efforts should be made
towards incorporating higher levels of automation and process innovation.
Operators were an important part of the project from the beginning; initially they contributed their
knowledge to the advanced control team and they were always informed about the progress of the
project. As a result of these practices, operators learned to use the APC system and now ACE has a 99 %
utilization rate. This result is not common as in many cases operators reject new controllers because they
do not trust them.
This represents a great advance over existing control strategies because it adapts itself to changing process
conditions, anticipates disturbances, knows the process behaviour. It also represents another step towards
the Fault Tolerant Control System that is part of ANDRITZ´s vision for Automation 4.0 for the Mining
Industry; it means completely automated and connected plants.

ACKNOWLEDGEMENTS
The authors wish to thank Compañía Minera Doña Inés de Collahuasi for permission and collaboration in
order to publish this paper. A thank you is also extended to the crusher operations group at CMDIC for
their patience and contributions to the success of these projects and for letting us perform the required but
unavoidable tests.

REFERENCES
Andritz (2016), http://www.andritz.com/, viewed June 29, 2016.
Clarke, D.W., Mohtadi, C., & Tuffs, P.S. (1987), Generalized Predictive Control – Part I. The Basic
Algorithm, Automatica, Vol. 23, No. 2, 1987, pp. 137-148.
Gough, W. A. (2011), BrainWave®: Model Predictive Control for the Process Industries, Advanced Model
Predictive Control, Dr. Tao ZHENG (Ed.), pp. 393-418.
Jara, M. & Gómez, F. (2016) Automation and his contribution to the development of mining excellence:
The reality of Collahuasi, Minexcellence 2016, Santiago, Chile.
Salgado, M.E., Goodwin, G.C. & Middleton R.H. (1988) Modified Least Squares Algorithm Incorporating
Exponential Resetting and Forgetting, Int. J. Control, Vol. 47, No. 2, pp. 477-491.
Zervos, C.C. & Dumont, G.A. (1988), Deterministic Adaptive Control Based on Laguerre Series
Representation, Int. J. Control, Vol. 48, No. 6, pp. 2333-2359.

6
Energy and Throughput Optimization for Fine
Grinding in Stirred Media Mills
Ilesh Shah
FLSmidth, USA

ABSTRACT
Energy and throughput considerations are paramount in selection and operation of comminution
(crushing and grinding) circuits and equipment. Increased attention to sustainability and
productivity has amplified the need to optimize the operation of all comminution devices.
Through decades of use and development, tumbling mills have been recognized as the device of
choice for grinding of hard rock minerals. Lately, however, stirred media milling has gained
traction in fine and ultrafine grinding of minerals due to increased energy efficiency it provides
compared to conventional grinding in tumbling mills. Due mainly to throughput and feed particle
size limitations, the applicability has been restricted mainly to regrind or in some cases to tertiary
grinding applications. Nevertheless, it is an important growth area in comminution technologies.
The present paper deals with the effect of one of the important operating parameters that
significantly influences the productivity of stirred media mills, viz. the mill speed. Limestone rock
dust was ground from a starting size of F80 of about 75 µm to a product P80 of less than 7 µm in a lab
scale stirred media mill. The energy utilization and evolution of particle size distribution was
tracked for three different rotational speeds. Higher impellor speed corresponds to increased input
energy and results in finer particle size distributions. It is concluded that grind optimization will
require a detailed trade-off analysis.

1
INTRODUCTION
Stirred Media Mills have been in use as comminution device for fine grinding in select industries
since 1930's. In minerals industry, they are now being accepted, and sometimes preferred, as an
alternate for less energy efficient conventional tumbling mills (Wills & Finch, 2015). This is
especially the case where the process demands grinding to finer than 75 µm. Typically, tonnage
(t/h) treated in these mills are modest, compared to primary and secondary grinding. However,
total installed power can still be sizeable, as specific energy (kWh/t) for fine grinding is usually
larger than crushing or intermediate grinding. Fine grinding, in general, differs from traditional
milling in one more important aspect. No industry standard has yet been established for estimating
throughput rate or energy requirements. Thus, systematic testing is needed to establish influence of
important operating parameters.
The FLSmidth® VXPmill is a vertical Stirred Media Mill (SMM), offered in a range of laboratory,
pilot, and production mill sizes. A multitude of adjustable design and operating parameters,
grouped in two classes of significant variables as process state and mill configuration variables
(Rahal, Erasmus & Major, 2011) makes it a versatile fine grinding device. This flexibility
nevertheless, demands supplementary testing for proper evaluation and fine tuning for optimizing
the mill performance. In previous papers (Shah et al., 2013; Hines, Shah & Rahal, 2014) have
touched on some research done on industrial and lab scale VXPmills. The present paper examines
the effect of one of the important operating parameters, mill speed, on energy and throughput.

EXPERIMENTAL DETAILS AND MATERIALS

A VXP10 (10 litre) laboratory mill, shown in Figure 1, was used in the research program for a set of
tests performed in FLSmidth Minerals Technology Center in Salt Lake City, Utah. The tests utilized
Limestone Rock Dust, gratefully provided by Graymont's Cricket Mountain Operation.

Figure 1 The VXP10 grinding mill

2
The skid mounted system and its operation has been described in an earlier paper (Hines, Shah &
Rahal, 2014). The mill has the flexibility to operate over a wide range of rotational speeds. For the
results described in the present paper, all the mill operational parameters were fixed, except the
mill speed. The tests used 2.2 - 2.8 mm ceramic beads with a volume fill of 60% of the net mill
volume. The Limestone feed, when analysed in Malvern gave a feed size of about 75 µm (P80, 80%
passing size).
About 70 litres of slurry, at a slurry density of 1.4 kg/L, was prepared for each of the three tests.
This was done by mixing a predetermined weight of limestone with known quantity of water. For
the test series, the only independent variable was the mill rotational speed. Three speed settings
were evaluated. First a routine test at 800 RPM was performed, followed by one at half the speed
(400 RPM). The last test was performed at double the usual speed (1600 RPM). A total of eight
pendulum passes were done for each speed setting. The usual test procedure is to perform two
passes at a fixed high slurry flowrate (depending on the feed particle size). Next two passes are
done at a lower flow rate and so on till the last pass. The first two tests (at 800 and 400 RPM)
followed this procedure. For the last test at higher speed, all eight passes were done only at the
flowrate of 8 litres per minute (LPM).
Table 1 presents a summary snapshot of test conditions and an overview of some results. Included
in the table are power data from a series of tests done with only water circulating through the mill
(no slurry). The mill still had the same 60 % of media charge as in the rest of the tests.

Table 1 Test Conditions and Results Snapshot

Rotational Speed, RPM

400 800 1600
Slurry Flow, LPM 8 to 2 8 to 2 8
Water & Media Power , kW 0.04* 0.32 0.88
Average Power with Slurry, kW 0.42 1.08 2.31
F80, µm 62.5 81.0 76.2
After 2 passes (both at 8 LPM), P80, µm 32.8 16.6 13.6
After 8th Pass, P80, µm 6.9 4.8 6.4
*Extrapolated value.

RESULTS AND DISCUSSION

Figures 2 & 3 depict log-linear particle size distributions (PSDs) obtained by analysing feeds and
products in Malvern size analyser. Figure 2 is for all three feed PSDs and also for products of 2nd
pass (P2) for the three tests. Figure 3 is for a complete test for all eight passes (P1 to P8), shown only
for the 800 RPM test. Clearly, the successive passes produced a measurable change in the PSD. The
size distribution plots of Figure 2 & 3 demonstrate that P80 can be used as a representative size for
indicating the progression of grinding at each successive pass.

3
 Figure 2 Feed and 2nd pass PSDs, limestone grinding

Figure 3 PSDs for 800 RPM test, limestone grinding

It was observed that the mill power, measured in kW, remained practically constant throughout the
eight passes of a given test. Figure 4 presents the average power measured as a function of mill
speed, as discrete points for the three tests (solid filled symbols). Clearly, rotating the mill impellor
at higher speed requires more power. Also displayed in Figure 4 are points representing measured
mill power when only water was circulating through the mill loaded with same amount of media as
the slurry tests. Notice that mill power increases linearly with speed for these media-water tests.
It is interesting to note that if a line is imagined going through the three data points for power when
slurry was circulating, it will not be parallel to the solid line for water only power. The divergent

4
nature is indicative of the fact that effective power due to presence of slurry (i.e. kW with slurry –
kW with water) is progressively higher at higher mill speed. The trend for this net mill power is
represented by the dotted line in Figure 4.

Figure 4 Mill power vs mill rotational speed

Traditional “signature plots”, showing cumulative specific energy of the mill (kWh/mt) versus
product size (represented by P80), on a log-log scale, are displayed in Figures 5. Evidently, specific
energy curve shifts upward and towards left at higher rotational speed. Comparing the two tests, at
400 and 800 RPM, and considering the product P80 size, it is seen that corresponding passes at
higher RPM generated finer P80. For example, after 2 passes, the test with mill speed of 400 RPM
produced a product P80 of 32.8 µm compared to corresponding value of 16.6 µm for the 800 RPM
test. The same trend continues for all successive passes. Note that the results are for two identical
tests in terms of operating parameters, except for the speed of rotation. For the final product, at the
end of the 8th pass, the two tests resulted in final P80 of 6.9 and 4.8 µm respectively. Obviously,
running the stirred media mill at higher rotational speed, all other operating conditions being
similar, provides a finer product.
Comparison of the two tests discussed so far with the third test at 1600 RPM is a bit trickier. All
eight passes in the 1600 RPM test were at a constant higher slurry flow rate. Looking at the product
after first two passes, the 1600 RPM test provided the finest product, also see Figure 2. Displayed in
Table 1 is the fact that the P80 of 13.6 µm (after 2nd pass) is smallest amongst the three speed tests.
Note that all three tests had identical slurry flow rate for the first two passes.

5
 Figure 5 Signature plots for limestone grinding

Now consider the product size of the final products for the three tests, at the end of the 8th pass
(refer to Table 1 and Figure 5). The 1600 RPM test gave a final product size of 6.4 µm which is finer
than 6.9 µm (400 RPM test) but coarser than 4.8 µm (800 RPM test). There is a convincing
explanation. At higher slurry flow rate, the effective residence time the particles remains in the
grinding zone is shorter. The overall residence time for the 1600 RPM test, with all eight passes at 8
LPM is shorter than that for the 800 RPM test that had progressively slower slurry flow rate after
the second pass.
Another observation from Figure 5 is that 8th pass of the 400 RPM test, 6th pass of 800 RPM test
and 7th pass of 1600 RPM test all gave a product P 80 of close to 7 µm. The PSDs for these passes are
shown in Figure 6. Note the close matching of the entire product size distributions for respective
passes of the three tests having same P80 of 7 µm.

Figure 6 PSD for limestone grinding to P80 of 7 µm

6
The discussions so far has focused only on product particle size. Another characteristic indicator in
the signature plot is the energy utilization. Figure 5 indicates that higher RPM tests produced finer
product, albeit at higher specific energy value. For production of P 80 of say 7 µm, the mill running
at 800 RPM corresponds to a specific energy value of 29 kWh/mt compared to about 22 kWh/mt at
400 RPM, and a value close to 48 kWh/mt for the 1600 RPM test. Thus from energy efficiency
perspective, the slower speed is preferable for this material. To summarize, we have two opposing
assertions; higher rotational speed corresponds to finer product albeit at higher specific energy.
In Figure 7, the traditional signature plot is supplemented with an additional variable. The solid
symbols and lines represent the specific energy, represented on the Y-axis on the left. Plotted on the
2nd Y-axis on the right are values for processed throughput (represented by empty symbols and
dashed lines), of dry metric tons per hour (dmtph).

Figure 7 Modified signature plots for limestone grinding

Focusing on the varying reference P80 of 12, 10 and 7 µm, Table 2 was generated by interpolating
values from the plots of Figure 7, for Specific Energy and Throughput. Note that from practical
perspective, selection of a grinding device is based on the considerations of capability to provide a
required grind size, at acceptable levels of energy and throughput. Towards the goal of identifying
optimization, the modified signature plot of Figure 7 (and data from Table 2) can be used to create a
throughput vs specific energy plot such as one displayed in Figure 8. The plot can be used to
demonstrate the effect of speed change on specific energy and throughput, for different target
particle size.

7
Table 2 Energy and Throughput Interpolations for P80 of 12, 10 and 7 µm

Target P80, µm 12 µm 10 µm 7 µm
Rotational Speed, RPM 400 800 1600 400 800 1600 400 800 1600
Specific Energy, kWh/mt 9.1 11.4 18.8 12.3 15.6 25.8 21.7 28.6 47.8
Energy Change 1.0 1.25 2.06 1.0 1.27 2.11 1.0 1.32 2.21
Throughput, dmtph 0.0418 0.0853 0.1214 0.0318 0.0634 0.0891 0.0186 0.0355 0.0487
Throughput Change 1.0 2.04 2.90 1.0 1.99 2.80 1.0 1.91 2.62

Figure 8 Throughput vs specific energy plots

The tests showcased in this paper indicates that limestone can be ground from a feed having F80 of
around 75 µm to a product PSD with P80 of 7 µm (solid line in Figure 8) with specific energy as little
as 22 kWh/mt. This is the case when the mill is operated at very low rotational speed of 400 RPM.
Doubling the rotational speed (to 800 RPM), the throughput value is almost doubled, at specific
energy increase of about 30%. However, the rate of rise in throughput is more modest in relation to
specific energy rise when the speed is doubled again, to 1600 RPM. Only about 35% incremental
improvement in throughput is obtained for 67% increase of specific energy, going from 800 to 1600
RPM. Overall though, approximately 2.6 times the throughput is possible if the speed used is 1600
RPM instead of 400 RPM, but the specific energy will be about 220% of that at 400 RPM. Notice that
for each of the three target P80 values considered (12 µm (long dash dot dot line), 10 µm (round dot
line) or 7 µm (solid line) in Figure 8), an analogous trend is observed.

8
CONCLUSION
Fine grinding in stirred media mills such a VXPmill is dependent on several key parameters. The
parameters relate to mill, media and material characteristics. The speed of rotation of the impellor is
an important operating parameter. It has a very significant effect on product particle size,
throughput rate, and specific energy. Energy utilization as low as 22 kWh/mt can provide a product
P80 of about 7 µm for Limestone dust, from a feed having 80% passing size of about 75 µm. The
relatively low energy to provide such a fine product PSD is obviously because of high grindability /
low hardness of limestone.
The choice of operating speed for optimized grinding performance needs careful analysis. It is seen
that operating the mill at high impellor speed results in finer product PSD. However, the increased
speed corresponds to higher input energy. The higher speed of rotation also allows higher
throughput (processing rates) from a given mill. Realizing that higher wear rates are to be
anticipated at higher speeds, the mill performance optimization will require a detailed trade-off
analysis. The decision to use higher speed of rotation, giving increased throughput from a given
mill size for a given target size, has to be balanced against the operating costs (major components
being additional energy and wear/maintenance requirements). The operator/ metallurgist needs to
take into account the suitable target PSD value in deciding the optimal operating parameters for a
stirred media mill.

ACKNOWLEDGEMENTS
The author gratefully acknowledges the contributions from following FLSmidth colleagues: D.
Hunter, C. Braun, and T. Olson. As mentioned earlier, Graymont's Cricket Mountain Operations
provided the limestone dust samples utilized in this study.

REFERENCES

Wills B. A. and Finch J. A. (2015). Will's Mineral Processing Technology, 8th Edition, Elsevier.
Rahal, et al. (2011). Knelson-Deswik Milling Technology: Bridging the Gap between Low and High
Speed Stirred Mills. Canadian Mineral Processors 2011. Ottawa, Ontario, Canada.
Shah, et al. (2013). Field study of an industrial stirred media mill. 13th European Symposium on
Comminution & Classification. pp 101-104.
Hines, et al. (2014). Effects of media density to slurry density ratio on fine grinding. SME Annual
Meeting 2014. Preprint 14-159.

9
Debottlenecking of Secondary Grinding Circuit: An
Hydraulic Analyze over the Process Results
Claudio Muñoz
Cía. Minera Doña Inés de Collahuasi, Chile

ABSTRACT
Doña Inés de Collahuasi is a copper mining company located in Chile’s First Region. Its operations
include a concentration plant with a processing capacity of 135,000 TDP, located at 4,320 meters
above sea-level. The flotation stage of the copper and molybdenum concentrator plant consists of a
rougher flotation circuit equipped with both conventional self-aerated cells and injected air cells.
The cleaning circuit, however, operates in two stages; the first uses conventional self-aerated cells
and the second cleaning stage is performed in columns.
The concentrator plant has 3 grinding lines working in parallel. Specifically, the line 3 has a 40’x24’
SAG Mill operating with two 26’x38’ Ball Mills. The secondary grinding circuit presented some
issues related with spillages prior the ball mill feed. Based on the above both a metallurgical study
as a hydraulic study was carried out to define the main constraints of this circuit.
The results obtained in those studies allowed to identify where the secondary grinding was
constrained. As a consequence, a low cost project was developed, which allowed to improve the
throughput of the plant and improve the confiability of the pumping system.

There is no full article associated with this abstract.

1
The Effect of Key Operating Conditions on Grinding
Media Consumption in Semiautogenous Mills
Levi Guzmán 1and Jaime Sepúlveda2
1. Moly-Cop Adesur, Perú
2. J-Consultores, Chile

ABSTRACT
Moly-Cop has been involved in analysing the different mechanisms of grinding media consumption
due to media being one of the main costs in the comminution process. A thorough understanding of
the independent contribution of each grinding media consumption mechanism is extremely
important, especially in semi-autogenous grinding (SAG) operations where macrowear (spalling)
and ball breakage cannot be ignored or neglected.
The high impact conditions imposed in modern SAG milling environments has created serious
challenges for grinding media manufacturers since their products must be able to perform with
minimal breakage and surface spalling.
Grinding media consumption rates and their dependence on key operating conditions in semi-
autogenous grinding (SAG) mills is analysed based on a detailed mathematical model recently
developed by Moly-Cop´s specialists. It describes and takes into account the three primary media
consumption mechanisms, namely abrasion/corrosion, surface spalling and impact ball breakage, as
well as the scats generation as part of the total consumption in SAG mills.
The influence of key mill operating conditions – such as ball filling, discharge grate opening, mill
speed, lifter profile, and ball quality – on their corresponding grinding media consumption rates is
explored via a variety of model simulations, discussing in every case the independent contribution
of each of the indicated consumption mechanisms to the estimated total ball consumption rate.

1
INTRODUCTION
A thorough understanding of the independent contribution of each grinding media wear mechanism
is extremely important, especially in semi-autogenous grinding (SAG) operations where macro wear
(spalling) and ball breakage cannot be ignored or neglected. The high impact conditions imposed in
modern SAG milling environments has created serious challenges for grinding media manufacturers
since their products must be able to perform with minimal breakage and surface spalling.

GRINDING MEDIA WEAR MECHANISM

For SAG milling operations, at least 3 different mechanisms have been identified as the main
contributors to the observed overall grinding media consumption, here referred to as: abrasion /
corrosion, ball breakage and surface spalling.

Wear by Abrasion/ Corrosion

The abrasive/corrosive wear mechanism has been discussed in depth in previous publications
(Benavente, 2007), (Guzman Levi, 2014), (Sepulveda & Morrow, 2014). The most widely accepted
approach to mathematically characterize the consumption (wear) kinetics of grinding media in rotary
tumbling mills is known as the Linear Wear Theory (Prentice, 1943; Norquist, 1950; Sepulveda, 2004),
which claims that – at each instant 't' after a grinding body enters the mill load – the rate at which the
body loses weight is directly proportional to its surface area exposed.
𝜕(𝑚)
𝛺𝑡 = = −𝑘𝑚 𝐴𝑏 (1)
𝜕(𝑡)
where:
Ωt = media consumption rate, kg/h
m = ball weight, kg; after t hours in the mill charge
Ab = surface area of the ball exposed to wear, m2
km = mass wear rate constant, kg/h/m2.
Taking into account the geometry of the grinding body (normally a sphere), Equation 1 converts to:
𝜕(𝑑) −2𝑘𝑚
= = −𝐾𝑑 (2)
𝜕(𝑡) 𝜌𝑏
where:
d = size (diameter) of the grinding body, after t hours in the mill charge, mm
ρb = density of the grinding body, gr/cm3 or ton/m3
kd = linear wear rate constant, mm/h.

Equation 2 above can be simply integrated for the most common case when k d remains constant as
the ball wears off; that is, the ball is sufficiently homogeneous from surface to center and k d is not a
function of time (or diameter); the condition required for linear wear kinetics:
𝑑 = 𝑑 𝑅 − 𝑘𝑑 𝑡 (3)
where:

dR = make-up size of the new grinding bodies periodically recharged to the mill, mm.

2
Ball Breakage
As originally proposed by the US Bureau of Mines (Blickensderfer &
Tylczak, 1985; Blickensderfer, Tylczak & Madsen, 1985) Drop Ball
Testing (DBT), (see figure 1) has been the experimental methodology
selected for the characterization of alternative grinding media product
attributes. As implemented around the Moly-Cop organization –
world-leading manufacturer of the full range of grinding media sizes
– such DBT. The methodology consists of repetitively dropping an
ensemble of N balls (one at a time) from a standard height of 10 m for
a cumulative minimum of 1000 drops of each ball (or until more than
10 balls get broken) (Sepulveda & Morrow, 2014).
The impact severity conditions so created in the DBT apparatus are
intended to be representative of the high impact environments in large
diameter SAG balls (ball-to-ball or ball-to-liner impacts). The target of
1000 drops per ball has been specified to represent approximately half
of the life of a ball in the mill charge.

Figure 1 The DBT J Tube

The main result of a properly conducted Drop Ball Test is the DBT Breakage Index defined as:
#Broken balls
𝐷𝐵𝑇𝐵𝐼 = ∗ 1000 (4)
Total # Drops

Surface Spalling
An equally important result of a DBT is the assessment of the rate of weight loss experienced by the
balls due to surface spalling after repetitive impacts. Since spalling has been observed to be
accelerated as the ball accumulates impacts (surface fatigue phenomenon), it is required to divide the
test into multiple sequential stages stopping at say 20,000, 40,000, 60,000… total drops, determining
the average weight of the remaining unbroken balls after each test interruption.
The contribution of surface spalling to the weight loss of the balls may then be characterized by the
DBT Spalling Rate, kdS defined as:

Incr.diameter loss, mm
𝑘𝑑𝑠 = Incr.#Drops × 1000 (5)
𝑁

where the ‘Incremental Diameter Loss’ between two consecutive test stops (k and k-1) may be simply
determined from the measured variation in average ball weight by the expression:

6𝑤𝑘−1 1⁄3 6𝑤 1⁄3

(𝑑𝑘−1 − 𝑑𝑘 ) = ( ) − ( 𝜌𝜋𝐾) (6)
𝜌𝜋
with wk representing the average weight of the remaining balls after the kth test interruption.
The DBT Spalling Rate, kdS, after DOSB (drops of the same ball) has been experimentally found to
satisfactorily conform to the power relationship:

3
 𝑘𝑑𝑠 = 𝐾0𝑠 × DOSB 𝜑 (7)
where k corresponds to the surface spalling rate after accumulating exactly 1000 DOSB’s and φ is
0S

an exponent accounting for the surface ‘fatigue’ effect; that is, the continuous increment in k dS as the
balls accumulate repetitive impacts. Figure 2.

Figure 2 Experimental DBT results illustrating the significant increment in surface spalling rate as the balls accumulate
repetitive drops through the DBT J-tube

The observed surface ‘fatigue’ effect is reflected by fairly high values of the slope φ. Interesting to
note that all types of different ball qualities tested so far appear to conform to similarly high
magnitudes of the exponent φ, such as the 2.38 value shown in Figure 2 above.

MATHEMATICAL MODEL FORMULATION

The theoretical considerations presented in the previous section provide the basis for the formulation
of a mathematical model, claimed to be applicable to any actual industrial SAG milling operation,
which includes and combines the independent effects of each of
the three identified media consumption mechanisms: Wear, Ball
Breakage and Surface Spalling. (Morrow & Sepulveda, 2015).
This model allows to estimate (depending on the geometry of
liner), the lifting capacity of the lifters; that is, the volume of
charge (balls-rocks-slurry) being lifted by each lifter bar per
revolution, as shown in the schematic representation of a cross
section between two neighboring lifter bars. Figure 3.

Figure 3 Lifter configuration

4
 Clearly, not all the balls being lifted by the action of the rotating
mill would suffer high-energy drop impacts comparable to those
created in the DBT device. Figure 4.
The vast majority of the balls being lifted will cascade down the
charge kidney without considerable damage and only a small
proportion of them will cataract with high-energy against other
balls by the kidney’s toe (ball-to-ball impacts) and even against
the exposed liner bars and plates (ball-to-liner impacts), this type
of severe impacts are here defined as ‘Critical Impacts’.

Figure 4 DEM Simulation into SAG Mill

MODEL CALIBRATION
Assuming the so defined critical DOSB impacts are comparable to standard DBT impacts, the
influence of the various media consumption mechanisms may be quantified as follows:
Diameter (mass) loss by wear:

d(t) = d(t-Δt) – kdΔt ; with kd constant with time t

Diameter (mass) loss by spalling :

d(t) = d(t-Δt) – kdSΔt ; with kdS = k0S * (DOSB/1000) φ

# of Ball Failures by breakage:

# BrokenΔT = (DBT Breakage Index)*(DOSBΔT/1000) * (d(t)/d(0))4

Where the (d(t)/d(0))4 factor accounts for the age of the ball in the mill, making it less prone to breakage
as the ball gets older and smaller.The calculation algorithm incorporated in the spreadsheet Media
Charge Wear & Spalling & Breakage of Moly-Cop Tools 3.0 (Sepúlveda, 2012). Allows for the
evaluation of the rate at which every independent ball size in the mill charge string is being
consumed by each of the 3 identified mechanisms plus the amount of steel rejected off the mill as
ball scats (cores). That is, those balls that gradually lose diameter down to the size of the mill
discharge grate slots without suffering massive ball breakage.
The calculation algorithm permits the evaluation of the influence of key operational conditions on
total media consumption and most importantly, permits to disengage and independently analyze
the contribution to total consumption by each of the three main identified mechanisms.
For the purposes of calibrating the mathematical model to actual SAG milling practices, the following
reference operational conditions were defined, covering various levels of mill charge impact severity
(volumetric ball/charge ratio, Jb/J) and ball type DBT performance parameters (Breakage and
Spalling). (Table No 1).

5
Table 1

Ball Quality

SAG Mill Impact Spalling

Environment Ball DBT Rate

Severity Jb/J Toughness BI k0s

Low <0.45 High 0.05 0.1

Medium >0.45, <0.6 Medium 0.25 1.0

High >0.6 Low 0.50 2.0

EXPLORATORY – WHAT IF SIMULATIONS

Once the model has been calibrated, it is possible to perform exploratory “what if simulations” about
the effect of operational conditions on grinding media consumption rates. See Figure 5
TM
Moly-Cop Tools (Version 3.0)
MEDIA CONSUMPTION RATES IN SAG MILLS

Remarks Base Case Example : 40'f x 25' SAG Mill.

The following operating conditions were
Mill Dimensions and Operating Conditions: Power, kW

Eff. Diam.
ft
Eff. Length
ft
Mill Speed
% Critical
Charge
Filling,%
Balls
Filling,%
Interstitial
Slurry
Filling,%
Lift
Angle, (°)
13,373
3,221
2,406
Balls
Rocks
Slurry
considered as a basis for simulations:
40.00 25.00 73.00 25.00 15.00 65.00 36.5 19,000 Net Total
L/D 0.63 8.84 rpm 5.00 % Losses
% Utiliz. hr/month 20,000 Gross Total
95.00 684.0 13,680 MWh/month

 Diameter (ft)
% Solids in the Mill 75.00 %
ton/m3
Ore Density,
Slurry Density,
Ore Feedrate,
2.80
1.931
2000.0
ton/m3
ton/hr
Charge
Volume,
Mill Charge Weight, tons
Ball O´size Interstitial
Apparent
Density
: 40´
 Length (ft)
45,600 ton/day m3 Charge Rocks Slurry ton/m3
Energy, (gross) 10.00 kWh/ton (ore) 222.84 621.72 149.75 111.88 3.964
Balls Density,
Ball Size,
Scrap Size,
7.75
125.00
50.0
ton/m3
mm
mm
: 25´
Liner Design :
# of Lifter Bars
Mill Speed,
40
354
Def ault Value
40
lifters/min
 Critical Speed : 75%
 Total Charge
Lifters Spacing, L0 37.70 inches
, L1
, L2
26.24
33.70
inches
inches
: 25
 Ball Charge
Lifter Height, hL 8.00 inches
Lifter Width, wL
Lifter Face Angle,
Forward, f1F
4.00

25.00
inches

(°)
: 12%
 G. Power, kW
Backward, f1B 25.00 (°)
Load Angle, f2
hx
45.00
22.98
(°)
inches
: 20,500
Lifting Cavity Filling, m3/lifter
Lifting Cavity Filling, %
0.084
7.1
Lifting Capacity :
Total Balls & Rocks, 1,782 m3(ap)/hr
 Throughput : 5,000 t/h
m3(ap)/hr

 % solids
Voids Fraction in Lifting Cavity, % 35.0 Rocks Lifting Rate, 713

Angle at Balls Release, wrt Vertical

Angle at Balls Impact, wrt Vertical
45.0
45.0
(°)
(°)
Balls Lifting Rate,

# of Lifts on same Ball (LOSB)

1,069
5,385
8.7
m3(ap)/hr
ton/hr
#/hr
: 75
Cushioning Constant, g0

 Ai
0.10 0.10 << Def ault Value
Equiv. DBT Height,

WEAR Rate Constants:

10.28 m Cushioning Constant, g1
Cushioning Factor
# of Critical Lifts on same Ball
1.00
0.060
0.5 #/hr
1.00 << Def ault Value
: 0.25
 pH
Def ault Values LOSB = Lifts on same Ball
Bond's Abrasion Index
Fresh Feed F80, mm
Slurry pH
0.25
125000
10.0
0.22
100000
10.5 BALL CONSUMPTION RATES
:9
k dB 1.362 1.292 gr/kWh gr/kWh
k dE (UPDATE)
kd
1.741
0.0374
mm/[kWh/ton]
mm/hr
gr/ton (gross) (balls) kg/hr

Ball Consumption by WEAR

ton/month % of Total
 F80 : 125mm
 Ball Diam
BREAKAGE Parameters: 352.5 35.25 52.73 705.1 482 89.1

Total
# of Drops
# of Balls
in Test
DBT Test Results
Drops on
same Ball
Broken
Balls
°/ 1 Broken
per k DOSB 43.2
Ball Consumption by BREAKAGE
4.32 6.46 86.4 59 10.9
: 5.0”
 Scats
20,000 35 571 10.00 0.50
Corrected for Equiv. DBT Height,
DOSB = Drops on same Ball
SPALLING Parameters:
0.51
0.1
Ball Consumption by SPALLING
0.01 0.01 0.1 0 0.0 : 50 mm
 kdE µm/kWh/t
Ball Cores (SCATS) Rejects

: 1.967
DBT Test Results 0.0 0.00 0.00 0.0 0 0.0
Total # of Balls Drops on mm (loss)/ Fatigue
# of Drops in Test Same Ball 1000 Drops Exponent OVERALL BALL CONSUMPTION RATES
20000 35 571 0.07 2.38 395.8 39.58 59.20 791.6 541 100.0

Spec. Area, m 2/m 3 (app) 36.56 Total Charge Area, m 2 4844 Purge Time, hrs 2,000

Figure 5 Moly-Cop Tools Grinding media Simulator

Effect of SAG Mill Impact Environment (Jb/J)

The effect of the ball consumption was simulated as a function of SAG Mill environment (J b/J) and
the Ball Breakage Index (BI).

6
 Figure 6 shows the strong effect of the Jb/J ratio on
overall media consumption. As expected, such
effect gets to be more significant as the toughness
decreases (BI of the balls increases).
Figures 7, 8, 9, and 10 represent the independent
contribution of each consumption mechanism. It
can be noticed that in all cases, the main contributor
is the wear mechanism, more than any of the other
3 mechanisms together.

Figure 6 Overall consumption

Figure 7 Consumption by Wear Figure 8 Consumption by Breakage

Figure 9 Consumption by Spalling Figure 10 Scats Generation

7
 As a function of the Spalling Rate (k0S), Figure 11
shows that higher grinding media consumption is
obtained in all cases at higher spalling rates, k0s
Figures 12, 13, 14 and 15 represent the contribution
of each wear mechanism to the overall grinding
media consumption. It can be noticed that, after
wear, the higher contributions are obtained by
spalling and breakage mechanisms.

Figure 11 Overall consumption

Figure 12 Wear Consumption Figure 13 Breakage Consumption

Figure 14 Spalling Consumption Figure 15 Scats Generation

8
Effect of SAG Mill Critical Speed
The effect of the SAG Mill Critical Speed (%) were
simulated, considering medium toughness
grinding media quality. Figure 16, shows that
higher consumptions are obtained at higher levels
of critical speeds.
Figures 17, 18, 19 and 20 show the contribution of
each independent mechanism. It may be noticed
that the spalling rate increases as mill speed
increases.

Figure 16 overall consumption

Figure 17 Wear consumption Figure 18 Breakage consumption

Figure 19 Spalling consumption Figure 20 Scats Generation

Effect of Lifter Profile

One of the main design conditions that is usually modified in SAG milling operations is the lifter
profile, with the main objective to increase the availability and increase the life of lifter. However an
analysis of the impact on grinding media consumption and performance is usually not evaluated.
Figure 21 shows the ball trajectory for an old lifter profile of 8.0” height and 25° face angle against a
new lifter design of 12” height and 30° face angle. It can be noticed that the new design projects a less
aggressive trajectory and impact energy would be almost the same. One could even think that
increasing critical speed could be beneficial.
The new lifter profile, because of its higher lifters, increases its lifting capacity, meaning that critical
impacts will increase substantially as shown in Table 2.

9
 Figure 21 Ball Charge trajectory – lifter profile comparison

Table 2 Lifting capacity analysis

Lifter Face Angle, ° 25 30

Lifter Height (in) 8 10 12 8 10 12
Balls Lifting Rate (t/h) 5,928 9,263 13,338 4,695 7,336 10,563
# of Critical Lifts on same Ball 0.51 0.80 1.15 0.40 0.63 0.91

Simulations of lifter profile effect on overall media

consumption is shown in Figure 22. Estimate results
show that higher consumption rates are obtained
using higher lifter heights at lower face angle.
Figure 23, 24, 25 and 26 show the contribution of
each mechanism to overall consumption.

Figure 22 Overall grinding media consumption

Figure 23 Wear consumption Figure 24 Breakage consumption

10
Figure 25 Spalling consumption Figure 26 Scats generation

CONCLUSIONS
The analysis performed using the new Moly-Cop Model is very useful for identifying the effect of
operating conditions as mill speed, ball charge, total mill level, and their effect in grinding media
consumption.
More importantly, this new methodology identifies the main media consumption mechanisms –
wear, breakage and spalling – in relationship with ball toughness and operating conditions, so it is
recommended to include this methodology in evaluations of alternative SAG media products, in
order to best anticipate their grinding media cost-effectiveness.

REFERENCES
Internal Reports, Moly-Cop Mexico, Moly-Cop Peru, 2015.
Latchireddi, S. (2014). Operational Optimization of AG/SAG Milling Circuits Shortcourse. IMEC
2014, San Luis Potosí, Mexico.
Sepulveda, J. E. & Morrow, R. (2014). Moly-Cop Tools Applications for the Assessment of Grinding
Media Performance at Full Industrial Scale. 46th Annual Canadian Mineral Processors Operators
Conference, Ottawa, Canada.
Guzman, L (2014), Updated Benavente Correlation for Estimating Grinding Media Consumption
Rates, XXVII IMPC, Santiago, Chile.
Sepulveda, J. E. (2004). Methodologies for the evaluation of grinding media consumption rates at full
plant scale. Minerals Engineering, vol 17, pp 1269–1279
Sepulveda, J. E. (2012). Moly-Cop Tools, Version 3.0: Software for the assessment and optimization
of grinding circuit performance. Available upon request at [email protected]
Benavente, H. (2005). Correlación empírica para la evaluación de consumos de bolas en aplicaciones
de molienda. Moly-Cop 2005: X Simposio sobre Proces

11
The Role of HPGR Recycle in Dry Minerals
Processing Circuits
Frank van der Meer1 and Javier Valdeavellano2
1. Weir Minerals, The Netherlands
2. Weir Minerals, Chile

ABSTRACT
Dry HPGR processing including fine particle recovery by air classification has entered the minerals
processing landscape, especially for reasons of project feasibility and environmental sustainability,
but also to generate an operable flow sheet where very fine sizing by dry screening would be less
favorable.
High Pressure Grinding Rolls (HPGR) in closed circuit arrangement with classification demonstrate
to be an applicable technology for a wide range of fine grinding applications like industrial
minerals or iron ore processing, even more so in the growing demand for reduction of (potable)
water addition and consumption.
Combined with a realistic reduction in energy consumption, the current HPGR equipment sizes can
serve the generation of materials at a very fine product size distribution at an unpreceded high unit
throughput rates.
This publication summarizes some of the specific challenges for dry HPGR processing, and
discusses experiences with some present applications and studies for near-future project.

1
INTRODUCTION
HPGR, as a feasible crushing and grinding technology in minerals processing, does in most cases
rely on effective classification to generate a suitable product for downstream processing.
Classification in combination with HPGR can be achieved in various ways.
If the feed size distribution from preceding secondary crushing generates a significant proportion
of fines or final product, a truncation of the HPGR feed may be considered. In that case, the HPGR
effective feed will display a steep size distribution and a low bulk density. In other cases than a
single, pass operation, an effective HPGR feed size distribution will evolve as combination of the
fresh feed and a recycle stream.
For a closed circuit with screening, a feed partly deficient of fines but richer in intermediate size
fractions will result, at an associated bulk density. In a PPR (Partial Product Recycle) circuit, where
a portion of the HPGR discharged is returned to the head of the circuit, a very broad effective
HPGR feed size distribution is generated, with a high proportion of fines [van der Meer,
Gruendken 2008].
Each case, especially using a dry system, does have a series of implications on HPGR operation,
wear and grinding efficiency.
Present day simulations of HPGR circuits do, for an important part, relate to the size reduction
effects, and the energy required for various circuit arrangements. For the sizing and scaling of
associated equipment (classifiers, conveyors, feeders and bins) however, the unit capacity and
effects of circulating load on HPGR feed size distribution and performance are important factors.

CLASSIFICATION AND RECYCLE; IMPLICATIONS FOR HPGR

Different classification methods, or a different classification cut size, do reflect on the effective
HPGR feed size distribution. In Figure 1 and Table 1, a number of different effective HPGR feed
size distributions are given as example, based on a common plant feed for a medium hard copper
ore.
Table 1 Summary Comparison of HPGR Effective Feed Particle Size Distributions

F80, mm F50, mm F20, mm F10, mm

Plant Feed 35.0 18.0 6.5 3.2

Truncated 38.0 22.0 10.0 7.5

HPGR Feed 8.00 mm 31.0 16.0 6.5 2.6

HPGR Feed 1.00 mm 25.0 9.0 2.4 1.3

HPGR Feed 0.25 mm 17.0 3.1 0.5 0.3

HPGR Feed 0.10 mm 10.0 0.6 0.1 0.1

PPR 100 % 25.0 8.0 0.5 0.1

2
 Figure 1 Comparison of HPGR Effective Feed Particle Size Distributions

A truncated feed, based on a truncation size of about 8 mm, increases the P50, P20 and P10 sizes
and diminishes the fines proportion in the feed. In plant conception, a truncation resulting in e.g.
25% or more material mass bypassing the HPGR might be assumed to provide advantages in
reducing equipment size (CAPEX and OPEX), as less material would need to be handled by HPGR.
When processing the truncated feed in single pass or closed circuit with screening, however, this
will have a clear effect on both specific throughput and specific energy consumption, whilst also
affecting the wear rate.
For a truncated feed as above the increase in void space in the HPGR feed results in reduction in
feed bulk density and as consequence effects a reduction in the operating gap and specific
throughput. Depending on material size reduction and applied pressure-energy input, this may not
lead to smaller equipment, and may result in an overall HPGR circuit product that has fewer fines
from a re-combination of the bypass flow and HPGR section product, at a similar or higher specific
energy input. Considering downstream grinding requirements and advantages in HPGR product
weakening, the overall balance may be that for, for instance, a HPGR-ball mill circuit the overall
energy input and capital expenditure may be higher for a truncation material bypass. Excessive
fines in the feed, on the other hand, may reduce HPGR grinding efficiency and may well warrant
some type of truncation of the feed.
For circuits involving classification at different cut size levels, the effective HPGR feed will display
different material size compositions. The different material size compositions will theoretically lead
to different maximum closed packing particulate densities. A relative over-representation of
particle size fractions in the mid-range, above the screen aperture size and below the HPGR
operating gap or P80 size, will result in a different compressibility of the material. Thus the mass
transport between the rolls, as result of a different bulk density, operating gap and material

3
compression, will be affected. Depending on phenomena such as packing arrangement and particle
shape, such different particle size distribution will in most cases result in different HPGR mass
transport phenomena (slippage, compaction, gap opening), and this will generally significantly
affect the specific throughput and energy consumption. Concurrent with the above, the size
reduction may be affected either from hindrance of a high proportion of mid-sizes in the feed (finer
classification), or from promotion of coarser particles crushing and generation of micro-cracks
when running at a lower fines proportion, e.g. at a coarser classification where less cushioning of
coarse particles by fines takes place.
In an extreme case, a circuit operating as PPR (Partial Product Recycle) can be applied. Such a circuit is
applicable for instance in cases where a high level of moisture or a required fine product size does
prevent dry screening, or where a strong flake formation does render material handling in
classification difficult. A PPR arrangement will effectively create a HPGR feed that is build-up from a
relatively large proportion of fine recycle material blending-in with fresh feed. This material originates
from finished or nearly finished material from recycled edge material and a proportion of the center
discharge. The high void filling of the HPGR feed, possibly also including flakes or flake fragments
carried with the HPGR discharge material, can potentially increase energy consumption and lower
specific throughput, and potentially could reduce the effective size reduction at higher recycle ratios.
In addition to the difference in effective circulating load, as consequence of cut size selection, the above
effects do have significant implications for HPGR sizing and selection. Additional effect will be, for
instance, that product handling such as bunker flow and intermediate material storage could be
affected from different HPGR feed and product size distributions, especially in view of potential effects
of contained moisture.
HPGR testing of copper ore particle compositions with different proportions of “fines” (< 3.15 mm) did
indicate that the specific throughput maximized at mid-range fines content. It is considered to fall in
line with an optimum pressure resistance in the particle bed from an optimum void filling by finer
particle sizes. A lower proportion may reduce this due to a higher void space and lower bulk density in
the feed, and a higher proportion of fines could lead to more slippage and therefore a smaller operating
gap. Flake strength maximized in line with the trend for specific throughput, thus confirming a
pressure resistance optimum. The net specific energy consumption reduced from the combined effect
of a lower achieved pressure level at a reduced operating gap. Supposing comparable grinding model
parameters (which actually do differ for the different feed and circumstances), simulation calculations
on actual test basis do display different product particle size distributions and reduction ratios for
different effective HPGR feed size distributions. This is illustrated in Table 2 and Figure 2.

4
Table 2 Summary Comparison of HPGR Discharge Particle Size Distributions

Trunc’d
Closed Circuit Full Feed Full Feed Full Feed Full Feed Full Feed
Feed

Type 8 mm 8 mm 1 mm 0.25 mm 0.10 mm Circuit

Dry Dry 100 %

Screening Screening Screening
Class'n Class'n PPR

Specific Throughput,
210 245 255 260 255 275
ts/hm³

HPGR Load,
116 155 270 345 450 200
% of fresh feed

P80, mm 14 13 11 4.2 1.45 8.0

P50, mm 4.3 3.4 2.0 0.57 0.18 1.25

P20, mm 0.73 0.42 0.23 0.09 0.06 0.09

RR 80 % passing size 2.7 2.3 2.2 3.8 6.6 3.1

RR 50% passing size 5.3 4.7 4.5 5.3 3.3 6.4

RR 20% passing size 18 19 10 5.9 2.4 6.7

5
 Figure 2 Comparison of HPGR Discharge Particle Size Distributions

The above reduction ratios (RR) are given comparing effective HPGR feed with effective HPGR
discharge. Obviously, when the comparison is done based on fresh feed, much higher numbers do
appear, reflecting the classification cut sizes (Table 3).

Table 3 Summary Comparison of Fresh Feed Based Reduction Ratio’s for HPGR discharge

Trunc’d
Closed Circuit Full Feed Full Feed Full Feed Full Feed Full Feed
Feed

Type 8 mm 8 mm 1 mm 0.25 mm 0.10 mm Circuit

Dry Dry 100 %

Screening Screening Screening
Class'n Class'n PPR

RR 80 % passing size 2.5 2.7 3.2 8.3 24.1 4.4

RR 50% passing size 4.2 5.3 9.0 31.6 100.0 14.4

RR 20% passing size 11.0 19.0 34.8 88.9 137.9 88.9

Furthermore, the reduction ratio’s comparing fresh HPGR feed with final HPGR circuit product do
show even higher numbers (Table 4).

6
Table 4 Summary Comparison of Fresh Feed Based Reduction Ratio’s for HPGR final product

Trunc’d
Closed Circuit Full Feed Full Feed Full Feed Full Feed Full Feed
Feed

Type 8 mm 8 mm 1 mm 0.25 mm 0.10 mm Circuit

Dry Dry 100 %

Screening Screening Screening
Class'n Class'n PPR

RR 80 % passing size 6.4 7.0 53.8 250.0 700.0 4.4

RR 50% passing size 9.0 12.0 72.0 257.1 720.0 15.0

RR 20% passing size 22.9 32.0 114.3 400.0 800.0 88.9

The implications from the above are that selection and sizing of HPGRs is affected by both physical
parameters (specific throughput, gap opening, specific energy consumption, circulating and conveyor
load) and size reduction efficiency.
The high reduction ratio at fine product sizes for dry classification do warrant a serious consideration
of this technology in applicable fields, for achieving a balance between product quality for overall
circuit performance at a given HPGR throughput and energy consumption. In this, moisture and
HPGR product desagglomeration effects in classification are to be considered as well.

CLASSIFICATION AND RECYCLE; CASE STUDY

was carried out a simulation case for the VasGold operation (Kazakhstan) for evaluating a potential
expansion of their present processing plant in Altyntau and increase in capacity of the final
crushing stage of the gold ore. The present operation is running at 8 Million t/y [van der Meer,
Romanchenko, Ibrayev, 2010], and an expansion is envisaged to 10 Million t/y.
The present circuit involves primary, secondary and tertiary crushing to a HPGR feed top size
about 30 mm. From this, two parallel processing HPGR lines at each 530 t/h take the material down
to a P80 size of about 5 mm, in closed circuit with a relatively large product circulation
arrangement. This product then is fed to a reverse circuit of cyclone classification and ball milling.
Ball mill circuit final product would be 90 % < 74 µm.
The choice for a product recirculation flow sheet (with a presently employed average circulating
load of 170 %) was based on measurements and observations of the initial ore being wet and
containing a significant proportion of clayey material. This upon HPGR compression generated a
product with a significant proportion of flakes and agglomerates, which would be difficult to
classify by screening, and require undesired large screening facilities.
The present operation and envisaged expansion would provide the plant with a more competent
ore from deeper levels, and with less clayey material. This would allow modifying the plant process
to incorporate a closed circuit HPGR operation with (wet or dry) HPGR discharge screening.

7
Was carried out a test program to compare effects of processing the envisaged ore in closed circuit
with screening and product recycle [KHD Humboldt Wedag 2012]. For this, test runs were carried
out using a product recycle with 50 % circulating load and 170 % circulating load, and an approach
with 5 mm screening. Following this, scale-up and simulation was carried out for situations with a
55 %, 110 %, and170 % circulating load, and for the cases closing the circuit with a 5 mm or 8 mm
screening.
The product recycle options did return a relatively wide particle size distribution with a high
proportion of fines (Table 5, Figure 3).

Table 5 Product Particle Size Distributions for Different Circuit Options

Size, Fresh PPR PPR PPR 5 mm 8 mm

mm Feed 55 % 110 % 170 % screen screen

Cum % Cum % Cum % Cum % Cum % Cum %

passing passing passing passing passing Passing

31.50 100.0 100.0 100.0 100.0

25.40 92.0 98.3 99.3 99.8

19.00 71.0 96.5 98.0 99.0

16.00 54.0 94.9 96.7 97.9

12.50 39.0 90.9 93.7 95.4

9.50 25.0 82.6 87.5 90.9 100.0

6.70 14.50 69.7 77.6 83.6 95.0

5.00 10.00 58.0 69.0 78.0 100.0 85.0

3.35 7.0 46.7 57.9 67.5 81.0 72.5

2.36 6.0 38.0 49.6 59.8 70.0 63.0

1.18 4.5 26.0 36.9 46.7 53.0 48.0

0.600 3.5 18.4 28.2 36.4 40.0 36.0

0.300 2.6 12.2 21.5 27.8 28.0 25.0

0.150 2.0 7.2 16.1 20.7 18.0 16.0

8
 0.075 1.6 4.0 11.3 14.8 11.8 9.0

P80, mm 19.0 8.9 7.4 5.7 3.2 4.4

Recycle, % 55 110 170 47 30

Figure 3 Product Particle Size Distributions for Different Circuit Options

The product size became finer with a higher recycle, from 58 % < 5.0 mm at the 50 % recycle to 78 % <
5.0 mm at the 170 % recycle, with a decreased in P80 size from 8.9 mm and 5.7 mm. The trends
observed do suggest that the relationship between recycle load and product size is a digressive one,
with product size reducing less than proportional to the recycle load. The screen undersize product did
show a steeper size distribution slope, with a P80 size between 4.4 and 3.2 mm.
An increase in recycle proportion, from 50 % to 170 % (% of fresh feed) resulted in a decrease in specific
throughput, from 260 ts/hm³ to 235 ts/hm³, and an increase in specific energy from 2.2 kWh/t to 2.5
kWh/t (Table 6). Applying a closed circuit with screening instead of product recycle did show a specific
throughput at an average level of 265 ts/hm³, as compared to 235 ts/hm³ for the 170 % product recycle
conditions, at a lower net specific energy of 1.85 kWh/t, as compared to 2.5 kWh/t. The HPGR
discharge product size coarsened, but the screen undersize product did show a fine size distribution of
100 % < 5 mm or 8 mm, as compared to the product recycle. Projected at a plant operation with 8 mm
screening, the screen undersize is expected to be about 80 % < 4 mm, well within the envisaged plant
ball mill feed specifications.

9
Table 6 Comparison Parameters for Different Circuit Arrangements

Circuit Type PPR PPR PPR 5 mm 8 mm

55 % 110 % 170 % Screen Screen
Specific Throughput, ts/hm³ 260 245 235 267 260
Net Specific Energy, kWh/t 2.2 2.4 2.5 1.9 1.8
HPGR Load, % of fresh feed 55 110 170 47 30
Overall Net Specific energy, kWh/t 3.4 4.9 6.9 2.7 2.3

Circuit Product
% < 5.0 mm 58.0 69.0 78.0 93.0 85.0
P80, mm 8.9 7.4 5.7 3.2 4.4

When considering this, and also given that both the total energy consumption and further operating
costs will tend to rate in proportion to the recirculating load, a lower recycle will benefit the operating
cost almost linearly, and operating at a lower recycle should be beneficial as long as the product size
achieved does match the downstream process requirements.
The implications from the above could be that, if moisture and flake strength would provide an
adequate screening efficiency and operation, a better product at a higher overall throughput and lower
energy consumption could be achieved by using a conventional closed circuit with dry or wet
screening. Obviously, when selecting for a screening process, this does require a capital investment in a
screening section.
Additional effect will be that a better product handling (bunker flow and effective size reduction) will
result from a HPGR feed containing less fines, and holding less moisture.

ASPECTS OF DRY CLASSIFICATION. SEGREGATION EFFECTS

Air classification further expands the practical use of HPGR into a fine particle production range
[Altun, Benzer, Dundar, 2016]. In these, HPGR could be arranged in combination with dry minerals
classification and processing, such as low intensity magnetic separation, for size ranges between
1,000 microns down to 10-30 microns for massive fine grained iron ores, limestone (minerals filler)
and breaking ground for applications in gold, copper and other minerals.
Basic elements for an integrated system involve an HPGR with a static classifier (cross flow) and
subsequent dynamic separator, followed by air cyclones and a bag-house for fines recovery.
Product specification for such application allows generating a fine product, tuned to achieve market
specifications. The classifiers operation is sensitive to generate a product of tightly defined
specification over the full width of the product size distribution and throughput.
Results of air classification depend on air flow, dynamic separator rotation speed, moisture, and
temperature, with Mass-to-Air loading as criterion for the fine classifier sizing and downstream
recovery units such as bag-house or scrubber. Depending on the classifier efficiency and associated
fish hook effect, a bypass and fines recycle mass will establish, which will determine the quality and
mass proportion of the recycle flow.
In dry classification, segregation of minerals may present an additional phenomenon to take into
account. The effect is more widely known in wet minerals processing such as in gravity treatment

10
of ores. The origin is that smaller grains of heavy minerals will display similar classification
properties and larger minerals of a lower density. In dry classification using static or dynamic air
classifiers, segregation does take place as consequence of this. Consider for instance a two-stage
classification of iron ore with a first stage at about 100 microns and a de-sliming at 10 microns. The
initial cut for a product size will, depending on liberation characteristics of the material, reject the
heavier but finer iron mineral values into the recycle stream, together with coarser gangue
minerals, whereas the size distribution of the recovered fine product will tend to show a higher
fineness of the iron minerals against a background of a relatively coarse gangue particle size
distribution. The effect thus is an enrichment of iron minerals in the recycle stream, and a potential
risk of over-grinding the iron minerals, rendering these less amenable for down-stream upgrading.
In addition, if a second stage, classification would be used for e.g. de-sliming (ahead of flotation
beneficiation); the same effect would cause a leaner slimes product to be rejected as compared to
the feed to this classification. Care should thus be taken in the selection of the classifier cut size
settings and efficiency to maintain an optimum size reduction, classification and beneficiation
result.
In Figure 4 an example of such a circuit balance is given. In this, a circuit is considered for a low
grade iron ore starting from a feed of 80 % < 60 mm, to generate a fine product by HPGR in
combination with static and dynamic classification. Fresh feed considered is 800 tons per hour, and
the final product is required to be de-slimed to allow efficient downstream beneficiation.
Given a plant feed containing 22 % magnetite (Satmagan assay), the circulating load, due to the
enrichment, did show an assay value of up to 31-36 % magnetite. When considering that at least
part of this stems from liberated material, recycling these grains to the grinding stage would incur
the risk of over-grinding the iron minerals and thus lowering the recovery in downstream
processing. In addition, part of the fine product contains liberated barren material of a lower S.G.,
indicating the necessity of removing these by subsequent beneficiation techniques like flotation.

11
 Figure 4 Mass Balance of Iron Ore Fine Air Classification

CONCLUSIONS
HPGR performance depends on the associated classification circuit design. Given different
classification systems, different effective HPGR feed sizes do result, and consequently different
HPGR parameters will be realized. These are considered to be result of a different particle size
distribution, bulk density and material packing. Depending on these conditions and associated
different breakage parameters, different size reduction ratios will establish, and different circulating
loads will result. These are to be considered for circuit selection and optimization.
For dry classification, a high size reduction ratio between fresh feed and classification product was
indicated. This would suggest that, where relatively dry ores or a very fine product is required,

12
circuit design incorporating dry classification may be a very viable option. In this, it should be
borne in mind that, once at liberation size, segregation of minerals depending on their specific
gravity may occur. This may help rejecting barren material, but may also pose a risk in over-
grinding the valuables.

REFERENCES
Van der Meer F.P, Gruendken A. (2010) Flow sheet Considerations for Optimal Use of High
Pressure Grinding Rolls. Comminution 2008. Minerals Engineering (Elsevier) 23, 2010 pp. 663-669

Van der Meer F.P, Romanchenko A, Ibrayev S. (2010) High Pressure Grinding at Vasilkovka Gold.
Paper 23, Proceedings, Procemin 2010. 8-10 December 2010, Santiago de Chile

KHD Humboldt Wedag Test Report Nr. B.09.1117.1.001. (2012) Test Report on HPGR Grinding of
Vasilkovka Gold Ore for the Altyntau Expansion Project. 30-11-2012
O. Altun, H. Benzer and H. Dundar (2016) Closed circuiting the HPGRs: air classification-their
operations and efficiencies. Proceedings, Comminution 2016 Conference. April 11-14 2016, Cape
Town, South Africa.

13
Characterization of Particulate Matter Emitted
During the Primary Crushing Simulation and their
Relationship with Geological Units
Javiera Gerding1 and Óscar Jerez2
1. Centro de Estudios Mineralógicos, GeoAtacama, Chile
2. Instituto de Geología Económica Aplicada, Universidad de Concepción, Chile

ABSTRACT
Mining in Chile has had sustained development during the last few decades, mainly due to the
expansion of existing mining operations and the creation of new centers for mineral extraction. This
process involves the emission of large amounts of particulate matter into environment, particularly
open pit operations, affecting the health of staff working at the site, surrounding communities and
the environment.
The object of this work was to study with a reduced number of geological units, the particulate
matter dependency with the lithology of the materials that is processed in the primary crusher. By
establishing relationships between the processed lithological unit and the amount of the produced
particulate matter, it allows the prediction of episodes of major environmental pollution by
designing Geo-environmental units according to the degree of PM10 emission.
The current work was conducted with samples from a porphyry Cu-Mo deposit, taken from
different lithological units, with different mineralogy, texture and hardness, in order to evaluate
their behavior in the crushing process. The tests were performed on a laboratory scale using a jaw
crusher. The dust collected during the crushing was characterized by chemical, particle size,
mineralogical and morphological analyses. Analytical techniques used included X-ray diffraction,
laser diffraction, light microscopy, scanning electron microscopy and automated mineralogy.
The results showed differences for PM generated for each lithological unit reaching up to 60%
difference for PM collected between different lithological units and showing a direct relationship
with hydrothermal alteration affecting the rock. Differences between the different units processed
were also evident in terms of the mineralogy, mainly for trace elements and heavy metals. It is
therefore possible to establish criteria that can guide the definition of Geo-environmental units,
based on the largest contribution of PM and mineralogical composition of the processed lithological
units.

1
INTRODUCTION
The present study focuses on the difficulties seen nowadays at mining sites and in exposure of
workers to the air pollution generated by ore treatment for metal recovery.
Csavina et al. (2012) published a literature review on air pollution caused by mining operations,
which found that few countries, such as Australia, England and the United States of America, have
addressed this issue.
In Chile, mining has been primarily focused on porphyry deposits, which are mostly exploited
through open pits. This type of mining exploitation is the main source of suspended particulate
matter, generating air pollution, both within mining installations as well as in the vicinity, and
affecting human health, the environment and even the performance of mining equipment.
Nowadays, all mining operations are required to control their dust emissions, either by blasting,
movement of material and/or mineral processing. These emissions must not exceed the quality
standard of 150mg/m3 daily set by the General Secretariat Ministry of the Presidence, the
Environmental Commission of the Chilean Government. If the limit is exceeded, the company
must suspend the operation of the mine. In this regard, it is estimated that some mining companies
must stop their operations during 10% of their working time every year (Lagos, 2001).
In mining, according to the EPA (1998) and the National Pollutant Inventory (2012) in Australia, the
main contaminating processes are crushing, transport of material and unloading.
Krueger et al. (2004) published a study which highlights the importance of knowing the mineralogy
of each particle, based on the interaction that these particles will have with the environment and the
possible atmospheric implications. Gill et al (2006) reported that the characterization of dust
particles from multiple sites and types of emissions can provide a chemical or mineralogical brand,
which allows the determination of the source of origin. In other countries such as Spain, India, and
Australia, some cities have characterized their pollution chemically, mineralogically and
morphologically to correlate it to the source (Rodriguez et al., 2006; Pipal et al., 2011).
The objective of the current study is to determine the relationship between the generation of PM10
and the lithology in different geological units situated in a porphyry deposit during the crushing
process in a laboratory simulation. Once results are obtained, the definition of geo-environmental
units will also be evaluated.

METHODOLOGY
Practical work was based on the collection and subsequent mineralogical characterization of
particulate matter generated through a simulation of the primary crushing process, conducted in 4
separate lithologies from a porphyry deposit. This simulation was performed in triplicate for each
geologic unit, generating twelve samples for analysis, three per unit.

Lithological Samples
Samples were chosen based on the following variables: lithology, alteration and hardness of the
rock. The aim of this selection process was to obtain samples representative of different
mineralogies and, in some cases, samples representing similar mineralogies but having varying
hardness values. For each lithological type, three rock samples of approximately 6,5 kg were
collected in situ from the extraction area in the deposit and were used for simulation.

2
Primary Crushing Simulation
This procedure was conducted at the GEA Institute, using a Fritsch jaw crusher, with a maximum
discharge size of 2 mm.
With the aim of collecting the largest amount of PM generated during the crushing process, a
device was designed to collect these particles through a vacuum pump and a membrane filter with
0.45mm pores. This device was placed in a strategic location within the jaw crusher, which had
been previously determined to generate the greatest emission of “dust”. To ensure the complete
collection of the particles, the “collector” was covered with a plastic sleeve. The dust collected
during this process was separated from the PM10 by centrifugation using Stokes’ law.

Analytical Techniques
The mineralogical, compositional and morphological characterization was conducted in the
laboratories at the GEA Institute of the University of Concepcion and at the CEM of Geoatacama.
Samples were previously described by optic microscopy, and the particulate matter (PM) was
characterized by X-ray diffraction and scanning electron microscopy with automated mineralogy.
X-ray diffraction was measured with the Bruker® D4 Endeavor, operated with Cu radiation and a
Ni Kβ radiation filter; automated mineralogy was conducted by the TIMA (Tescan Integrated
Mineral Analyser) software, and a TESCAN electron microscope (VEGA model).
Williamson et al. (2013) suggest the use of automated mineralogy to best characterize PM in order to
access minority phases that may be important when evaluating the toxicity of the particulate
matter.
The granulometric profile for each PM collected was carried out by analysis in water in the Fritsch
Analysette 22 Microtec Plus laser particle sizer, with a measuring range of 0.08 - 2000 m.

RESULTS AND DISCUSSION.

Petrography
The petrographic analysis yielded information regarding texture and mineralogies present in each
type of sample, indicating different textures (recrystallization in Tuff and Rx. Obliterated) and
different mineralogies for each unit, highlighting the presence of biotite in the andesite unit, and
potassium-rich feldspar in the porphyry unit. These mineralogies were confirmed by DRX.

Laboratory Crushing Test

Dust collection results from the crushing process of each lithological unit are shown in figures 1 and
2, where it can be seen that the unit with the most suspended particulate matter produced was Rx
obliterated along with Tuff, while the unit generating the least amounts of suspended PM was
Andesite.

3
 0,6
0,5
0,4
Grams

0,3
0,2
0,1
0
Rx Andesite Tuff Porphyry
Obliterated
Geology Unit

Figure 1 Graph showing weight of Figure 2: Graph showing grams of

particulate collected during crushing PM10 and PM2.5 for each unit
simulation

The granulometric profile of the collected dust indicates the unit with the greatest quantity of PM10
and PM2.5 is the Porphyry unit. When the data is cross-referenced with the laser granulometry and
PM weights collected at each unit, it can be observed in figure 2 that there is a similarity in the
contribution of PM10 for each unit, emphasizing Andesite and Porphyry for its low contribution to
PM10.
The quantification of PM generated indicates a relationship between hardness of the rock and the
generation of particulate material. As was expected, the lower the hardness value, the greater the
generation of PM, in this case the Rx Obliterated unit. Additionally, there is a relationship between
PM generation and the hydrothermal alteration affecting the rock; the units that generated the most
PM are those associated with dioctahedral mica alterations (muscovite and illite) and clays.
However, if the granulometric profile is analyzed, the particle sizes generated by rocks having
greater hardness values (Porphyry) are smaller than those generated by rocks having smaller
hardness values (Rx Obliterated and Tuff).
Oyarzún and Arévalo (2011) propose a direct relation between rock hardness and the texture
exhibited by minerals under the microscope, going as far as suggesting that rocks containing the
same minerals but having different textures would have very different hardness values.
Based on the suggestion by Oyarzún, the petrographic observation of the samples in the current
study indicates that the Porphyry and Andesite units exhibit textures that infer high interfacial
energy, while the Tuff and Rx Obliterated units exhibit a recrystallization and/or replacement by
mainly quartz and mica, indicating lower free interfacial energy. This indicates a relationship
between texture and dust generation in the mining process, suggesting that samples exhibiting
textural characteristics associated with high interfacial energy will generate lower quantities of
particulate matter when processed, although the PM generated would be finer than the PM
generated by rocks with low interfacial energy.

4
Mineralogy of the PM
The analyses conducted with DRX and automated mineralogy (TESCAN/TIMA) show similar
majority mineralogies in all lithological units, with quartz and dioctahedral micas (illite and
muscovite) being the most significant. Two units stand out, the Andesite unit for the presence of
biotite and the Porphyry unit for the presence of potassium-rich feldspar. Among metallic minerals,
the presence of pyrite, chalcopyrite, molybdenite, chalcocite and magnetite is highlighted in the
Tuff unit. Furthermore, accessory minerals were only detected in two units (Porphyry and
Andesite).
The main mineral phases of the rocks coincide with the main concentrated phases in the suspension
material. Similar to the work by Moreno et al. (2009), quartz and dioctahedral mica were the phases
exhibiting the greatest concentrations among all PM samples. As well, mineralogical differences
were observed that originated from the original sample, which imprints a mark on all particulate
matter generated.
In terms of particle size, both the main phase as well as the trace phase were detected by DRX and
TESCAN/TIMA, the Porphyry unit presented the highest percentage of particulate matter below
2.5um. On average, this unit presented 56% of quartz, chalcopyrite, potassium-rich feldspar, clays
and albite particles in the PM2.5.

CONCLUSION
The results obtained in this study indicate that it is possible to quantify the amount of PM10
generated by different lithological units, and can be differentiated by the quantity of PM generated.
In addition, PM10 can be characterized by DRX and TESCAN/TIMA, and it is possible to even
identify levels of trace minerals present in the PM, as well as being able to determine the particle
size of each mineral phase.
The mineral phases present in the PM differ among geological units, contrasting the different PM
generated, both in main mineralogy as well as in trace mineral phases.
The definition of geo-environmental units can be conducted based on texture, mineralogy and
hardness of the rock. The definition should also comprise all geological units in the deposit,
including those that will not be processed due to sterility. In order to characterize the geo-
environmental units, a DRX analysis and automated mineralogy of the PM collected during the
laboratory crushing of the samples obtained by diamond drilling is needed, in addition to the
granulometric profile by laser diffraction. ICP analysis, in conjunction with the aforementioned
techniques, would be useful for the definition of these units.

ACKNOWLEDGEMENTS
The authors would like to express their gratitude for the collaboration of the staff at the Centro de
Estudios CEM Geoatacama and the GEA Institute.

ABREVIATIONS
PM10 Suspended particles having an aerodynamic diameter of 10 um or less
PM2.5 Suspended particles having an aerodynamic diameter of 2.5 um or less

5
REFERENCES
Csavina, J.,Field, J.,Taylor, M., Gao, S., Landezuri, A., Betterton, E., Sáez, E. (2012). A review on the
importance of metals and metalloids in atmospheric dust and aerosol from mining operations.
Science of the Total Environment 433:58-73.
EPA. (1998). Emissions Factors & AP 42, Compilation of air pollutant emission factors, section 11.24.
U. S. E. P. Agency. United States.
Krueger, B.J., Grassian, V.H., Cowin, J.P., Laskin, A. (2004).Heterogeneus chemistry of individual
mineral dust from different dust source regions: the importance of particle mineralogy.
Atmospheric Environment 38:6253-6261.
Lagos G. (2001). The Impact of Public Policy on Environmental Quality and Health: the case of land
management and planning. Book Review. Resources Policy, 27: 273-278.
Moreno, T., Amato, F., Querol, X., Alastuey, A., Elvira, J., Gibbson, W. (2009).Bedrock controls on
the mineralogy and chemistry of PM10 extracted from Australian desert sediments. Environmental
Geology 57:411-420.
National Pollutant Inventory. (2012). Emission estimation technique manual for mining.
Department of Sustainability, Water, Population and Communities, Australian Government.
Oyarzún, M. y Arévalo, A. (2011). Rock texture and Bwi relationships, El Teniente ore deposits,
Chile. The first aussim international geometallurgy conference. Brisbane. 181 – 185.
Pipal, A., Kulshrestha, A., Taneja, A. (2011).Characterization and morphological analysis of
airborne PM2.5 and PM10 in Agra located in north central India. Atmospheric Environment
45:3621-3630.
Rodríguez, I., Gali, S., Marcos, C. (2006). Cuantificación mineralógica mediante rietveld del aerosol
atmosférico de Oviedo. Macla 6:407-409.
Williamson, B., Rollinson, G., Pirrie, D. (2013). Automated mineralogical analysis of PM10: New
parameters for assesing PM toxicity. Environmental Science Technology 47:5570–5577.

6
Mine to Plant Commissioning in Salobo mine
Jair Alarcon1, Ronaldo Fonseca2, Grazielle Gontijo3 and Stephen Jeric4
1. Process Plant Area/Technology Solutions/Orica, Latin America
2. Process Area/ Salobo/ Vale, Brazil
3. Process Plant Area/Technology Solutions/Orica, Brazil
4. Marketing Area/Global Marketing Team/Orica, Global

ABSTRACT
The experience of the authors supports the “Mine to Plant” concept that blasting fragmentation can
have a positive impact over the processing plant performance. However, it is very common in
industry that when commissioning new plants, the “Mine to Plant” approach is postponed until the
plant is running in a steady state condition. Due to the high complexity in the technical analysis to
determine the corresponding impact of each initiative initiated in parallel of the plant start-up. By
focusing the targets of the mine and plant departments to execute “Mine to Plant” throughout all
stages of the operation, the Vale Salobo copper mine in Brazil has effectively demonstrated this
concept through the successful implementation of this methodology.
In Q2-2012, the Vale Salobo I (12 Mtpy) initiated its operations. In Q2-2014, Salobo II (24 Mtpy)
plant expansion was implemented. Vale began Q1-2015 with the commissioning of the “Mine to
Plant” Methodology. This paper presents the results to increase the plant throughput
corresponding to the first stage of “Mine to Plant”, validating the impacts in every comminution
steps: primary crushing, secondary crushing, HPGR and grinding. The key sizes to improve the
blasting fragmentation curve in the initial steps of the project: top size, P80 and P50 to minimize the
primary crushing delays due to over size material impacting downstream the entire process chain.
In the first 11 months of the “Mine to Plant” implementation, it was determined that the blasting
fragmentation increased the throughput 13% in primary crushing, 10% in the secondary crushing
and 2% in HPGRs and grinding circuits. Currently work is ongoing to build empirical models that
establish the relation of the other key sizes generated from blasting with plant throughput
including those related to the fine particle size distribution.

1
INTRODUCTION
Salobo is a mine that produces copper concentrate, that is found in the state of Pará in Brazil, 110
kms from Parauapebas and 297 kms from Marabá. Operation on Salobo I, the first phase, began in
Q2 of 2012 with a design capacity of 12 Mtpy, and in Q2 of 2014, the second phase, Salobo II, began
operating, reaching a design capacity of 24 Mtpy.

Figure 1 Salobo Location of the Salobo mine

In general, when the processing plants are in commissioning, all of the efforts of every department
and area from the mine to the plant are focused on achieving and stabilizing the united operations,
postponing the Mine-to-Plant integration initiatives. However, Salobo still working on achieving
and stabilizing their operations, taking advantage of the change in explosives service provider in
February 2015, still not reaching the rated capacity of Salobo II, took on the challenge of working
under the Mine-to-Plant methodology; that is to say, integrating the blasting as the first stage of the
comminution in order to improve the entire chain of mineral processing.

Define blast
fragmentation
target
according to
Process Plant

Design,
Adjust to plant
Modelling and
or blast
Simulation of
fragmentation
Drill and Blast

Drill, Blast and

Plant
Fragmentation
Evaluation
Measurement

Figure 2 Methodology Mine-to-Plant (Alarcón et al., 2015)

2
The objective of this methodology is to constantly look to maximize the treatment in the processing
plant through the improvement in fragmentation in the blasting; in this case the first step to
improve the fragmentation was evident, see Figure 3; the sizes and quantities of the oversized
particles caused a lot of problems in the loading, hauling, crushing and processing plant stages.

Figure 3 Photographic evidence Q4-2014

BACKGROUND

Rock properties

The ore in the Salobo mine is very hard for ball grinding and has a high density (Salobo ore
characterisation database, Cobre Salobo – Resumo Resultados WI. xls), more than ~80% of the
samples taken in 2014 have a Bond Ball Mill work Index above 20 kWh/t, and the density of the
mineral is between 2.7 and 4.0 g/cm3, which is unusual for copper minerals.

40% 35%
Hard Very Hard Extremetely Hard
35% 30%

30%
25%

25%
Frequency (%)
Frequency (%)

20%
20%
15%
15%

10%
10%

5%
5%

0% 0%
<16 16-18 18-20 20-22 22-24 24-26 26-28 28-30 >30 2.6-2.8 2.8-3.0 3.0-3.2 3.2-3.4 3.4-3.6 3.6-3.8 3.8-4.0 4.0-4.2 4.2-4.4
Bond Ball Mill Work Index (kWh/t) Density (g/cm3)

Figure 4 Histograms of samples taken in 2014: Bond Ball Mill Work Index and Density

Furthermore, from the point of view of the blast, the ore is considered to be hard and extra-hard,
see Figure 5; over ~90% of the samples are considered to be hard rocks and more than ~50% of the
samples are considered extra-hard. Therefore, the communition of the ore is a great challenge for
Salobo, and it makes sense to work on the blasting, crushing, HPGRs and grinding in an integrated
manner.

3
 40% 60%
Medium Hard Extra Hard Extremetely Hard
35%
50%

30%

40%
25%
Frequency (%)

Frequency (%)
20% 30%

15%
20%

10%

10%
5%

0% 0%
<60 60-100 100-140 140-180 180-220 220-260 >260 0-20 20-40 40-60 60-80 80-100
UCS (MPa) RQD (%)

Figure 5 Histograms of samples taken in 2014: UCS and RQD

Simplified flowsheet
A simplified flowsheet of the communition circuit is presented in Figure 6. It has two grinding lines
(Salobo I and Salobo II), each one contains two HPGRs and two ball mills.
The primary crushing stage contains two 60” x 89” 600 kW giratory crushers, which, under
standard conditions, one is operating and the other is on stand-by. The product of both feed a
stockpile of coarse particles, which has a total capacity of ~47,000 t.
The secondary crushing circuit is closed inverse, with three 12’ x 24’ double deck vibrating screens,
and each vibrating screen feeds one of the cone crushers with 750 kW motors. Under standard
conditions, two of the three crusher lines are operating and the other is on stand-by. The secondary
crushing product feeds the stockpile of fine particles that has a total capacity of 110,000 t, and which
has eight feeders underneath it, four for each grinding line (Salobo I and Salobo II). The HPGR
circuit are fed by both lines.
The HPGR circuit has four units of HPGR that are 6.56 feet in diameter (2 m) and 4.92 feet long (1.5
m), and each one has two 1.85 MW motors, which are fed by the secondary vibrating screen
product. The product of each one of the HPGRs is vibrated on two double deck vibrating screens;
the product that is retained is recirculated to the HPGR circuit and the small size particles feeds the
ball grinding circuit.
The ball grinding circuit is composed of four 26`x 44` ball mills with 16.5 MW of installed potential
and variable velocity. Each ball mill is operated on a closed circuit with a cluster of 10 26-inch
cyclones with 6.1 inches (155 mm) of apex and 10 inches (254 mm) of vortex.

4
 Figure 6 Simplified flowsheet of the communition circuit of Salobo (I and II)

METHODOLOGY
As was commented in the introduction, Salobo started operating in Q2-2012 (Salobo I) and
expanded in Q2-2014 (Salobo II). Furthermore, in December 2014, it increased the power of the
conveyor belt of the secondary crushing circuit, and in February 2015, it installed a more robust
rock breaker in the primary crusher. Considering that the mine-to-plant methodology began
working in March 2015, the ability to isolate and determine the effect of each change was
challenging.
In the first stage of the implementation of the mine-to-plant methodology it was evident that it was
best to work with the oversized particles and in the P80 of the blasting, over 16 inches (>400 mm).
The problems with loading, hauling, crushing and the processing plant were due to the coarse
fragmentation of the explosion. Because of this, in the first period we chose to work, as the blast
fragmentation objective, on reducing the oversized particle’s size, reducing the P80 and reducing
the P50.
To achieve the objective fragmentation, we worked on implementing the new blasting service
company’s operational discipline in blasts, maintaining the perforation and blasting designs and
increasing the energy in the well by changing the explosive.

5
Operational discipline of new explosive Service company
The first thing that must be done to improve the blast fragmentation is to improve the on site
control of the design implementation (Chiappetta, 2016) of the perforation and blast, height of
perforation (bench plus subdrill), position of wells, amount of explosive loaded per well, height of
stemming, position of the initiation system, among others, in addition to the detonation sequences
for the interaction between shock waves.

Explosive energy was increased

The change in the explosives service company brought with it the change in explosives, which are
more energetic than those that were being used. The pure applied emulsions, with the change in
explosives, the effective energy relative to the ANFO increased from ~103 to ~147 RBS (Relative
Bulk Strength – RBS) and in the blended emulsions, it increased from ~153 to ~167 RBS, see Figure
7.

Applied explosives comparison

180

160

140

120
RBS (%)

100

80

60

40

20

0
100 70
% Emulsion

Base New

Figure 7 Comparison of explosives, according to their effective energy RBS

Since in this case, the explosive load columns per blast well were maintained, the effective energy
relative to the ANFO, RBS, allows us to quickly understand the relationship between energies
applied by the explosives, since it is calculated per unit of volume, that is to say, for each meter of
explosive loaded.

RESULTS
In order to evaluate the impact of the new blast fragmentation, it was necessary to construct a
baseline of each communition stage: primary crushing, secondary crushing and plant that includes
HPGRs and grinding.

Primary crushing
In the same month that the mine-to-plant project started, a new, more robust rock breaker was
installed, for which, to be able to evaluate the impact of the new blast fragmentation on the primary
crusher, four KPIs were defined.

6
To be able to evaluate the impacts of the new blast fragmentation on the primary crusher, the
oversized rocks were quantified, for which two KPIs were defined. The first is related to the
quantity of oversized particles that fed each 100 kt of mineral to the crusher and the second with the
downtimes for each event related to this. With this data, the effect of the new rock breaker can be
isolated, see Figure 8.
Primary Crushers: Downtimes by oversizes, Primary Crushers: Downtimes by oversizes,
event/100 kt h/event
25 0.60

0.50 0.50
20

16.5 0.44
0.40 0.37 0.40
15 0.38
0.34 0.330.34
0.30 0.30
9.7 0.28 0.29 0.28
10 8.4
6.8 6.6 0.20
4.3 4.8 4.9
5 4.3 3.7
2.6 0.10
1.2

0
0.00

Primary Crushers: Downtimes by caging, Primary Crushers: Downtimes by caging,

events/100 kt h/event
20 0.60

18
0.50
16 0.45
0.45
14
0.40
12 11.5 0.36
0.33 0.33 0.34
10 9.2 0.30 0.31 0.30

8 0.24
0.20 0.21
6 5.0 0.17
4.3 0.17
4 3.4
2.9 2.5 0.10
1.5 1.8 1.4
2 1.3 1.1

0 0.00

Figure 8 KPIs for oversized particles and caging

The graph on the upper left of Figure 8 shows that there was an important reduction in oversized
particle events in the feeding of the primary crushers; they were reduced to almost a third.
However, the graph on the upper right shows that the effect of the new rock breaker was not as
significant. In a similar manner, the downtimes for caging were also reduced, and the effect of the
length of these downtimes was greater than with the oversized particles.
Additionally, it was measured in the effect of the improvements in management of the maintenance
that was done on the primary crusher. In Figure 9 the summary of the effects of the rock breaker,
maintenance and blast are shown.

7
 Primary Crushing: Effects of initiatives
(March 2015 to January 2016)
13%
140% 17%

120% 100% 3%

100%

80%

60%

40%

20%

0%
Baseline Rockbreaker Maintenance Blast

Figure 9 Effects of all initiatives on primary crusher in the period

Figure 9 shows that in the period between March 2015 to January 2016 that in the primary crusher,
the new rock breaker achieved an increase of 3% of processed mineral, the management of the
maintenance improved by 17% and finally the blast by 13%.

Secundary crushing
As previously noted, in December 2014, the motors of the conveyor belts of the secondary crusher
were repowered, which would define a new baseline to evaluate the impact of the improvements of
the blast fragmentation, see figure 10.

16%
Secundary Crushing: tph effectives

14%

12%

10%
% Frecuency

8%

6%

4%

2%

0%

tph
Baseline Baseline + Repowering belt Evaluation period

Figure 10 Comparison of frequencies of tph in secondary crushing

Figure 10 shows that an increase in the frequency of the tph exists with higher values, form the
initial baseline, for when the conveyors were repowered, and finally when the blast fragmentation
improved, shown by the blue bars.

8
 Secundary Crushing: Effects of initiatives
(March 2015 to January 2016)
10%
140%
7% 3%
11%
120% 100%

100%

80%

60%

40%

20%

0%
Baseline Repowering 2° Crushing Primary Blast
belt Maintenance Crushing

Figure 11 Effects of all initiatives on the secondary crusher made during the period

Figure 11 shows the effects evaluated on the secondary crusher that impacted the treatment: the
repowering of the conveyor motors (11%), management of the maintenance on the second crusher
(7%), improvement of primary crusher without effect on blast (3%) and finally the improvements of
blast fragmentation, 10%. As expected, the impact of the effect of the blast reduces as it advances in
the communition stages.

HPGR and grinding circuits

The effect of the improvement of the blast fragmentation on the HPGRs and grinding circuit were
evaluated jointly, considering that they work in continuous processes. The improvement of the
blast fragmentation is also reflected in the fine particles, for which an effect is seen on the
circulating load in the HPGRs circuit.

HPGR circuit: % Circulating load Grinding: tph effectives

25%
120%
110% 110% 107%
101%
97% 98% 20%
100% 92%
90% 91% 90%
83%
76%
% Frequency

80% 15%
% Circulating load

60%
10%

40%
5%

20%
0%

0%

Baseline Evaluation period

Figure 12 To the left the % circulating load of HPGRs circuit and effective tph of the grinding circuit

The improved fragmentation showed a reduction in the percentage of the circulating load of the
HPGRs circuit, the effective tph of the grinding considerably increased their frequency in the ranges
of 3,000-3,200 and 3,200-3,400. However, there were other initiatives that also favored the increase

9
in the tph, like: new operating practices, management of the maintenance of the HPGRs and
grinding circuits, operating output, reduction of downtime, among others, see Figure 13.

HPGRs & Grinding: Effects of initiatives

(March 2015 to January 2016)
2% 1% 2%
140% 11% 1%
100% 11% 4%
120%

100%

80%

60%

40%

20%

0%

Figure 13 Distribution of effects of initiatives on the HPGR and grinding circuits

To be able to determine the effects of each initiative, the tph of the grinding, % of circulating load of
the HPGR circuit, data about plant downtime, availability and use of equipment were analyzed.
As expected, in this stage of communition, the effect of the blast is less than in the primary and
secondary crushing; in this evaluation period, an impact of 2% was reached, which translates to the
same percentage of production increase of additional fine coppers, otherwise this initiative would
have been carried out.

CONCLUSION
Although Salobo has been constantly improving its processes and operations, having begun the
testing of the mine-to-plant methodology in this stage has demonstrated that it is beneficial,
positively impacting the primary crushing stage by 13%, the secondary crushing stage by 10%, and
finally the HPGR and grinding circuits by 2%, equivalent to a 2% increase in fine copper in this
evaluation period, March 2015 to January 2016. This confirms that the decision to begin the project,
without waiting for the stabilization of the processes, has been valuable for Salobo, leaving a good
base for large improvements in the blast fragmentation that will take place in the near future, since
the empirical models that relate the fragmentation to the plant treatment show that it can make an
even bigger impact.

ACKNOWLEDGEMENTS
We are grateful to Vale Salobo for their confidence in and integration to this joint project with Orica
and for allowing the publication of the results obtain so far.

10
REFERENCES
Chiappetta, F. (2016). Innovative Blasting Technologies to Improve Fragmentation, Plant
Throughput and Productivity. Orica’s technical seminar, Santiago, Chile.
Alarcón, J. et al. (2015). Mine to Plant methodology in Codelco Andina mine. Procemin 2015,
Santiago, Chile.
Gontijo, G et al. (2015-2016). Technical reports of Mine to Plant Service of Orica in Vale Salobo mine.
Internal reports of service.
Salobo ore characterisation database, Cobre Salobo – Resumo Resultados WI. Xls.

11
CHAPTER 2

Flotation: Fundamentals
and Practical Applications
Adsorption and Flotation Response of the
Rhodococcus Erythropolis Bacteria to Hematite
Carlos Castañeda, Antonio Gutiérrez, Jhonatan Soto and Maurício Torem
Pontifícia Universidade Católica do Rio de Janeiro, Brazil

ABSTRACT
The biotechnology research fields regarding mineral bioflotation have showed that the use of
microorganisms as flotation reagents is a promising alternative because of its green chemistry and
as an alternative solution to treat uneconomic mineral deposits of low grade. Thus, this research
aimed to study the adsorption and flotation response of the R. erythropolis bacteria to hematite
through zeta potential measurements, adsorption experiments, microflotation tests, Fourier
transform infrared spectroscopy (FTIR) and scanning electron microscopy (SEM) analyses. The
isoelectric point (IEP) of hematite showed a significant change after bacterial interaction, decreased
from pH 5.3 to pH 2.1. The bacteria-mineral adsorption experiments were verified by the FTIR
spectrum, where NH, CH2, CH3, C=O, PO2- and COO- groups were identified on the mineral surface
after bacterial interaction. The microflotation tests were carried out in a modified Hallimond tube
and the utmost floatability of hematite reached was of approximately 83.86% in the finest particle
size (53 - 38 μm), using 200 mg/L of the bacteria at pH value of 6.0 and after 10 minutes of flotation.
Finally, the results showed that R. erythropolis bacteria has potential as a biocollector of hematite
and may be a solution against treatment of low-grade mineral deposits.

1
INTRODUCTION
In flotation, the use of microorganisms such as bioreagents rather than potentially toxic chemicals is
very attractive and has given hopes for improving the recovery of minerals, using economical and
environmentally safe processes (Farahat et al., 2009). The separation of minerals by biobeneficiation
is governed by selective adhesion of microbial cells onto mineral surfaces and the changes occurred
on mineral surfaces after the biotreatment (Sharma & Rao, 2002; Farahat et al., 2009). Thus many
works have pointed out that several microorganisms, as the Rhodococcus opacus (Natarajan & Deo,
2001; Mesquita, Linsb & Torem, 2003; Merma et al., 2013), R. erythropolis (Yang et al., 2013) and
Rhodococcus ruber (Lopez et al., 2015) strains, have the ability of turning into hydrophobic, a
hydrophilic mineral surface, due to the different functional groups presented on their cell wall.
Such functional groups present microbial cell surface may be non-polar (hydrocarbon chains) and
polar (carboxyl, phosphates, hydroxyls), which can change interface properties and thereby change
the amphipathic characteristics of a mineral surface (Mesquita, 2000; Dwyer et al, 2012).
The non-pathogenic R. erythropolis strain is widely found in nature in a great variety of sources
including soils, rocks, boreholes, groundwater, marine sediments, animal dung, the guts of insects
and from healthy and diseased animals and plants (Bell et al., 1998; Carla & Manuela, 2005). In
addition it has been widely used for bioremediation of oil-contaminated water and soil (Liu & Liu,
2011; Carla, 2012)
In order to gain a greater understanding in the flotation of the mineral hematite, this paper
develops the basic fundamentals of adsorption of the R. erythropolis bacteria to hematite.

METHODOLOGY

Sample Preparation
The hematite sample, purity of 94%, was provided by local supplier from Araçuaí in Minas Gerais.
The samples were jaw-crushed and dry-screened to -3 mm. These samples were then dry-ground in
a porcelain mortar and wet-screened for obtaining the desired size fractions (Table 1).

Table 1 Experiments and particle sizes

Experiments Size
Zeta potential < 38 μm
Adsorption (53 – 38 μm)
(150 – 106 µm)
Flotation (106 – 75 µm)
(53 – 38 µm)

2
Microorganism, Media and Growth
R. erythropolis was obtained from the CBMAI – UNICAMP. The microorganism was inoculated in a
TSB (Tryptic soy broth) liquid medium (pancreatic digest of casein 17.0 g/L, papaic digest of
soybean 3.0 g/L, dextrose 2.5 g/L, sodium chloride 5.0 g/L and dipotassium phosphate 2.5 g/L), on a
rotary shaker, maintained at 150 rpm and 28 °C, during 70 h.
After that, the bacterial suspension was centrifuged and the obtained biomass was twice washed
with deionized water, then the cells were re-suspended in a 10-3 M NaCl solution. Finally, the
bacterial concentrate was autoclaved in order to avoid further bacterial development. For the
measurement of cell concentration in suspension, the dry weight method was adopted, which was
determined by the difference between the final weight and the initial weight, expressed in g/L.

Zeta Potential Measurements

Zeta potential measurements for the mineral sample and the microorganism were carried out on
the micro-electrophoresis apparatus “Zeta Meter System +4”. The concentration of the biomass
suspension was 50 mg/L and NaCl 10-3 M was used as indifferent electrolyte, while for mineral
samples a concentration of 70 mg/L was used, pH was adjusted using diluted HCl and NaOH
solutions. The evaluation of the zeta potential profiles of cells was carried out before and after
interaction between the bacterial cells and the mineral sample.

Adhesion Experiments
All bioadhesion experiments were carried out in a duplicate run basis. The suspensions were
prepared by mixing 0.25 g of the mineral sample and known quantities of the bacteria cells (50, 100,
150 and 200 mg/L) in 35 mL of NaCl 10-3 M solution, the suspension was agitated on a rotary shaker
at a constant shaking of 150 rpm during 5 minutes to ensure that equilibrium was reached. The
studies were performed as a function of pH (range 2.0 to 9.0). The concentration of the bacteria was
determined using the optical density method (Shimatzu UV-1800 spectrophotometer).
The amount of bacterial adhered was calculated by:
V(Co −Cf )
q= (1)
M

Where: C0 and Cf are the initial and equilibrium concentrations of the bacterial (g/L) respectively, M
is the weight of mineral (g) and V is the solution volume (L).

Flotation Experiments
The microflotation evaluation was carried out in a modified Hallimond tube. An amount of 1.0 g of
mineral was added to 0.16 L total volume suspension of known bacterial concentration. The mineral
was conditioned with the bacterial suspension inside the Hallimond tube under constant stirring for
about 5 minutes, and the pH was adjusted to the desired value. Finally, the mineral flotation tests
were carried out using air at a flow rate of 23 mL/min for 2 min. The settled and floated fractions
were carefully separated, washed, dried and weighed. The floatability was then calculated as the
ratio of floated and non-floated mineral amounts and the total weighed sample.

3
RESULTS AND DISCUSSION

SEM image of the bacteria

The characteristics of the bacteria can be seen in the Fig. 1 (SEM image). A bacillus shape with (0.4-
0.5) μm of diameter and (0.7–2.8) μm of length were observed.

Figure 1 SEM image of the R. erythropolis cells

FT-IR spectroscopy of the bacteria

The FT-IR spectra of the R. erythropolis cells exhibited many bands in the range 400-4000 cm-1, as can
be clearly seen in Fig. 2. Similar spectra were pointed out in different works (Schmitt & Flemming,
1998; Sharma, 2001; Garip, Gozenb & Severcan, 2009; Yang et al., 2013). The band observed in 3292
cm-1 are characteristic of hydroxyl groups and stretching NH; the band 2924 cm -1 are due to C–H
stretching mode of the CH2; the strong band at 1652 cm-1 characterizes the (C=O) stretching
vibration of the amide I, while the presence of amide II can also be found in the band 1538; and the
band at 1452 cm-1 was due to CH2 from lipids. Moreover, the band observed at 1393.5 cm -1
characterizes the stretching of CH2 and CH3; band at 1237 is due to P=O stretching from nucleic acid
and phospholipids; band at 1067 is due to PO-2 symmetric stretching in nucleic acids and the band
at 698 can be attributed to COO-, C=O bending of the carboxyl group.

4
 Figure 2 FT-IR Spectrogram of the R. erythropolis bacteria

Zeta Potential Measurements

Fig. 3 shows the zeta potential profile of the bacterial cells, and hematite before and after interaction
with the cells. It was observed that the surface of the R. erythropolis cells is negatively charged over
a wide range of pH, presenting an isoelectric point (IEP) value at around pH 2.2. Moreover, from
the zeta potential profile of the hematite it was possible to observe an IEP at pH 5.3, this value is in
accordance with those obtained by Sharma et al., 2001 and Yang et al., 2013; however after
interaction of the hematite particles with bacterial cells its IEP shifted to pH 2.1. This modification
could be related to the adhesion of the bacterial cells and/or adsorption of any metabolic product
onto the mineral surface.

Figure 3 Zeta Potential Curves (electrolytic support: NaCl 10-3 M)

Adhesion Experiments
The effect of the pH solution on the adhesion capacity of the bacteria onto the mineral surface can
be observed in the Fig. 4 (a). The highest adhesion was attained at pH 2. It was also observed that

5
higher the initial concentration of the bacteria higher the adhesion capacity. Thus, for the
concentration of 200 mg/L it was achieved an adhesion capacity of 8.7 mg/g. This value is similar to
that found by Mesquita, Linsb & Torem, 2003.

Figure 4 a) Effect of the pH bacteria adhesion and b) Effect of time on the bacteria adhesion (pulp pH 2)

R. erythropolis adsorption onto hematite surface was also studied as a function of adhesion time. For
bacterial concentrations of 150 mg/L and 200 mg/L, the tests revealed a rapid adhesion in the first 5
minutes, as shown in Fig. 4 (b). The maximum bacterial adhesion was around 8.6 and 9.5 mg of
bacterial cells per gram of mineral with bacterial concentrations de 150 mg/L and 200 mg/L,
respectively and after 10 minutes de contact. Therefore, according the results, the optimal
adherence time to be considered for further studies was 5 minutes, not depending on the
concentration of biomass.
The adhesion of the bacterial cells onto the hematite surface can be observed through scanning
electron microscopy. Thus, from the Fig. 5 (a) and (b) it is possible to affirm that the bacterial cells
were adhered onto the hematite surface.

Figure 5 SEM Images (a and b) of the bacterial cells onto the hematite surface. Adhesion time 10 min

Bioflotation Experiments
The floatability of hematite as a function of pH, and particle size, using the R. erythropolis as
biocollector, is presented in the Fig. 6 (a). It is observed that higher floatability values are achieved

6
between pH 5 – 6. This effect was also observed by Merma et al, 2013. Moreover, it was observed a
higher floatability for lower particle size.
These results are not in accordance with the adhesion results, where the higher adhesion was
achieved at pH 2. The IEP of the bacteria was around pH 2, at this pH the cells has no charge and
electrostatic repulsion is inhibited, therefore the cells can form agglomerates and flocs, which will
reduce the bacterial concentration of the suspension at this pH, which affect the adhesion capacity
value.
The effect of the bacterial concentration on the hematite floatability can be observed in Fig. 6 (b). As
expected higher bacterial concentration higher the hematite floatability. However, this effect was
observed until a bacterial concentration of 200 mg/L, beyond this concentration value, the
floatability decreased. This effect was also observed by Yang, Zhang & Jiang, 2007 (hematite
bioflotation), Botero, Torem & Mesquita, 2008 (magnesite bioflotation) and by Merma et al., 2013
(apatite bioflotation). However, the value observed by Yang, Zhang & Jiang, 2007 was significantly
lower at around 16 mg/L, but it is difficult to compare the results, since, different bacteria were
used and experimental conditions differ significantly.

Figure 6 a) Effect of the pH on the mineral floatability (bacterial concentration, 50 mg/L) and b) Floatability of
hematite as a function of bacterial concentration (pulp pH 6)

According to Yang et al., (2013), the hematite recovery increased rapidly as concentration of R.
erythropolis increased from 0 to 75 mg/L, but leveled off at above 75 mg/L and its maximum
hematite recovery was 89.67% at a pulp pH of 6.
According to Lopez et al., (2015), the bioflotation of hematite using R. ruber as bioreagent depends
on the pH value, bacterial concentration and particle size. The higher floatability of hematite was
found at a pH value around 3, with a percentage of 84%, using 0.6 g/L and a particle size range
between -53 µm and +38 µm.
Finally, the floatability of hematite as a function of time can be observed in Fig. 7. The highest
floatability (84%) of hematite was achieved after 10 minutes of flotation; this effect was the same for
the three particle size range used.

7
 Figure 7 Floatability of hematite as a function of time: pH 6, bacterial concentration 200 mg/L

FT-IR spectroscopy of the cells-mineral interaction

After interaction mineral-bacterial cells, FT-IR tests were carried out in order to demonstrate the
adsorption of any functional group of the cells onto the mineral surface. The FT-IR spectra of the
hematite before and after the interaction can be seen in Fig. 8. Before interaction, the hematite
surface contained hydroxyl groups and stretching vibrations of Fe-O. However, after interaction,
the hematite surface contained several functional groups of R. erythropolis with a minimal
dislocation of their respective bands. As for example the hydroxyl groups and stretching NH were
dislocated from 3291 cm-1 to 3425 cm-1, the asymmetric stretching of the CH2 groups to 2923.6 cm-1,
the stretching vibration of the amide I to 1633 cm-1, the stretching NH from amide II to 1540 cm-1,
the stretching of CH2 and CH3 to 1400 cm-1 and the PO-2 symmetric stretching was dislocated to
1090 cm-1. Therefore, the presence of these bands and their dislocation pointed out several
adsorption mechanisms of the cells onto the mineral surface as chemical interaction between the
carboxylic groups and the hematite surface.

8
 Figure 8 FT-IR spectra of hematite (Before and after interaction with the microorganism)

CONCLUSIONS
The results presented in this work showed that the R. erythropolis cells have a strong affiliation to
hematite surface, and its application as a biocollector of hematite is feasible. For a constant ionic
strength (10-3 M NaCl), the zeta potential evaluation of hematite particles before and after
interaction with bacterial cells showed the modification of hematite zeta potential profile. The
maximum bacterial adhesion was around 9.5 mg of bacterial cells per gram of mineral with
bacterial concentration de 200 mg/L and after 10 minutes de contact. The bioflotation of hematite
using R. erythropolis as bioreagent depends on the pH value, bacterial concentration and particle
size. The higher floatability (83.9%) was achieved at pH 6, using 200 mg/L of bacteria, in the particle
size range of (53 - 38 µm) and after 10 minutes of flotation.

ACKNOWLEDGEMENTS
The authors acknowledge PUC-Rio (Pontifical Catholic University of Rio de Janeiro), CNPq
(Conselho Nacional de Desenvolvimento Científico e Tecnológico), FAPERJ (Fundação Carlos
Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro) and ITV-VALE for the financial
support.

REFERENCES
Bell K.S., Philp J.C., Aw K.J.W. & Christofi N. (1998) ‘The genus Rhodococcus’, Journal of Applied
Microbiology, vol. 85, pp. 195–210.
Botero, Ana Elisa C.; Torem, M.L. & Mesquita, Luciana Maria S. (2008) ‘Surface chemistry
fundamentals of biosorption of Rhodococcus opacus and its effect in calcite and magnesite flotation’,
Minerals Engineering, vol. 21, pp. 83–92.

9
Carla C.C.R. and M. Manuela R. (2005) ‘The remarkable Rhodococcus erythropolis’, Applied
Microbiology and Biotechnology, vol. 67, pp. 715-726.
Carla C.C.R. (2012) ‘Adaptation of R. erythropolis cells for growth and bioremediation under
extreme conditions’, Research in Microbiology, vol. 163, no. 2, pp. 125–136.
Dwyer, R.; W.J. Bruckard, W.J.; Rea, S. & Holmes, R.J. (2012) ‘Bioflotation and bioflocculation
review: microorganisms relevant for mineral beneficiation’, Mineral Processing and Extractive
Metallurgy, vol. 121, pp. 65-71.
Farahat Mohsen, Hirajima Tsuyoshi, Sasaki Keiko, Aiba Yuuki & Doi Katsumi (2008) ‘Adsorption of
SIP E. coli onto quartz and its applications in froth flotation’, Minerals Engineering, vol. 21, pp. 389–
395.
Garip, Sebnem; Gozenb, Aysegul Cetin & Severcan, Feride (2009) ‘Use of Fourier transform infrared
spectroscopy for rapid comparative analysis of Bacillus and Micrococcus isolates’, Food Chemistry,
vol. 113, pp. 1301–1307.
Liu, C.W. & Liu, H.S. (2011) ‘R. erythropolis strain NTU-1 efficiently degrades and traps diesel and
crude oil in batch and fed-batch bioreactors’, Process Biochemistry, vol. 46, no. 1, pp. 202–209.
Lopez, Leslie Y., Merma, Antonio G. Torem, Mauricio L. & Pino, Gabriela H. (2015) ‘Fundamental
aspects of hematite flotation using the bacterial strain Rhodococcus ruber as bioreagent’, Minerals
Engineering, vol. 75, pp. 63–69.
Merma, Antonio G., Torem, Maurício L., Morán, José J.V., Monte, Marisa B.M. (2013) ‘On the
fundamental aspects of apatite and quartz flotation using a Gram positive strain as a bioreagent’.
Minerals Engineering, vol. 48, pp. 61–67.
Mesquita, Luciana Maria Souza (2000) ‘Bioflotação de Hematita e Quartzo-Um Estudo de
Seletividade’, Tese de Doutorado em Engenharia de Materiais e de Processos Químicos e Metalúrgicos da
PUC-Rio, Rio de Janeiro.
Mesquita L.M.S., Linsb F.F. & Torem, M.L. (2003) ‘Interaction of a hydrophobic bacterium strain in
a hematite–quartz flotation system’, International Journal of Mineral Processing, vol. 71, 2003, pp. 31–
44.
Natarajan, K.A. & Deo, Namita (2001) ‘Role of bacterial interaction and bioreagents in iron ore
flotation’, International Journal of Mineral Processing, vol. 62, pp. 143–157.
Schmitt Jiirgen & Flemming Hans-Curt (1998) ‘FTIR-spectroscopy in microbial and material
analysis’, International Biodeterioration & Biodegradation, vol. 41, pp. 1–11.
Sharma P.K. & Rao K. Hanumantha (2002) ‘Analysis of different approaches for evaluation of
surface energy of microbial cells by contact angle goniometry’, Advances in Colloid and Interface
Science, vol. 98, pp. 341–463.
Sharma, P.; Hanumantha Rao, K.; Forssberg, K. & Natarajan, K. (2001) ‘Surface chemical
characterisation of Paenibacillus polymyxa before and after adaptation to sulfide minerals’,
International Journal of Minerals Processing, vol. 62, pp. 3-25.
Sharma, Prashant K. (2001) ‘Surfaces Studies Relevant to Microbial Adhesion and Bioflotation of
Sulphide Minerals’, Doctoral Thesis of Technology,Department of Chemical and Metallurgical Engineering
of Lulea University, Lulea.

10
Yang, H.; Tang, Q.; Wang, C. & Zhang, J. (2013) ‘Flocculation and flotation response of R.
erythropolis to pure minerals in hematite ores’, Minerals Engineering, vol. 45, pp. 67–72.
Yang, H., Zhang, Q. & Jiang, Z. (2007) ‘Adsorbability of Mycobacterium phlei on hematite surface’,
Journal of University of Science and Technology Beijing, vol. 14, 2007, pp. 103-106.

11
Bioflotation of Dolomite and Calcite Using
Rhodococcus Opacus Strain
Vanessa Coelho, Ronald Rojas, Antonio Gutiérrez and Mauricio Torem
Pontifícia Universidade Católica do Rio de Janeiro, Brazil

ABSTRACT
Flotation is an important segment in the mineral processing field. The high necessity to achieve
lower operating costs with more stringent specifications of minerals concentrates, as well as stricter
environmental legislation have led to a significant increase in the number of studies in order to find
better processing techniques and more effective flotation reagents.
The aim of this research is to study the bioflotation of calcite and dolomite minerals, evaluating the
bacterial strain Rhodococcus Opacus as a bioreagent. The methodology used to assess the interaction
between the hydrophobic strain and mineral samples, calcite and dolomite, followed the next:
preparation of mineral samples, obtaining the bacterial samples, conducting experiments involving
zeta potential measurements and microflotation tests in Hallimond tube. In addition, there were
carried out chemical and mineralogical analysis, scanning electron microscopy and infrared
spectroscopy in order to assist and to interpret the results obtained in the experiments mentioned
above. The results showed that this environmental friendly bioreagent is very hydrophobic and is
able to produce a stable froth during flotation.

1
INTRODUCTION
Flotation is the most versatile and efficient technique of minerals concentration. Thus, the
bioflotation has ascended due to the use of microorganisms as bioreagent and has been an
environmental friendly alternative to replace conventional methods, which use more aggressive
chemical reagents.
The use of microorganisms as bioreagents in mineral flotation is very attractive for their surface
features, such as negative electrical charge and hydrophobicity. Thus, it is notorious the importance
of research focusing on understanding the surface properties of microorganisms and their
interaction with the mineral surface. The similarity with conventional reagents and their
biodegradability are important factors for an industrial scale application. (Sharma et al., 2001;
Botero et al.,2007; Farahat et al., 2008).
Calcite (CaCO3) and dolomite [CaMg (CO3) 2] are important and are extensively carbonate minerals
extracted in mining activities in Brazil (Sampaio and Almeida, 2005). Because of their solubility and
have physico-chemical characteristics, similar selectivity in the flotation process becomes more
complex. The objective of this research is to study the fundamental aspects of bioflotation of
dolomite and calcite, and to evaluate the bacterial strain Rhodococcus Opacus as bioreagent.
In this work, the effect of the bacterial strain Rhodococcus opacus as biocollector in microflotation of
two carbonate mineral, calcite and dolomite, was investigated through analysis of electrophoretic
behavior and of infrared spectroscopy of the mineral samples. This analysis was carried out before
and after interaction with the bacterial strain. Scanning electron microscopy images were captured
to identify cells of the bacteria adhered on the mineral surface.

MATERIALS AND METHODS

Minerals preparation
Samples of pure dolomite and calcite were provided by a local supplier (Estrada Mining, Belo
Horizonte, Minas Gerais State). The samples were properly ground and classified on Tyler Mesh
sieves. The desired size fractions were obtained and showed in Table 1. Finally, all samples were
stored in a desiccator.

Table 1 Particle size range for each experiment

Experimental Procedures Particle Size

Tests of Zeta Potential < -38 µm
Tests of FTIR -38 µm
-106 +75 µm
Tests of Microflotation -75 +53 µm
-53 +38 µm

Microorganisms, media and growth

R. opacus strain was supplied by the Brazilian Collection of Environmental and Industrial
Microorganisms (CBMAI-UNICAMP). It is a non-pathogenic, Gram-positive and

2
chemoorganotrophic organism, which presents a high hydrophobicity (Mesquita et al., 2003). It was
cultivated in a maintenance solid medium (10.0 g/L glucose, 5.0 g/L peptone, 3.0 g/L malt extract,
3.0 g/L yeast extract and 10 g/L agar). Stocks of the bacteria were prepared and renewed
periodically using this medium in Petri plates and saving them in a refrigerator at 20 °C. Then, the
bacteria was sub-cultured in a liquid medium (10.0 g/L glucose, 5.0 g/L peptone, 3.0 g/L malt extract
and 3.0 g/L yeast extract) at pH 7.2, on a rotary shaker, maintained at 175 rpm and 28 °C, during 72
h. Time needed for the bacterial cells to reach the beginning of the stationary phase of their growth.
After that, the bacterial suspension was centrifuged and the obtained biomass was twice washed
with deionized water.
Finally, the bacterial concentrate was inactivated in the autoclave to avoid further bacterial
development. Concentration of the cellular suspension was quantified by optical density in a
spectrophotometer (UV-1800, Shimatzu UV-spectrophotometer) at a wavelength of 620 nm.

Zeta-potential evaluation
Zeta-potential measurements for R. opacus cell, dolomite and calcite were carried out on a micro-
electrophoresis apparatus Zeta Meter System 4.0+. The concentration of the mineral and biomass
suspension were both of 100 mg/L; NaCl 0.01 M was used as an indifferent electrolyte.
The isoelectric point (IEP) of pure dolomite and calcite minerals and R. opacus suspension were
evaluated. The evaluation of zeta-potential profiles for dolomite and calcite were also carried out in
the absence and presence of R. opacus.

Infrared spectroscopy studies

The infrared absorption spectra of pure samples of dolomite and calcite were recorded before and
after interaction with Rhodococcus opacus strain using a Nicolet Fourier transform infrared
spectrophotometer (FTIR). For spectra of mineral and bacteria interactions, mineral calcite and
dolomite samples were conditioned separately in a cell concentration of 0.3 g / L for 5 minutes at
pH 7 with constant stirring. All samples were separately washed and dried at 60 ° C in an oven for
24 hours. After the samples were properly dried, they were properly mixed with potassium
bromide (KBr) in a 1% (w / w). Measurements of infrared spectroscopy were made with 32 scans.

Scanning Electron Microscopy

Scanning Electron Microscopy (SEM) studies were carried out with the aim to observe the bacterial
attachment onto the mineral particles. The minerals samples of dolomite and calcite after
interaction with the biomass suspension was centrifuged at 4000 rpm for 5 min. After that, water
remains were removed and later were dehydrated in graded series of ethanol or acetone, and air
dried under vacuum. Dry samples were loaded for SEM studies in a FEI Quanta 400.

Microflotation experiments
The microflotation assays were carried out in a modified Hallimond tube. An amount of 1 g of the
minerals samples was added to 160 ml total volume suspension; the pH was adjusted with 0.1 M of
HCl or NaOH solution. A variable concentration of R. opacus solution was added. The bioreagent
was conditioned onto the suspension for 5 min. Five minutes was the flotation time adopted and
the air flow was .15 ml/min.

3
The floated and unfloated minerals were carefully washed, filtered and weighed. The floatability
was then calculated as the ratio of floated and unfloated minerals and the total weighed mineral

RESULTS AND DISCUSSION

Zeta-potential studies
In aqueous media, the cells of the bacteria Rhodococcus opacus behave as charged particles positively
or negatively, and thus a zeta potential profile of bacterial strain curve can be obtained. This may be
due to the presence of polysaccharides and amides phosphate groups that confer to the cell wall
surface of a negative or positive charge depending on the pH value of the solution (Poortiga et al.,
2002).
The Zeta Potential profile of Rhodoccocus opacus strain as a function of pH is illustrated in Figure 1, it
is possible to observe that the isoelectric point (IEP) of bacteria is approximately at pH 3.2.

10

5 (a)
0

-5
Zeta potential (mV)

-10

-15

-20

-25

-30

-35

2 3 4 5 6 7 8 9 10 11 12
pH

5 5
Dolomite
0
(b)
Calcite 0 (c)
Calcite after interaction Dolomite after interaction
-5 -5

-10 -10

-15 -15
Zeta potential (mV)
Zeta potential (mV)

-20 -20

-25 -25

-30 -30

-35 -35

-40 -40

-45 -45

-50 -50
2 3 4 5 6 7 8 9 10 11 12 13 2 3 4 5 6 7 8 9 10 11 12 13
pH pH

Figure 1 Zeta potential of R. opacus (a); calcite (b) and dolomite (c) samples before and after R. opacus
interaction. NaCl 10-3 M as background electrolyte. Bacterial concentration, 0.10 g/L.

4
The zeta potential curve for both pure sample of the minerals dolomite and calcite can also be seen
at Figure 1 and it is observed that the values obtained were negative for all pH ranges. The minerals
dolomite and calcite, both are carbonate minerals and have similar physical and chemical
characteristics, highlighting this fact, it is observed in Figure 1 that the electrokinetic behavior of
calcite resembling dolomite mineral.
The hydrolysis phenomena of slightly soluble minerals can explain the negative values in Zeta
Potential tests. This behavior is possibly assigned by the predominance of anions CO 3 , HCO , OH-
2- 3-

2-
compared to cations H+, Ca2+, Mg2+, probably due to hydrolysis of CO3 groups in the mineral
surface (Charlet et al.,1990; Smani et al., 1975).
Figure. 1 also presents the zeta potential curves of dolomite and calcite after interaction with R.
opacus. In acid pH values, the carbonate minerals solubility is highest. Therefore, the surface of both
minerals is more exposed to the hydrolysis phenomena. Note that there was a noticeable change in
electrokinetic behavior of dolomite and calcite, indicating that there was an interaction between
bacterial strain and minerals surface. However, this behavior does not necessarily indicate an
electrostatic interaction, suggesting a possible chemisorption. For this reason, these results show
that may have been specific adsorption of ionic groups present in the cell wall of the bacteria,
thereby modifying the surface of the mineral.

FTIR studies
The IR spectrum of dolomite, before and after interaction with bacterial cells, is observed in Figure
2. It is clearly observed that interaction with the cells promoted changes in the spectrum. It is
noticeable the presence of the peak 3421 (cm -1) correlated with hydroxyl groups (OH and NH) and
the peak of 2897 (cm-1) corresponding to the CH3 groups. These functional groups are associated to
organic compounds present in the cell wall of Rhodococcus opacus bacteria and their presence can be
related to the adsorption of the bacterial cells on the surface of dolomite.

..... Dolomite
..... Dolomite + R. opacus
Absorbance

Wavenumber (cm-1)

Figure 2. FTIR of dolomite before and after interaction with R. opacus strain

5
The IR spectrum of calcite is presented in Figure 3, this also shows changes that have occurred after
interaction with the bacteria, but the intensity of the absorbance was lower than the spectrum of
dolomite. It can be identified the presence of the band 3296 and 2924 cm-1 corresponding to
hydroxyl groups (OH and NH) and CH3, respectively; which indicates that these compounds have
interacted with the surface of calcite. (Bötcher et al., 1997; Botero, 2007; Veerasingam and
Venkatachalapathy, 2014).

..... Calcite
..... Calcite + R. opacus
Absorbance

Wavenumber (cm-1)

Figure 3 FTIR of calcite before and after interaction with R. opacus strain

Microflotation tests
Figure 4 presents the microflotation of dolomite, the highest floatability value was around 75%,
using 0.4 g/L at pH 7 and a particle size range of (-53+38)µm. This concentration is higher compared
to literature (Mesquita et al., 2003; Merma et al., 2013), and it is due to a higher contact surface area
between the bacteria and the mineral, thereby increasing the consumption of bioreagent.
These results are according to infrared and microscopy analysis, showing a higher affinity of R.
opacus cells for dolomite surface. According to Sheng et al. (2001) the functional groups of the
polymers present in the cell wall of Mycobacterium phlei bacteria would be the responsible for the
adhesion of bacteria on the dolomite surface and this affinity, facilitate the separation of dolomite-
apatite system.

6
 80
75 R. opacus concentration
70 0,1 g/L
0,2 g/L
65
0,3 g/L
60 0,4 g/L
55
50
Floatability (%)
45
40
35
30
25
20
15
10
5
0
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 4 Floatability of dolomite sample as a function of cellular concentration and pH. Particle size: (-53+38)
µm; flotation time: 5 min

The results of calcite microflotation showed a lower floatability (50%) relative to the mineral
dolomite. According Beveridge and Murray (1976) and Fein et al. (1997) certain functional groups
present in the cell wall of bacteria have higher affinity for certain metal ions that allow the acid-base
connection.
60

55 R. opacus concentration
0,1 g/L
50 0,2 g/L
0,3 g/L
45
0,4 g/L
40
Floatability (%)

35

30

25

20

15

10

0
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 5 Floatability of calcite sample as a function of cellular concentration and pH. Particle size: (-53+38) µm;
flotation time: 5 min

7
Sarvamangala and Natarajan (2011) have studied the floatability of calcite with B. subtilis, reaching
maximum values of floatability around 74% at neutral pH with 1 × 109 cells / ml. Padukone and
Natarajan (2011) also studied the flotation of calcite using S. cerevisiae and a lower floatability value
was achieved (around 50%), also at neutral pH, with 4 × 108 cells / ml. In this research, the flotability
of calcite achieved a similar value (50%) than Pakudone and Natarajan (2011), however it was
achieved at a different pH value, which is logical because a different bacteria was used (R. opacus).
Although, calcite and dolomite minerals belong to the same group of carbonate minerals, the
microflotation tests showed differences in their flotabilities. The differences in results between
dolomite and calcite can be attributed to the ionic groups of Mg2+ and Ca2+, respectively. According
to Beveridge and Murray (1976), the affinity is higher for ions Mg2+ than Ca2+ ions.

The froth profile analysis

One of the more important factors in bioflotation is the flotation time; this affects the performance
of the process especially due the peculiarity of the bacteria to form froth, which depends on pH and
bacterial concentration. The microflotation results showed a higher floatability of minerals at pH 7.
So, the ability to form froth as a function of bacterial concentration can be seen in Figure 6. As can
be seen this is not a quantitative result however, it can help to support the ability of the bacteria to
form froth. These images could help to understand the froth behavior of the bacteria in aqueous
solution. Note that there is a greater density of small air bubbles in Figure 6-D and it is observed
that as the bacterial concentration is decreased the density of air bubbles is reduced and,
concomitantly, the air bubble size is increased (Figure 6-A). In conclusion, the R. opacus bacteria can
produce a stable froth with higher bacterial concentration.

A B

C D

Figure 6 The froth formed in microflotation tests with R. Opacus suspension as function of cellular
concentration at pH 7.; (a) 0,1 g/L; (b) 0,2 g/L; (c) 0,3 g/L; (d) 0,4 g/L

8
SEM analyses
The Figure 7 illustrates the bacteria R. opacus adhered on the surface of the mineral samples of
dolomite and calcite. These images indicate the probably of adhesion of R. opacus cells on the
surface of mineral samples. Analyzing the micrographs, it can be seen that the amount of deposited
cells on the dolomite was higher compared to calcite. These results may explain the greater
floatability for dolomite in relation to calcite.

A B

Figure 7 Scanning electron micrographs of (a) dolomite and (b) calcite particles floated, showing the R. opacus
cells adsorbed onto the mineral surfaces

CONCLUSIONS
The authors concluded that the use of the R. opacus bacteria in carbonate minerals flotation was
viable, showing its great potential as dolomite biocollector. In addition, the bacterial cells produce a
stable froth, proving to be very promising for future application in flotation industry. The Zeta
potential measurements indicated that the isoelectric point of Rhodococcus opacus bacteria is around
pH 3.2. Moreover, the Zeta Potential of both mineral presented negative values for the pH range
studied. The results of infrared spectra of the dolomite and calcite after interaction with Rhodococcus
opacus confirmed the presence of functional groups such as OH, CH 2, CH3, C = O, NH, characteristic
of the bacterial cell wall. Rhodococcus opacus cells adhered to the surface of the mineral dolomite and
calcite were observed by Scanning Electron Microscopy (SEM), this adhesion was more pronounced
for dolomite. The highest flotability value for dolomite was found at pH 7 and achieved a value of
about 75%, using 0.4 g/L of bacteria and with a particle size range of +38 -53 micrometers. On the
other hand, the highest flotability value (50%) for calcite was obtained at pH 5. Finally, according to
the results it was possible to identify a window of selectivity for the separation of dolomite and
calcite at pH 7 with 0.1 g/L of bacteria.

ACKNOWLEDGEMENTS

9
The authors acknowledge CNPq (Conselho Nacional de Desenvolvimento Científico e
Tecnológico), ITV-VALE, CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior)
and FAPERJ (Fundação Carlos Chagas Filho de Amparo à Pesquisa do Estado do Rio de Janeiro) for
the financial support.

REFERENCES
Beveridge T. J., Murray R G., Uptake and retention of metals by cell walls of Bacillus subtilis.
Journal of Bacteriology, v 127(3), p. 1502-1518. 1976.
Bötcher M. E., Gehlken P.L., Steele F., Characterization of inorganic and biogenic magnesian calcites
by Fourier Transform infrared spectroscopy. Solid State Ionics, v 101-103, p. 1379-1385,1997.
Botero, Ana Elisa Casas. Bioflotação de magnesita, calcita e barita usando Rhodococcus opacus
como biorreagente. Tese (Doutorado) PUC-Rio, 2007, 130 p. 120
Botero, Ana Elisa Casas; Torem, Mauricio Leonardo; de Mesquita, Luciana Maria Souza.
Fundamental studies of Rhodococcus opacus as a biocollector of calcite and magnesite. Minerals
Engineering, v. 20, p. 1026–1032, 2007
Charlet, L., Wersin, P. and Stumm, W. Geochim. Cosmochim. Acta, 54, p. 2329-36, 1990.
Fein J B., Daughney C, Yee N, Davis T. A., A chemical equilibrium model for metal adsorption onto
bacterial surfaces. Geochimica et Cosmochimica Acta, v 61(16), p. 3319-3328, 1997
Mesquita, L.M.S., Lins, F.A.F., Torem, M.L. Interaction of a hydrophobic bacterium strain in a
hematite–quartz flotation system. International Journal of Mineral Processing, v.71, p.31-44, 2003.
Poortiga A., Boss R., Norde W., Bussher H., Electric double layer interactions in bacterial adhesion
to surfaces. Surface Science Reports, v 47, p. 1-32. 2002.
Sarvamangala, H.; Natarajan, K.A. Microbially induced flotation of alumina, silica/calcite from
haematite. International Journal of Mineral Processing, v.99, ed.1-4, p.70-77, 2011.
Sheng, X.; Arps, P. J.; Smith, R. W. Adhesion of two bacteria onto dolomite and apatite: their effect
on dolomite depression in anionic flotation International Journal of Mineral Processing. V 62, p 159-
172, 2001.
Smani, M., Cases, J. and Blazy, P., Trans. Am. Inst. Min. Eng., 258: p. 176—180, 1975.
Usha Padukone, S.; Natarajan, K.A. Microbially induced separation of quartz from calcite using
Saccharomyces cerevisiae. Colloids and Surfaces B: Biointerfaces, v.88, ed.1, p.45-50 2011.
Veerasingam, S.; Venkatachalapathy, R. Estimation of carbonate concentration and characterization
of marine sediments by Fourier Transform Infrared Spectroscopy. Infrared Physics & Technology,
2014.

10
Investigation of Uncertainties as an Indicator for
Process Evaluation in the Recovery Stages of the
Flotation Circuits
Marcelo Montenegro1, Yousef Ghorbani2, Glenn Araya3 and Manuel Leiva3
1. CICITEM, Chile.
2. University of Exeter, UK
3. Universidad de Antofagasta, Chile

ABSTRACT
This article´s main objective is to determine and analyze the shape of the probability distributions
functions (PDF), its uncertainties, evaluation and selection based in metric indicators comparing
one indicator or all indicators for multiple objectives, for the global recovery and concentrate grade.
The case studies include the uncertainty for the PDF in the recovery stages for PDF normal, uniform
and triangular with at least one different PDF and with standard deviation levels of 1,3,5 y 7. A
total cases were studied which correspond with twelve circuits (eight of them was included the
rougher, cleaner, recleaner and scavenger in the rest four circuits the recleaner stages was not
considered. For each case 2000 mass balance was performed Montecarlo simulation. İn whole
twelve circuit’s approximately 12 million simulations were generated. Among the conclusions
obtained in this work are the following: 1) The shape of PDF in the output variables is dependent
on the Input PDF, the flotation circuit structure and the uncertainty level. 2) The values of global
recovery and concentrate grade do not depend on the the type of PDF and level uncertainty, it just
depends on the type of flotation circuits structure. 3) The circuits that have less uncertainty, they
tend to show more favorable for both the global recovery to the concentrate grade results, which
concludes “The circuits are more stable they are the most efficient.

1
INTRODUCTION
For the decision making in flotation processes, it is necessary to integrate different flotation circuit
structure, selection by indicators and knowledge of the uncertainties of the input and output
variables. The first methodologies could give response to the comparison of alternatives of circuits
with operational and economic indicators (Mendez et al. 2009) and the incorporation of multiple
objectives with algorithm genetic(Guria et al. 2005a; Guria et al.2005b). The work Montenegro et al.
2013, create the methodology considers the construction of a superstructure that includes
alternatives to be analysed the representation of variables by distribution functions, A Monte Carlo
simulation, the incorporation of indicators and multiple objective analysis., Montenegro et al. 2015,
analyses the effect of the uncertainty in the recovery of each concentration stage on the global
recovery and concentrate grade. The uncertainties are represented by normal, uniform and
triangular probability distributions functions (PDFs). The case studies include the uncertainty in the
recovery of each stage separately and simultaneously across all of the stages with different
deviations levels. This work aims to extend the article Montenegro et al, 2015, taking the cases of the
recovery of each when the PDF are different using FDP normal, uniform and triangular with levels
of deviation standard of 1,3,5 y 7, for 6000 cases for all circuits and 12 million simulations.

METHODOLOGY
The methodology used was developed by Montenegro et al. 2013 and is briefly reviewed here. The
circuits considered in this paper are represented by a superstructure to identify all possible circuits
in the analysis. In this work, twelve circuits alternatives are considered (figure 1).These alternatives
are represented by a string where the letters indicate when the cleaner tail, scavenger concentrate,
and recleaner tail are sent. For example the RRR circuit is built by sending the cleaner tail to the
rougher stage, the scavenger concentrate to the rougher stage and the recleaner tail to the rougher
stage (figure 2). For process without a recleaner stage two letters are included, and the SC process is
constructed by sending the cleaner tail to the scavenger stage and the scavenger concentrate to the
cleaner stage. The alternatives without a stage recleaner are RR,RC,SR y SC and the alternatives
with a recleaner stage are RRR, RCR, RRC, RCC, SRR, SCR, SRC and SCC.

2
 Figure 1 Superstructure of the flotation circuit

After selecting the circuit to be tested one or more of the variables that have are selected for
analysis. The uncertainty is represented by a normal, uniform and triangular PDF, including
different cases in the recovery in each stages (at least one different PDF in the cases studied) for the
global recovery and concentrate grade. The standard deviation (DVS) values considered for the
stage recoveries were 1, 3, 5 y 7. A total of 500 combinations and 6000 total cases from twelve
different flotation circuits were analysed in this study eight of them was included the rougher,
scavenger, cleaner and recleaner in the rest four circuits the recleaner stage was not considered. For
each case 2000 mass balances simulations were performed using Monte Carlo simulation.
The uncertainties of the selected variables are defined. For each variable different distributions
within which they are normal, uniform, triangular and represented by an average value for
multivalued standard deviation were considered, standard deviation 1,3,5 y 7 were considered and
used subindex N,U and T are ( where N,U,T are considered normal, uniform and triangular PDFs
respectively). For example, the case 1N1U5T5T, correspons to the case where the DVS of recovery
rougher, recovery cleaner, recleaner are 1,1,5,5 with PDF, normal, uniform, triangular and
triangular respectively (figure 2).

3
 Figure 2 Nomenclature used in this study

In this paper the Global copper recovery and copper concentrate grade were considered. The
indicators the kurtosis and asymmetric factor represent the shape of PDF output, the average is the
value of the indicators and deviation standard its uncertainty, too spider graphics are used multiple
objectives.

RESULTS AND DISCUSSION

The main results of this work are presented and are as follows:

PDF Global Recovery:

The figure 3, describes the results of the shape of the PDF for global copper recovery, using the
kurtosis and assimetric factor for the twelve circuits, which are represented by circles of the
different colours for to show the results obtained in the simulations, on the other hand the red
contour line represent the results for normal PDF, the green contour line green for uniform PDF, the
light blue contour line for triangular PDF, the purple line contour line for hybrid PDF and the
yellow contour line for not know PDF. The main results are: 1) There are no results for triangular
PDF 2) There are few results for uniform PDF 3) The 54% the results are hybrid PDF 4) the 19% the
results are not know PDF 4) The other results are normal PDF, we can finally indicate the shape of
PDF in the output variables is dependent on the Input PDF, the flotation circuit structure and the
uncertainty level.

4
 Figure 3 Skewness and Kurtosis PDFs Global copper recovery, 6000 cases and 12 million simulations

PDF Concentrate Grade:

The figure 4, describes the results of the shape of the PDF for copper concentrate grade, using the
kurtosis and assimetric factor for the twelve circuits, which are represented by circles of the
different colours for to show the results obtained in the simulations, on the other hand the red
contour line represent the results for normal PDF, the green contour line green for uniform PDF, the
light blue contour line for triangular PDF, the purple line contour line for hybrid PDF and the
yellow contour line for not know PDF. The main results are: 1) There are no results for triangular
PDF 2) There are few results for uniform(6%) PDF 3) The 70% the results are hybrid PDF 4) the 1%
the results are not know PDF 4) The 23% the results are normal PDF, we can finally indicate the
shape of PDF in the output variables is dependent on the Input PDF, the flotation circuit structure
and the uncertainty level.

5
 Figure 4 Skewness and Kurtosis PDFs copper concentrate grade, 6000 cases and 12 million simulations

Relationship between DVS and average in the global recovery:

Figure 5, shows the DVS versus the average value of the global recovery PDF. The circuits that sent
the cleaner tail to the rougher delivered the best recoveries, which was expected. An interesting
aspect is that these circuits also yielded smaller SDs, i.e., they were less sensitive to the uncertainties
of the stage recoveries. Comparisons should be made between circuits with the same number of
stages because those with fewer stages also have less uncertainty. 1) The values of global recovery
and concentrate grade do not depend on the the type of PDF and level uncertainty, it just depends
on the type of flotation circuits structure. 3) The circuits that have less uncertainty, they tend to
show more favorable for both the global recovery to the concentrate grade results, which concludes
“The circuits are more stable they are the most efficient.

6
 Figure 5 The average versus the SD of the global recovery

CONCLUSION
The main conclusions obtained in this work are the following:
1) The shape of PDF in the output variables is dependent on the Input PDF, the flotation circuit
structure and the uncertainty level.
2) The values of global recovery and concentrate grade do not depend on the the type of PDF and
level uncertainty, it just depends on the type of flotation circuits structure.
3) The circuits that have less uncertainty, they tend to show more favorable for both the global
recovery to the concentrate grade results, which concludes “The circuits are more stable they are
the most efficient.

ACKNOWLEDGEMENTS
The authors are grateful our respective organizations, to Conicyt and Antofagasta Regional
Government for allowing the publication this paper.

NOMENCLATURE
PDF probability distribution function
DVS standard deviation
N normal probability distribution function
U uniform probability distribution function

7
T triangular probability distribution function

REFERENCES
Guria, C., Verma, M., Gupta, S.K., Mehrotra, S.P., 2005a. Simultaneous optimization of the
performance of flotation circuits and their simplification using the jumping gene adaptations of
genetic algorithm. International Journal of Mineral Processing 77, 165–185.
Guria, C., Verma, M., Mehrotra, S.P., Gupta, S.K., 2005b. Multi-objective optimal synthesis and
design of froth flotation circuits for mineral processing, using the jumping gene adaptation of
genetic algorithm. Industrial & Engineering Chemistry Research 44, 2621–2633.
Méndez, D.A., Gálvez, E.D., Cisternas, L.A., 2009. State of the art in the conceptual design of
flotation circuits. International Journal of Mineral Processing. 90, 1-15.
Montenegro, M.R., Sepúlveda, F.D., Gálvez, E.D., Cisternas, L.A., 2013. Methodology for process
analysis and design with multiple objectives under uncertainty: Application to flotation circuits.
International Journal of Mineral Processing 118, 15–27.
Montenegro, M.R., Gálvez, E.D., Cisternas L.A., 2015. The effects of stage recovery uncertainty in
the performance of concentrations circuits. International Journal of Mineral processing143, 12-17.

8
Typical Reproducibility of Metal Balances in
Flotation Plants
Luc Lachance and Donald Leroux
Triple Point Technology, Canada

ABSTRACT
Each and every flotation plant must produce daily/weekly/monthly production reports in which
metal balances for payable and penalising elements are presented. The content of those reports
origins from raw measurements taken in harsh production environments. The ensuing
measurement/estimation errors always cause irreproducibility. As such, no matter the fact that they
are balanced through statistical data reconciliation, reported numbers are nothing more than just
the most probable values of probability distributions whose parameters must be representatively
assessed. Several authors have documented how such irreproducibility in production reports can
be a source of significant financial losses. Best practices of metal accounting thus prescribe that
production reports be accompanied by measurement quality reports where the reproducibility of
each reported data is compared with a predefined target. However, this is not yet a current practice
for several reasons including a deficiency for some stakeholders to establish a tangible connection
with risk management.
This paper exposes a process for managing the reproducibility of all metal accounting variables. It
starts from the outset at the measurement process and reach the reporting process after having been
through the statistical data reconciliation process. This full process has seldom been published in
the holistic perspective exposed herein which encompasses hardware, software and people
dimensions. The authors have used this estimation process on industrial data coming from several
flotation plants located around the world. This allows the unprecedented reporting of typical levels
of measurement reproducibility that can be used as benchmarks for reproducibility levels that can
be realistically expected within the current practice.

1
INTRODUCTION
Each and every flotation plant must produce daily/weekly/monthly production reports in which
metal balances for payable and penalizing elements are presented. Such production reports are the
most expected output from a metal accounting system (MAS). However, as shown in Figure 1
(Gaylard, Morrison, Randolph, Wortley, & Beck, 2009), a MAS encompasses way more than just a
reporting system. Furthermore, acknowledging that the content of production reports indubitably
origins from raw measurements done is harsh conditions has important implications for how both
the reports and MAS should be monitored.

Figure 1 Three dimensions for the components of a MAS

In view of corporate governance (Gaylard, Randolph, & Wortley, 2014), a MAS should inform
senior management about business risks related to metal accounting. A given MAS must thus not
only deliver production reports in a timely manner, but also report data whose accuracy matches
the business risk level endorsed by management. Therefore, production reports must be
accompanied by measurement quality reports. However, this is not yet a current practice as some
stakeholders still fail to establish a tangible connection with risk management despite the fact that
several authors (Carrasco, Carrasco, & Jara, 2004; Gy, 2004) have documented how measurement
errors can be a source of significant economic losses. Indeed, despite the abundance of available
information about hardware (Gaylard, Randolph, Wortley, & Ralston, 2006; Holmes, 2015; Wortley,
2009), practitioners experience a lack of published management frameworks for measurement
quality. To this extent, the authors have previously published (Lachance, Leroux, & Gariépy, 2015a)
a full lifecycle management process for managing the reproducibility of all metal accounting
variables. Since this paper has used simulated data for both measurements and their related
reproducibility levels, readers have then asked to see typical reproducibility levels for metal
balance variables found in actual plants.
The authors have thus used this aforementioned reproducibility assessment and management
process on industrial data coming from several flotation plants located around the world. This
allows for the unprecedented reporting of typical levels of measurement reproducibility that can be
used as benchmarks for reproducibility levels that can be realistically expected within the current
practice.

2
REPRODUCIBILITY AS A COMPONENT OF MEASUREMENT QUALITY
Table 1 shows three components of measurement quality that should be continuously managed.

Table 1 Three components used to manage measurement quality

Component Description
Availability Availability of a given measurement device at an expected quality level.
Thus the capacity of a MAS to avoid replacement values.
Accuracy Amplitude of the systematic error that occurs when the average value of a
measurement is consistently different from its true value.
Reproducibility Amplitude of the random error that is caused by random variations caused by either
the environment or the measurement process.

This paper focuses on the reproducibility of assays {Cu, Zn, Pb, etc.} as the availability of those
measurements in flotation plants is typically higher than for mass flowrates. This should not be
interpreted, however, as a negation of the importance of either the other two components of Table 1
or of mass flowrate measurements.
Gaussian distributions (see Figure 2) are used herein to characterize the distributions of random
measurement errors. This is acknowledged as appropriate for correct sampling errors as long as the
measured quantity is not found in traces and the sampled mass is not order of magnitudes under
what would theoretically be needed (Gy, 2004). The commonly used quantitative benchmark for
estimation reproducibility is the relative standard deviation being the standard deviation σ divided
by the related reported value. Another commonly used terminology is precision which is
equivalent to twice the standard deviation or about 95 % confidence interval as shown in Figure 2.

Figure 2 Gaussian distribution with its parameters μ (mean) and σ (standard deviation)

Reproducibility (or precision) is not solely restricted to measured variables. Error propagation
theory (Xiao & Vien, 2003) shows that non measured but estimable variables inherit estimation
errors from measured variables through the mass balance equations used in metal balancing. As
such, it should be kept in mind that each and every reported number found in production metal
balance is nothing more than the most probable value μ of a probability distribution (see Figure 2)
whose parameters must be assessed. Thus reported numbers should not be taken at face value.

3
MANAGING THE REPRODUCIBILITY OF ALL METAL ACCOUNTING VARIABLES
The introduction of probability distributions (Figure 2) to characterize random measurement errors
enables the use of statistical data reconciliation. This method should be favored over ad hoc
methods such as neglecting some measured variables. Fully solving a statistical data reconciliation
problem (in a metal accounting context) using best practices is an iterative procedure including
several steps:
1. Initial assessment of the measurement network of the plant
2. Estimability and redundancy analysis
3. Assumption for hardware reproducibility
4. Statistical data reconciliation
5. Sensitivity analysis
6. Reduced residuals analysis
7. Report approval or feedback loop on the assumed hardware reproducibility

Step 1: Initial assessment of the measurement network of the plant

This step includes several sub-tasks such as obtaining valid plant configurations along with their
available measurements. Figure 3 shows an example of a given configuration for a simplified
flotation plant along with a table including potential/typical measurements. This generic case is
used to support following discussions.

Stream Ore Components

Weight % Weight
1 Measured Measured
2 — Measured
3 — Measured
4 — Measured
5 — Measured
6 — Measured
7 — Measured
8 — Measured

Figure 3 Simplified flotation plant configuration along with typical measurements

Another important sub-task is the characterization of processed ore via heterogeneity testing
(Minnitt & Assibey-Bonsu, 2010). Such heterogeneity testing should be performed for each
geological domain and be performed using correct sampling practices (Gy, 2004). The objective here
is to obtain parameters (K and α) allowing the estimation of the fundamental sampling error (FSE)
which is the minimal value for the global estimation error (GEE). Computing FSE also requires
selected mass for each sample to be known as well as the nominal particle size of each stream.

4
Step 2: Estimability and redundancy analysis
This analysis should be performed for each potential combination of plant configuration and
available measurements. It should be assessed if all non measured variables are estimable and
which measured variables are redundant (Lachance & Flament, 2011). Best practices in metal
accounting require significant measured variables to be redundant.

Step 3: Assumption for hardware reproducibility

Most statistical data reconciliation algorithms such as Algosys Bilmat end up solving a constrained
nonlinear weighted least-squares problem. For a given combination of plant configuration and
measurements, the inputs (see Figure 4) are the measured values along with their assumed level of
hardware reproducibility. While measured values are readily available, obtaining representative
numbers for hardware reproducibility is a recurrent issue in many metal accounting cases. As the
purpose of obtaining such numbers in a representative manner is commonly misunderstood, it is
thus wrongly believed that any number will do as long as some misguiding quality criteria
(Lachance et al., 2015a) are satisfied.

Assumed Measured
hardware values
reproducibility

· Eliminates the node imbalances

Data reconciliation while taking into account assumed
hardware reproducibility

· Turns hardware reproducibility

Sensitivity analysis
No into software reproducibility

· Measured values
Reduced residuals analysis · Reconciled values
· Software reproducitibility

· Termination criterion
Is there an adequate match between
Are reduced the assumed hardware reproducibility
residuals ok? and the actual measured values?
Are the reported values acceptable
from a business perspective?
Yes

Approved report · Production numbers

· Representative software reproductibility

Figure 4 Iterative process for report approvals

Reproducibility levels are commonly represented by relative standard deviations that are implicitly
converted into variances. These variances are used as weighing factors inside the constrained
nonlinear weighted least-squares optimization problem. Indeed, the necessity along with the
adequate use of such weighing factors is truly what distinguishes statistical data reconciliation from
other ad hoc reconciliation procedures.

5
It should be clear that the assumed reproducibility levels are related with the aforementioned GEE.
In cases where the GEE is neither directly determined via interleaved sampling nor a variographic
experiment, the combination of minimal numbers (FSE) made available from the above step 1 with
the current iterative procedure (Figure 4) helps with the assessment although uniqueness won’t
generally be possible. See below for reproducibility levels that are typically found in flotation
plants.

Step 4: Statistical data reconciliation

Due to the presence of nonlinear constraints in the metal conservation equations, the statistical data
reconciliation itself (see associated box in Figure 4) is iterative. However, the algorithm is often
commercially packaged in a way that it is merely perceived as an input–output process. It must be
understood that, by construction, any set of input measured values fed to the algorithm will turn
into a balanced output data set. Reaching a reconciled output data set with null node imbalances is
therefore never an issue.

Step 5: Sensitivity analysis

The sensitivity analysis has two important roles in the full statistical data reconciliation procedure
shown in Figure 4. Its first and most tangible role is to compute the reproducibility levels of
reconciled metal balances (referred to as software reproducibility) in response to the hardware
reproducibility associated with the actual measurements. Thus, even though the software
reproducibility levels are seldom reported in practice, sensitivity analysis is there to confirm that
measurement irreproducibility does not vanish in the depth of databases. Its second role is to feed
software reproducibility levels to the reduced residuals analysis (Step 6).

Step 6: Reduced residuals analysis

The use of non redundant metal balance models is not a best practice (AMIRA International, 2007)
as it gives the wrong impression that measurement errors do not matter because they do not create
any non null node imbalance. Furthermore, as shown in Table 2, the absence of such node
imbalances precludes the validation of the assumed reproducibility using mass/metal balancing. It
is thus the crucial role of the reduced residuals analysis (Lachance, Leroux, Gariépy, & Flament,
2014) to ensure that the assumed hardware reproducibility (Figure 4) is representative of the actual
hardware reproducibility. Since the latter is unknown, this can only be achieved indirectly when
the assumed hardware reproducibility levels match the observed/observable node imbalances.

Table 2 Comparative analysis of properties of redundant and non-redundant variables

Variable type Reduced residuals Reproducibility Software reproducibility

assessment

Redundant Defined Enabled Better than assumed hardware

reproducibility

Non-redundant Undefined Disabled Same as assumed hardware

reproducibility

6
Step 7: Report approval or feedback loop on the assumed hardware reproducibility
Once a preliminary metal balance is issued, the appropriate criteria for its approval need to be
determined as neither the avoidance of negative estimates nor the achievement of balanced
numbers constitutes a satisfactory quality warranty by themselves (Lachance, Leroux, & Gariépy,
2015b). Reaching an acceptable (details are beyond the scope of this paper) distribution for reduced
residuals is the primary criterion for assessing the quality of a given metal accounting report.
Therefore, a report should only be issued when the reduced residuals are adequately distributed.
This is achieved either by adjusting the assumed hardware reproducibility (feedback loop in Figure
4) or by removing faulty measurements when necessary. However, it should be kept in mind that
reduced residuals distribution only tells if assumed reproducibility levels match actual
reproducibility levels, not if the achieved software reproducibility is fit for purpose (Lachance et al.,
2015a).

TYPICAL NUMBERS FOR HARDWARE REPRODUCIBILITY

The statistical data reconciliation technique described above has been used by flotation (and other
mineral processing) plant operators around the world for more than three decades. However, very
few of the published results show either the hardware or software reproducibility levels. There are
thus only a few benchmarks against which achieved reproducibility results can be compared.

Methodology
A number of flotation plants (12, located in four different countries) were selected from the base
metals industry. For each plant, a recalibration of the assumed hardware reproducibility numbers
has been performed using the aforementioned procedure. Step 1 has been partially skipped since
no results from heterogeneity testing were available. A few dozens of metal balances were
considered for each participating flotation plant. The greater the number of metal balances available
conducted under similar conditions (ore, ore processing, plant configuration, etc.), the more
representative is the assessment of the actual hardware reproducibility.
Among all the diversified plants and available measurements, four base metals {Cu, Ni, Pb, Zn }
and two critical streams {feed, concentrates} were chosen for illustration. This choice is justified
both by conciseness and the fact that these are variables typically involved in grade/recovery
assessment thus among the most critical variables for metal accounting. Furthermore, it must be
underlined that, among all the possible product streams of a given plant, hardware reproducibility
(relative standard deviation) for each metal assay has been reported for the stream where it is
concentrated.

Results
Figures 5-8 show achieved results for hardware reproducibility using the aforementioned
methodology. Each chart shows results for a given couple {metal assayed, stream}. Figures 5 thus
show the reproducibility of Pb on feed and concentrate streams for five flotation plants. Results are
shown in an arbitrary order inside each chart but a given position for a given metal is associated to
a given plant. The number of reported data for each metal assayed is limited by the availability of
the measurements at each participating site. Red lines are the average values.

7
 Figure 5 Hardware reproducibility levels for Pb on feed (left) and concentrate (right)

Figure 6 Hardware reproducibility levels for Zn on feed (left) and concentrate (right)

Figure 7 Hardware reproducibility levels for Cu on feed (left) and concentrate (right)

Figure 8 Hardware reproducibility levels for Ni on feed (left) and concentrate (right)

8
Analysis
Although we do not claim to have reached a significant sample of the entire practice, Figures 5-8
have the merit of showing what actually exists in some flotation plants. One can readily see that, for
a given metal assayed, there are more variations between sites than between the two streams at a
given site. This is reinforced by the fact that the paired averages for the reproducibility of a given
metal assayed are very close. This could be surprising in view that sampling a concentrate stream is
easier (at least for the FSE) than a feed stream because of higher grade and finer particle size.
However, it is difficult to draw more specific conclusions in the absence of detailed information
about the actual sampling protocols of each flotation plant.
This analysis also lacks the needed information to determine if achieved reproducibility levels
(Figures 5-8) are too high or even too low for a given metal in a given stream at a given site. Only a
business analysis based on value of information coupled with risk tolerance (Lachance et al., 2015a)
can adequately answer this question. Reproducibility targets will thus differ from site to site, not to
mention that they can also change over time. Nevertheless, in view of Figures 5-8, it can be broadly
stated that attention should be paid to reproducibility levels that are above 15 %. In other words,
although we do not state that 15 % (one standard deviation) is the economically optimal target as it
can be found well below, it is less than likely that this optimal target be found above 15 %.

CONCLUSIONS
Subscribing to the seven steps of the full statistical data reconciliation process unraveled in this
paper is a commitment to risk management via the management of measurement reproducibility.
Important indicators such as reduced residuals must be monitored on a continuous basis as they
provide warnings about changes in measurement reproducibility. It should be reminded that
although the metal accounting software warns about these problems, it cannot correct them at their
root causes. Managing hardware and software reproducibility should thus be part of a
measurement quality management framework.
Some key features of this framework involving people, hardware and software are:
· All people involved in the metal accounting process (Figure 1) should be trained on the
good practices of metal accounting
· Sampling and laboratory protocols should be periodically submitted to both internal and
external reviews.
· Heterogeneity testing should be performed for each ore type/geological domain.
· Adequate samplers should be in place
· Each important measured variable should be redundant
· The reduced residuals of each redundant variable on a given admissible time window
should be normally distributed with a zero mean and a unitary variance

This paper has made some steps toward the synthesis of a measurement quality management
framework by gathering disseminated information. However, the convergence is not completed
yet. Directions for future work could involve deeper connections with the theory of sampling and
allow for the extensive use of interleaved sampling and variographic experiments.

9
ACKNOWLEDGEMENTS
Triple Point Technology would like to thanks our participating customers for having made their
data accessible to this study aiming at reporting the state of the art in metal accounting. Specials
thanks should also be delivered to Marie-Phillipe Filteau for having performed the data processing
needed for this paper.

REFERENCES
AMIRA International. (2007) P754 : Metal Accounting - Code of Practice and Guidelines: Release 3,
Melbourne, Australia.
Carrasco, P., Carrasco, P., & Jara, E. (2004) The economic impact of correct sampling and analysis
practices in the copper mining industry, Chemometrics and Intelligent Laboratory Systems, 74(1),
209–213.
Gaylard, P. G., Randolph, N. G., & Wortley, C. M. G. (2014) Metal accounting and corporate
governance, The Journal of the Southern African Institute of Mining and Metallurgy, 114(January),
83–90.
Gaylard, P. G., Morrison, R. D., Randolph, N. G., Wortley, C. M. G., & Beck, R. D. (2009) Extending
the application of the AMIRA P754 code of practice for metal accounting. Base Metals Conference
2009 (pp. 15–38). The Southern African Institute of Mining and Metallurgy.
Gaylard, P. G., Randolph, N. G., Wortley, C. M. G., & Ralston, I. (2006) Designing for metal
accounting, Journal of the South African Institute of Mining and Metallurgy, 106(October), 683–686.
Gy, P. M. (2004) Sampling of discrete materials—a new introduction to the theory of sampling,
Chemometrics and Intelligent Laboratory Systems, 74(1), 7–24.
Holmes, R. J. (2015) Sample station design and operation, WCSB7 (Vol. 2015, pp. 119–128).
Bordeaux, France.
Lachance, L., & Flament, F. (2011) A new ERA tool for state of the art metallurgical balance
calculations, Proceedings of the 43th Annual Meeting of the Canadian Mineral Processors (pp. 331–
346). Ottawa, Canada: Canadian Institute of Mining, Metallurgy and Petroleum.
Lachance, L., Leroux, D., & Gariépy, S. (2015a) Managing business risks also involves managing
measurement quality, Procemin 2015 (pp. 1–12). Santiago, Chile: Gecamin.
Lachance, L., Leroux, D., & Gariépy, S. (2015b) Role of metal accounting in assessing and managing
the business risks involved in production planning. Copper Cobalt Africa - 8th Base Metals
Conference 2015. Victoria Falls Livingstone, Zambia: The Southern African Institute of Mining and
Metallurgy.
Lachance, L., Leroux, D., Gariépy, S., & Flament, F. (2014) Detecting sampling biases in metal
accounting, Sampling 2014 (pp. 109–120). Perth, Western Australia.
Minnitt, R. C. A., & Assibey-Bonsu, W. (2010) A comparison between the duplicate series analysis
method and the heterogeneity test as methods for calculating the sampling constants K and alpha,
The Journal of the Southern African Institute of Mining and Metallurgy, (June), 251–268.
Wortley, C. M. G. (2009) Mass measurement for metal accounting — principles, practice, and
pitfalls. Fourth World Conference on Sampling and Blending (pp. 121–128). Cape Town, South

10
Africa.
Xiao, Z., & Vien, A. (2003). Limitations of variance analysis using propagation of variance. Minerals
Engineering, 16(5), 455–462.

11
Fundamental Aspects of Hematite Fines
Electroflotation Using the Bacterial Strain R. Opacus
as Bioreagent
Ronald Rojas, Mauricio Torem, Vitor Coelho and Antonio Gutiérrez
Pontifícia Universidade Católica do Rio de Janeiro, Brazil

ABSTRACT
Conventional flotation processes are not suitable to treat fine particles due to their low collision
efficiency. This difficulty can be overcome by decreasing the bubble size or by increasing the
apparent particle size. However, it is worth to highlight that the use of small bubbles or micro
bubbles also involves some disadvantages such as their low momentum resulting in a slower
process comparing with the conventional ones. Electroflotation is an attractive alternative for
treating fine particles with small bubbles, smaller than 100 µm, generated by water electrolysis.
Hydrogen and oxygen bubbles can be generated by electrolysis. Various factors influence the
production and size of the bubbles, such as: current density, electrode material, type and electrolyte
concentration and pH. Flotation tests were conducted with a hematite-Rhodococcus opacus system in
a 0.38 L modified Partridge–Smith electroflotation binary cell and the following process variables
were studied: bioreagent concentration, particle size, pH and current density with hydrogen
bubbles that were electrolytically generated. Thise electroflotation study revealed that R. opacus is
present as a bioreagent and reveals its promising application in treatment of hematite fines.

1
INTRODUCTION
The particle size is an important factor in the operating performance of flotation processes.
Consequently flotation strongly depends on the selection of a suitable particle size distribution. The
flotation of fine particles (-100+10 µm) and ultrafines (-10 µm) is critical, resulting in a low recovery,
making the process more dependent on the size of the bubble applied, and the hydrophobicity of
the particles (Gontijo et al., 2007). Low recovery in the flotation of fine (F) and ultrafine (UF)
particles is caused by the low probability of collision and bubble-particle adhesion, the difficulty to
overcome the energy barrier between them, entrainment and entrapment, the high reagents
adsorption, low selectivity of the ultrafine paticles, surface oxidation and mineralogical changes
(Trahar, 1981; Yoon, 2000; Peng, et al., 2015).
In order to solve the problem of low buoyancy of F and UF particles, various flotation technologies
have been developed to increase the bubble-particle collision efficiency. Among the various
technologies highlights the electroflotation. This process is based on the use of the hydrogen and
oxygen bubbles formed by water electrolysis. These two gases can be used separately or together.
Different authors (Sarkar, 2010; Chen, 2004; Baierle, et al., 2015) describe complexity of the
electroflotation process, as this technology involves different factors that control the amount or
production and the size of bubbles, such as the current density, material and type electrode,
electrolyte concentration and pH of the system (Ketkar et al., 1991, Chen, 2004,; Wei, et al., 2011)
The electroflotation has three unique characteristics compared to the other flotation techniques: I)
extremely small bubbles size (5-95 µm), II) different amounts of microbubbles and III) the fact that
electrode selection allows to configure the system for a specific process. These characteristic, when
compared to other conventional flotation processes explain the floatability of fine particles, mainly
by generation of smaller bubbles of 100 µm (Liuyi, et al., 2014). The smaller the size of the bubbles,
the better the recovery of fine particles due to the greater bubble surface area (Yao, et al., 2016). The
collision and attachment phenomenon play a determinant role in the flotation process and are
related with the bubble size. In addition, the electrocoagulation process successfully addresses the
challenges related to the probability of the collision and attachment of the bubble-particle, resulting
in increased recovery of fine particles. This process becomes a biotechnological alternative when
microorganisms are used as bioreagent for separation of the fine particles.
Certain microorganisms are used in the mineral flotation, that due to its amphipathic character are
similar to those of synthetic surfactants used in conventional flotation processes. The composition
of the cell wall of microorganisms enables the adhesion on the mineral surface, thus turning the
particles hydrophobic or hydrophilic. Different authors (Mesquita, et al., 2003; Merma, et al., 2013;
Yang, et al., 2014) attribute the degree of hydrophobicity to the greater proportion of the fatty
groups, proteins and of the basic/acidic character of cell wall. The hydrophilicity is attributed to the
higher polysaccharides proportion of cell wall (Kim, et al., 2015; Lopez, et al., 2015; Sanwani, et al.,
2016). In this sense, this study aims to evaluate the use of bioreagents in the recovery of fine and
ultrafine mineral particles by electroflotation.

MATERIALS AND METHODS

Sample mineral preparation

Pure hematite was used in this study. The sample was provided by a local supplier, Estrada
Mining, Belo Horizonte, Minas Gerais State. First the sample was crushed in a jaw crusher below 3

2
mm. Then, the sample was ground in a rod mill, ending with the classification to obtain the desired
size fractions (-38+20 µm). The ground hematite was first washed with 10-4 M HCl, and then with
double-distilled water until the pH of the mineral suspension was the same as initially. Finally, all
samples were stored in a desiccator.

Biomass obtaining from bacterial strain growth

R. opacus was used as the bacterial strain from which the electroflotation bioreagent was obtained,
and it was supplied by the Culture collection of Fundação Tropical de Pesquisa e Tecnologia André
Tosello – Campinas, São Paulo. R. opacus is a non-pathogenic and gram-positive microorganism,
obtained by isolation from the ground. This bacteria was sub-cultivated in the laboratory using
(Yeast Malt Agar) culture medium YMA from Himedia® , composed of 3 g/L yeast extract, 3 g/L
malt extract and 10 g/L of agar. Stocks of the bacteria were prepared and renewed periodically
using this medium in Petri plates and saving them for 48 h at 20 °C in a bacteriological stove. After
the bacteria cultivation in the liquid medium (pH: 7.2), the flasks were disposed on a rotary shaker,
maintained at 175 rpm and 28 °C, for 72 h. After this time the bacterial suspension was centrifuged,
twice washed in deionized water and finally concentrated in 100 mL of 10-3 M NaCl solution. To
avoid further bacterial development, the concentrate was inactivated in the autoclave. From now
on, the cellular suspension is named bioreagent or biomass, and its concentration was quantified
first by dry weight in an oven at 60 °C for 2 h, and by optical density in UV-1800 Shimadatzu
Spectrophotometer, at a wavelength of 620 nm.

Zeta potential measurements

Zeta potential measurements were made using the Zeta Meter 4.0+ apparatus (Zeta-Meter, Inc.,
Staunton-USA), using the electrophoresis technique. The evaluation of the zeta potential profiles of
pure hematite was carried out before and after interaction with the R. opacus biomass suspension
and the pH was an adjusted using diluted HCl and NaOH solution. The indifferent electrolyte 10-3
M NaCl was used as aqueous medium, for concentrations of 0.1 g/L for hematite and 0.1 g/L for the
bioreagent

Bubble size measurements

The bubble size measurements were conducted at room temperature and atmospheric pressure.
Figure 1 show the setting of the equipment, which consists of a binary electroflotation cell (a
modified Partridge-Smith cell flotation) made of glass in order to measure bubble sizes and
distributions, two platinum meshes (anode and cathode), and a salt bridge to connect the
electrodes. The oxygen bubbles were produced on the anode and the hydrogen bubbles on the
cathode. The floor area of each electrode mesh used was 14.3 cm 2 (mesh diameter: 40 mm; diameter
wire: 0.02 mm). Na2SO4 solution was used as electrolyte. The Anglo Platinum Bubble was used to
measure the bubble size, and images of the hydrogen and oxygen bubbles were taken and
processed by the software ImangeJ 1.45.

Electroflotation experiments
An amount of 1.0 g of mineral was added to 0.380 L total volume suspension of known bacterial
concentration; the pH was automatically adjusted with diluted HCl and NaOH solutions. The
mineral was conditioned with the bacterial suspension inside the modified Partridge–Smith cell

3
under constant stirring for about 10 min, and then the mineral flotation tests were carried out using
bubbles of electrolysis of water during 5 min.

H2 (g) O2 (g)

Figure 1 Modified Partridge–Smith electroflotation binary cell

RESULTS AND DISCUSSION

Bubble size measurements

Figure 2 shows the hydrogen bubble size behavior at different pHs, with a current density of 5.28
mA/cm2 and a concentration of 0.05M sodium sulfate. It can be seen that the hydrogen bubbles
reach a d32 of 76.1 µm in acid medium, which is higher than the diameters achieved in neutral and
alkaline conditions. This can be attributed to the electrostatic interaction between the hydrogen
bubbles and the cathode, as was proposed by Brandon and Kelsall (1985a). At acidic conditions, the
attraction between the positively charged hydrogen bubbles and the negatively charged platinum
electrode increases the restraining forces, and hence the nucleation time, producing larger diameter
bubbles; before buoyancy forces cause bubble departure. On the other hand, at alkaline pHs, the
bubbles and the electrode are negatively charged repelling each other by electrostatic forces
resulting in smaller bubbles. In another study, Brandon and Kelsall, (1985b) showed that such
bubbles can be charged even under surfactant-free conditions and they possess a point of zero
charge between a pH of 2 and 3.
Figure 3 shows the oxygen bubble size behavior at different pHs, with a current density of 5.28
mA/cm2 and a concentration of 0.05M sodium sulfate. Note that the trend of the oxygen bubbles
size was opposed to the behavior of the hydrogen bubbles. In this case, the oxygen bubbles reached
a d32 around 64 µm in acid medium, which was the minimum value. As the pH increased, the
oxygen bubble size reached a d32 around 67 µm, at basic conditions. Since the point of zero charge
of the oxygen bubble is the pH range between 2 and 4, there may be electrostatic repulsion between
the oxygen bubbles and the anode at acid conditions, decreasing the restraining forces and resulting
into lesser diameter bubbles, before buoyancy forces cause bubble departure. On the other hand, at
alkaline conditions, the electrostatic attraction between the bubbles and the anode favors the
formation of greater diameter bubbles.

4
 80
Current density
2
5.28 mA/cm
75
2
6.69 mA/cm
2
8.31 mA/cm
70
d32(µm)

65

60

55

50
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 2 Variation of the hydrogen bubble size as a function of pH for different current densities with Sodium
sulphate (0.05M)

80
Current density
2
5.28 mA/cm
75
2
6.69 mA/cm
2
8.31 mA/cm
70
d32(µm)

65

60

55

50
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 3 Variation of the oxygen bubble size as a function of pH for different current densities with Sodium
sulphate (0.05M)

5
Zeta potential studies
Zeta potentials of R. opacus are shown in Figure 4. The surface of the R. opacus cells is negatively
charged over a wide range of pH, presenting an IEP (isoelectric point) value around 2.8. This result
is in agreement with the values found in the literature (Mesquita et al., 2003; Botero et al., 2008;
Merma, et al., 2013; Kim, et al., 2015), that mentions an IEP around 3. The interaction between R.
opacus and hematite reduced the charge on the hematite surface significantly. After the interaction,
IEP of hematite decreased from pH 5.3 to 3.5. The surface of the hematite interacts strongly with R.
opacus in the pH range between 2 to 7. Based on these results, it is possible to identify a pH range
with the potential to facilitate electrostatic interactions between hematite and the R. opacus cells, i.e.,
pH 2 − 7.

25
20 R. opacus
Hematite
15
Hematite + R. opacus
10
5
0
Zeta potential (mV)

-5
-10
-15
-20
-25
-30
-35
-40
-45
1 2 3 4 5 6 7 8 9 10 11 12
pH

Figure 4 Zeta potentials of hematite before and after interaction with Rhodococcus opacus. Electrolyte
concentration: 10-3 M. NaCl

Electroflotation experiments
The floatability tests were conducted to evaluate the potential of using R. opacus as bioreagent in the
flotation of hematite with hydrogen and oxygen bubbles separately. The hematite floatability as a
function of the bioreagent concentration and pH values with hydrogen bubbles is presented in
Figure 5. It can be observed, that the highest floatability was attained at pH values around 6 and
using 200 mg/L of the bioreagent, beyond this value the floatability decreased. Hematite floatability
increased when the bioreagent dose was rised. The floatability of hematite changed significantly as
the concentration of R. opacus increased. When the concentration R.opacus was < 200 mg/L, the

6
hematite floatability increased with the concentration. However, when the concentration of R.
opacus was > 200 mg/L, the hematite recovery decreased as the concentration increased. At the
optimum concentration of 200 mg/L, the hematite floatability was close to 65%, which showed the
significant affinity of R. opacus for hematite particles. This implies that, R. opacus clearly has
potential practical applications as a bioreagent during hematite flotation.
The floatability of hematite as a function of the bioreagent concentration and pH values with
oxygen bubbles is show in the Figure 6. The floatability of hematite with oxygen bubbles presents
the same behavior as the floatability of hematite with hydrogen bubbles, but the optimum
bioreagent concentration was of 300 mg/L, the hematite recovery was around to 55%.

70
65 Concentration of bioreagent
60 100 mg/L
55 200 mg/L
300 mg/L
50
45
Floatability (%)

40
35
30
25
20
15
10
5
0
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 5 Hematite electroflotation with hydrogen bubbles as a function of R. opacus concentration and pH.
Particle size: (-38+20 µm); flotation time: 5 min

7
 70
65 Concentration of bioreagent
60 100 mg/L
55 200 mg/L
300 mg/L
50
45
Floatability (%)

40
35
30
25
20
15
10
5
0
2 3 4 5 6 7 8 9 10 11 12
pH

Figure 6 Hematite electroflotation with oxygen bubbles as a function of R. opacus concentration and pH.
Particle size: (-38+20 µm); flotation time: 5 min

CONCLUSIONS
The bubble size measurements showed a peculiar behavior of the gas phases; the hydrogen bubble
size decreases with higher pH values while the opposite was observed for oxygen bubble. The zeta
potential evaluation of the mineral particles before and after the R. opacus interaction showed that
the bioreagent modified the zeta potential profiles of hematite. The electroflotation of hematite
particles using R. opacus as bioreagent was also found to be dependent on the pH value and the
bacterial concentration. The higher floatability of both minerals was found at a pH value around 6,
achieving a recovery around 65% and 55% for hydrogen and oxygen bubbles respectively, using
200 mg/L and 300 mg/L of the bioreagent for hydrogen and oxygen respectively. Flotation with
hydrogen and oxygen bubbles has the potential to provide selective separation of fine minerals.
This largely is the result of the increase in the collision efficiency between the fine particles and the
bubbles.

8
ACKNOWLEDGEMENTS
The authors acknowledge CNPq (Conselho Nacional de Desenvolvimento Científico e
Tecnológico), VALE, CAPES (Coordenação de Aperfeiçoamento de Pessoal de Nível Superior),
CETEM (Centro de Tecnologia Mineral) and FAPERJ (Fundação Carlos Chagas Filho de Amparo à
Pesquisa do Estado do Rio de Janeiro) for the financial support.

REFERENCES
Baierle, F. et al., Biomass from microalgae separation by electroflotation with iron and aluminum
spiral electrodes. Chemical Engineering Journal 267 p. 274–281, 2015.

Botero, A.E.C.; Torem, M.L.; de Mesquita, L.M.S. Surface chemistry fundamentals of biosorption of
Rhodococcus opacus and its effect in calcite and magnetite flotation. Minerals Engineering, v. 21, p.
83-92, 2008.

Brandon, N. P. and Kelsall, G. H. (1985a) Growth kinetics of bubbles electrogenerated at

microelectrodes. Journal of Applied Electrochemistry, 15, 475-484.

Brandon, N. P. and Kelsall, G. H. (1985b) Interfacial electrical properties of electrogenerated

bubbles. Journal of Applied Electrochemistry, 15, 485-493.

Chen, G.; Chen, X.; Yue, P.L. Electrocoagulation and electroflotation of restaurant wastewater.
Journal Environment Engineering. 126 v.9, p.858–863, 2000.

Gontijo, C.D.F.; Fornasiero, D.; Ralston, J. The limits of fine and coarse particle flotation. The
Canadian Journal of Chemical Engineering, 85, p.739-747, 2007.

Ketkar, D.R.; Mallikarjunan, R.; Venkatachalam, S. Electroflotation of quartz fines. International

Journal of Mineral Processing, v.31, n.1–2, p.127–138, 1991.

Kim, G.; Park, K.; Choi, J.; Gomez-Flores, A.; Han, Y.; Choi, S. Q.; Kim, H.. Bioflotation of malachite
using different growth phases of Rhodococcus opacus: Effect of bacterial shape on detachment by
shear flow. International Journal of Mineral Processing 143 p. 98–104, 2015.

Lopez, L.Y.; Merma, A.G.; Torem, M.L.; Pino, G.H. Fundamental aspects of hematite flotation using
the bacterial strain Rhodococcus ruber as bioreagent. Minerals Engineering Volume 75, p. 63 – 69,
2015.

Merma, A. G.; Torem, M. L.; Morán, J.V.; Monte, M.B.M. On the fundamental aspects of apatite and
quartz flotation using a Gram positive strain as a bioreagent. Minerals Engineering, v.48, p61-67,
2013

Mesquita, L.M.S., Lins, F.A.F., Torem, M.L. Interaction of a hydrophobic bacterium strain in a
hematite–quartz flotation system. International Journal of Mineral Processing, v.71, p.31-44, 2003.

9
Sarkar, M.S.K.A.; Donne, S.W.; Evans, G.M. (2010). Hydrogen bubble flotation of silica, Advanced
Powder Technology 21, p.412-418.

Sanwani, E.; Chaerun, S.; Mirahati, R.; Wahyuningsih, T. Bioflotation: Bacteria-Mineral Interaction
for Eco-friendly and Sustainable Mineral Processing. Procedia Chemistry 19 p. 666 – 672, 2016.

Trahar, W.J. A rational interpretation of the role of particle size in flotation. International Journal of
Mineral Processing, v.8, p.289-327, 1981.

Yang, H. F.; Li, T.; Chang, Y. H.; Luo, H.; Tang, Q. Y. Possibility of using strain F9 (Serratia
marcescens) as a bio-collector for hematite flotation. International Journal of Minerals, Metallurgy
and Materials, v.21, ed.3, p.210, 2014.

Yao, J.; Yin, W.; Gong, E. Depressing effect of fine hydrophilic particles on magnesite reverse
flotation. International Journal of Mineral Processing 149 p. 84–93, 2016.

Yoon, R.H., 2000. The role of hydrodynamic and surface forces in bubble–particle interaction.
International Journal Miner. Process. 58 (1–4), 129–143.

10
Continuous, Real Time Pulp Chemistry
Measurements and what They Tell Us about
Metallurgical Performance
Christopher Greet and Kyle Selga
Magotteaux, Australia

ABSTRACT
The Pulp Chemistry Monitor (PCM) has been installed and operated in a number of concentrators in
Australasia, and has produced some very interesting results. PCM measures the pH, Eh, dissolved
oxygen, temperature and oxygen demand of process streams within a concentrator continuously and
in real time.
In the example presented in the paper the cleaner feed pulp chemistry is measured, and related back
to changes in the mineralogy as well as variations in metallurgical response. For example, a decrease
in Eh and dissolved oxygen as well as an increase in the oxygen demand. These changes in the pulp
chemistry appear to coincide with increases in the feed grades, probably higher pyrite
grades. Further, these variations in pulp chemistry and feed mineralogy have a negative impact on
the concentrate grade and recovery.
The implication is that if the pulp chemistry is monitored it is possible to predict changes in the
mineralogy of the system, and from this the metallurgical response. Further, the changes in pulp
chemistry can be used in a control system to counter the changes in mineralogy and maintain the
metallurgy.

1
INTRODUCTION
It has long been recognized at a fundamental level that the flotation behaviour of sulphide minerals
is dependent on the pH and Eh of the system (for example: Tolun and Kitchener, 1964; Woods, 1976).
However, measuring pulp chemical parameters within an operating concentrator has had limited
attention. Woodcock and Jones (1970a and b) were among the first to actually collect data from
operating plants. Jones and Woodcock (1984) identified some of the chemical complexities of flotation
pulps, and noted that most concentrators already operate at high levels of metallurgical performance
but there was still considerable scope for improvement. These gains would, in part come from more
effective chemical control. In the same volume, Healy (1984) discussed metal ion hydrolysis and
precipitation, redox chemistry and how the surface chemistry of sulphide minerals can be influence
flotation behaviour. The impact of grinding environment on the pulp chemistry and flotation
response of a complex polymetallic sulphide ore is discussed by Forssberg and Subrahmanyam
(1993). Their work essentially contrasts the pulp chemical and flotation performance of pulps ground
with conventional ferrous grinding media to a fully autogenous system. The data suggested that the
reduction or removal of iron hydrolysis contamination from the system could have a positive
influence on flotation behaviour, and perhaps the first stage of “controlling” the chemistry of the
system should be in the grinding circuit. Grano et al (1995) demonstrated that the collection of pulp,
solution and surface chemical data could identify weaknesses within a concentrator, and from these
data develop, test and implement solutions that improved the process. Greet et al (2006) showed that
the pulp chemical trends observed through primary grinding and rougher flotation circuits of most
sulphide flotation concentrators are approximately the same. There are differences in the magnitude
of various pulp chemical parameters (pH, Eh, dissolved oxygen, temperature and oxygen demand)
as a result of variations in the feed mineralogy and processes employed. Further, it was noted that
the Eh shifted to more reducing values over the ten days measurements were collected, and this
change could be traced to an increase in the iron sulphide content of the feed. Metallurgically, this
resulted in an increase in nickel recovery through the recovery of nickeliferous iron sulphides, and a
decrease in nickel concentrate grade.
The literature tells us that the sulphide mineral flotation is chemically driven, and that an
understanding of how the chemistry is changing in an operating plant may actually lead to firstly,
and understanding of why things happen and secondly the ability to react to these changes to
improve process efficiency.

INSTRUMENTATION
The Pulp Chemistry Monitor (PCM®), patent pending, samples continuously the process stream(s) of
interest within the operating concentrator. A subsample of this slurry is directed to the sample
chamber, which is filled, the measurements (pH, Eh, dissolved oxygen, temperature, conductivity
and oxygen demand) taken before the sample chamber is emptied and the probes cleaned with water
and made ready to receive the next sample. The excess slurry and wash water are returned to the
process. Figure 1 shows the general layout of PCM®.

2
 (b)
(a)

Figure 1 PCM® configuration: (a) Sampling module; and (b) Instrumentation cabinet

The raw data collected is sent to PCM®’s on-board PLC where it is processed, stored and dispatched
to the plant DCS for display in the control room. Other critical bits of data may also be extracted from
the array of information collected by the multitude of sensors within the concentrator, and these are
assimilated to produce meaningful trends to assist the plant operator.

CIRCUIT CONFIGURATION
The PCM® is installed, and has been operating since January 2015, within a cleaner circuit of a
copper/gold operation to collect the pulp chemical data from the Jameson Cell feed and tailing
(Figure 2). PCM® measured the pH, Eh, dissolved oxygen, temperature and oxygen demand of both
process streams. In addition to these data, the cleaner feed throughput, Jameson Cell feed,
concentrate, Cleaner 1 feed, concentrate and tailing on-stream analysis (OSA) data were extracted
from the DCS.

RESULTS
A typical day’s worth of data was analysed and presented below.

Pulp chemistry comparisons

Initially, the trends for pH, Eh, dissolved oxygen and oxygen demand were examined (Figure 3). The
pulp temperatures were around 38°C for both the Jameson Cell feed and tailing as did not vary
significantly. Conductivity measurements have been taken previously and found to be of limited use
mainly due to the high ionic strength of the process water making it difficult to measure meaningful
changes.

3
 Pulp Chemistry monitor PLC DCS and Control Room

OSA data
Jameson
Cell
Rougher con

Cleaner tailing
Cleaner 1

Cleaner 2

Cleaner 3

Final copper concentrate

Figure 2 Schematic of PCM® installation

The data in Figure 3(a) shows that the Jameson Cell operates at a very high pH ranging from 11.8 to
12.2. The elevated pH is consistent with the sites pyrite depression strategy. Broadly, the pH of the
system decreased from around 12.2 at the beginning of the day to around 11.8 at its end. The reasons
for this are not clear. In general, the pH of the Jameson Cell tailing was slightly lower than that
recorded in the feed, and it is presumed that this decrease resulted from the oxidation of pyrite.
The Eh of the system (Figure 3(b)) shows reasonable variation, ranging from nominally -60 mV up to
values around 0 mV. The Jameson Cell feed tended to have slightly more reducing Eh values
compared with the tailing, and this can be attributed to pulp aeration that occurs in the flotation
process.
Figure 3(c) contains the dissolved oxygen data. The dissolved oxygen content of the pulp varies from
around 2 to 5 ppm, with the tailing data closely matching that of the Jameson Cell feed. It was
interesting to note that the Jameson Cell tailing tended to contain marginally lower concentrations of
dissolved oxygen compared to the feed. It is presumed that this is related to the consumption of
oxygen from the pulp as the pyrite is oxidised during flotation.
The oxygen demand data for the Jameson Cell feed and tailing appear in Figure 3(d). It is thought
that this number is a better indicator of what is occurring as it is a measure of the rate at which the
pulp takes up oxygen. So, it is a kinetic measure of the reactivity of the system. The oxygen demand
varies considerably from 0.15 to 0.55 min-1, with the Jameson Cell tailing a slightly higher demand
than the feed. Again, this is probably a reflection of the reactivity of the pyrite in the system. These
data also show that the oxygen demand is not static, but varies with time presumable due to changes
in the feed to the Jameson Cell.

4
 13.0 6

5
12.5

Dissolved oxygen, ppm

4
12.0
pH

11.5
2

11.0
1

10.5 0
0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Time, min Time, min
(a) Jameson Cell feed Jameson Cell tailing (c) Jameson Cell feed Jameson Cell tailing

80 0.7

60 0.6

40

Oxygen demand, min-1

0.5
20
0.4
Eh, mV

0
0.3
-20

0.2
-40

-60 0.1

-80 0
0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Time, min Time, min
(b) Jameson Cell feed Jameson Cell tailing (d) Jameson Cell feed Jameson Cell tailing

Figure 3 PCM® pulp chemical data collected around the Jameson Cell: (a) pH, (b) Eh, (c) dissolved oxygen
and (d) oxygen demand

Influences on oxygen demand

The two parameters that appear to have an influence on the oxygen demand are the throughput
(Figures 4 (a) and (b)), and feed composition (Figures 4 (c) and (d)). It should be noted that the OSA
provided both copper and iron feed grade information. The copper feed grade was chosen because it
is known that the vast majority of the copper in the feed is associated with copper sulphide minerals
(chalcopyrite, chalcocite and bornite, predominantly). The iron assay on the other hand can be
attributed to not only sulphide but also non-sulphide gangue species. For example, the iron is
associated with pyrite, chalcopyrite, bornite, magnetite and a host of other minerals (e.g. silicates,
carbonates). Therefore, it was reasoned that the copper OSA grades would be the better
approximation for the sulphide content of the feed.
The time series plot given in Figure 4(a) indicates that the Jameson Cell feed throughput increased
by some 50 tonnes per hour over the day. This coincided with an increase in the Jameson Cell feed
oxygen demand. When the oxygen demand was plotted against throughput (Figure 4(b)) it is
apparent that there does appear to be weak relationship with an R2 value of 0.27.
Similarly, the copper feed grade and Jameson Cell feed oxygen demand time series plot were
constructed in Figure 4(c). These data show that the copper feed grade increased from around 2.5
percent at the start of the period, increasing to 5.0 percent just over midway before decreasing to
nominally 3 percent. The oxygen demand appeared to rise and fall as the copper feed grade varied.

5
The time series plot shows very good agreement between the two pieces of data, with increases in
oxygen demand coinciding with increases in copper feed grade and conversely decreases in oxygen
demand when the copper feed grade decreased. A plot of the oxygen demand against copper feed
grade (Figure 4(d)) suggests that there is potentially a reasonable relationship between the two
parameters. The R2 for this relationship was 0.22.
Both relationships deserve further investigation on a larger data set.
0.7 350 0.7 6.0

0.6 300 0.6 5.0

Jameson Cell feed tonnes, tph

Oxygen demand, min-1

Oxygen demand, min-1

Cu feed grade, %
0.5 250 0.5 4.0

0.4 200 0.4 3.0

0.3 150 0.3 2.0

0.2 100 0.2 1.0

0.1 50 0.1 0.0

0 200 400 600 800 1000 1200 1400 1600 0 200 400 600 800 1000 1200 1400 1600
Time, min Time, min
(a) (c)
Jameson Cell feed Jameson Cell feed tonnes Jameson Cell feed OD Jameson Cell feed Cu grade

350.00 6.0

300.00
5.0
Jameson Cell feed tonnes, tph

Cu feed grade, %

250.00

y = 107.7x + 223.68 4.0

200.00 R² = 0.2737
y = 3.6352x + 2.0169
R² = 0.2236
150.00 3.0

100.00
2.0

50.00

1.0
0.00 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
(b) (d) Oxygen demand, min-1
Oxygen demand, min-1

Figure 4 Oxygen demand relationships: (a) time series plot with Jameson Cell throughput, (b) oxygen
demand versus throughput, (c) time series plot with copper feed grade and (d) oxygen demand versus copper
feed grade

Impact on metallurgical response

So, the oxygen demand moves around, does the change in reactivity have an impact on metallurgy?
On the limited set of data discussed in this paper there does appear to be some trends that could be
meaningful. A time series plot of the Jameson Cell feed oxygen demand and copper recovery is
providing in Figure 5. These data show that for the first half of the period the oxygen demand was
0.31 min-1 (or lower), and the copper recovery to the Jameson Cell concentrate averaged 46 percent.
During the second half of the period the Jameson Cell feed oxygen demand increased to 0.41 min-1
(or higher), and the copper recovery decreased to an averaged 35 percent. This is rather surprising
given that the Jameson Cell copper feed grade increased from nominally 3 percent to 4 percent for
the corresponding timeframe, and it is broadly accepted that recovery usually increases with
increasing feed grade.

6
 0.7 70.0

0.6 60.0

Oxygen demand, min-1

0.5 50.0

Cu recovery, %
0.4 40.0

0.3 30.0

0.2 20.0

0.1 10.0

0 0.0
0 200 400 600 800 1000 1200 1400 1600
Time, min
Jameson Cell feed Jameson Cell Cu recovery

Figure 5 The Jameson Cell feed oxygen demand and copper recovery time series plot

The oxygen demand versus Jameson Cell copper recovery data are plotted in Figure 6, and although
there is considerable scatter in the data the regression line does indicate a drop in copper recovery as
the oxygen demand increased.
80.0

70.0

60.0
Cu recovery, %

50.0

40.0

30.0

20.0 y = -40.134x + 55.112

R² = 0.08
10.0

0.0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7
Oxygen demand, min-1

Figure 6 The Jameson Cell feed oxygen demand versus copper recovery data

To confirm this observation the data for Cleaner 1 was examined. Figure 7 shows the time series plot
of the Jameson Cell tailing oxygen demand and the copper recovery to Cleaner 1 concentrate, with
respect to Cleaner 1 feed. Again, at times when the oxygen demand is high the copper recoveries (as

7
described by the black arrows). When the oxygen demand versus copper recovery was plotted a
similar relationship to that observed in Figure 6 was revealed.
0.6 100.0

Oxygen demand, min-1 0.5 97.0

0.4 94.0

Cu recovery, %
0.3 91.0

0.2 88.0

0.1 85.0

0 82.0
0 200 400 600 800 1000 1200 1400 1600
Time, min
Jameson Cell tailing OD Cleaner 1 Cu recovery

Figure 7 The Jameson Cell tailing oxygen demand and Cleaner 1 copper recovery time series plot

Apart from potentially having an impact on the copper recovery variations in oxygen demand may
also have an effect on the copper concentrate grade. Figure 8 displays the time series plot for the
Jameson Cell feed oxygen demand and the copper concentrate grade. These data suggest (as depicted
by the arrows) that as the oxygen demand increases there is a subtle decrease in the copper
concentrate grade, and a decrease in reactivity yields a higher copper grade.
0.7 32.0

0.6 30.7
Oxygen demand, min-1

Concentrate grade, %

0.5 29.3

0.4 28.0

0.3 26.7

0.2 25.3

0.1 24.0
0 200 400 600 800 1000 1200 1400 1600
Time, min
Jameson Cell feed Jameson Cell Cu concentrate grade

Figure 8 The Jameson Cell feed oxygen demand and copper concentrate grade time series plot

8
As the OSA data only contains copper and iron assays, it is difficult to determine what diluent species
is causing the shift in copper concentrate grade as the oxygen demand varies. Again, further work
on a larger data set will provide greater confidence in these observations.

CONCLUSION
The PCM® was installed and has been operating since January 2015 in the cleaner circuit of a
copper/gold operation collecting pulp chemical data from the Jameson Cell feed and tailing. To
demonstrate the data captured a day’s worth of information has been analysed. The following
conclusions can be drawn from the analysis:
 The trends observed for the various pulp chemical parameters (pH, Eh, dissolved oxygen,
temperature and oxygen demand) measured in the Jameson Cell feed were similar to those
noted in the tailing.
 The oxygen demand is thought to be a kinetic measure of the reactivity of the system, and
thereby an excellent parameter to use to describe the chemistry of the system.
 The two metrics that appear to influence the oxygen demand in this system were the
throughput and the copper feed grade.
 Changes in the oxygen demand appeared to effect the copper recovery and copper
concentrate grade. That is, as the oxygen demand increased both the copper recovery and
copper concentrate grade decreased. This was observed despite the copper feed grade
increasing.
Further analysis on a larger data set is required to validate these observations.

REFERENCES
Forssberg E and Subrahmanyam T V (1993), Grinding, pulp chemistry and particle floatability, in the
Proceedings of the XVIII International Minerals Processing Congress, pp 1-6 (The Australasian
Institute of Mining and Metallurgy: Melbourne).
Grano S R, Wong P L M, Skinner W, Johnson N W and Ralston J (1995), Detection and control of
calcium sulphate precipitation in the leas circuit of the Hilton Concentrator of Mount Isa Mines
Limited, Australia, in the Proceedings of the XIX International Minerals Processing Congress,
Volume 3: Flotation Operating Practices and Fundamentals, Chapter 29, pp 171-179.
Greet C J, Kinal J and Mitchell I (2006), Is measuring pH enough? In the Proceedings of the
Metallurgical Plant Design and Operating Strategies Conference, pp x-y (The Australasian Institute
of Mining and Metallurgy: Melbourne).
Healy T W (1984), Pulp Chemistry, surface chemistry and flotation, Principles of Mineral Flotation
(The Wark Symposium), pp 43-56 (The Australasian Institute of Mining and Metallurgy: Melbourne).
Jones M H and Woodcock J T (1984), Applications of pulp chemcistry to regulation of chemical
environments in sulphide mineral flotation, Principles of Mineral Flotation (The Wark Symposium),
pp147-183 (The Australasian Institute of Mining and Metallurgy: Melbourne).
Tolun R and Kitchener J A (1964), Electrochemical study of galena-xanthate-oxygen flotation systems,
in the Transactions of the Institute of Mining and Metallurgy, 73, pp 313-322.

9
Woodcock J T and Jones M H (1970a), Chemical environment in Australian lead-zinc flotation plant
pulps: I. pH, redox potential and oxygen concentration, in the Proceedings of the Australasian
Institute of Mining and Metallurgy, 235, pp 45-60 (The Australasian Institute of Mining and
Metallurgy: Melbourne).
Woodcock J T and Jones M H (1970b), Chemical environment in Australian lead-zinc flotation plant
pulps: II. Collector residuals, metal ions in solution and other parameters, in the Proceedings of the
Australasina Institute of Mining and Metallurgy, 235, pp 61-76 (The Australasian Institute of Mining
and Metallurgy: Melbourne).
Woods R (1976), Electrochemistry of sulphide minerals, in Flotation, A M Gaudin Memorial Volume
(Edited by M C Fuerstenau), Volume 1, pp 298-333 (The American Institute of Mining and
Metallurgical Engineers: New York).

10
Improving Image Analysis of Online Bubble Size
Measurements with Enhanced Algorithms
Claudio Acuña1, Luis Vinnett1 and Seng How Kuan2
1. Automation and Supervision Centre for Mining Industry, CASIM. Universidad Técnica
Federico Santa Maria, Chile.
2. Department of Mechanical and Material Engineering, Universiti Tunku Abdul Rahman,
Malasia

ABSTRACT
In the flotation process, one of the variables directly related with flotation kinetics and fundamental
models is the bubble size in the collection zone, which is normally represented bubble mean Sauter
diameter. However, sampling and measuring bubble size distributions in the collection zone is
complex due to the presence of dark particles and bubble clusters, making it difficult the bubble
detection. Therefore, connecting flotation kinetics with bubble size is limited to the measurement
technique.
The most widely used technique for bubble size measurement is the McGill University bubble viewer.
It has been observed that irregular bubbles and small bubbles (below 0.5 mm) are not completely
detected, the former due to the use of the shape factor criteria to discriminate bubbles and the latter
due to resolution constraints. The exclusion of these bubbles introduces a bias into BSD
characterizations. To tackle these limitations, enhanced image analysis algorithm was developed.
The algorithms use a selection criteria based on the ratio of object convex area and object area, to
include small bubbles and irregularly sized bubbles. In order to solve the problem of uneven
illumination or presence of dark particles, a variable threshold (based on intensity histogram) was
applied multiple times to the same image, subtracting the detected bubbles. The proposed algorithms
allowed for increase the reliability of the image analysis technique by including small bubbles and
irregular bubbles usually omitted in a classical bubble size analysis. As a result, 80% of the sample
bubbles are neglected and therefore significant bias is produced in bubble size measurement
technique if only the shape factor criterion is used. The proposed algorithm, based on convex area
and multiple thresholding, appears to significantly improve the reliability of the McGill Bubble
Viewer technique, and therefore predict kinetics with better accuracy.

1
INTRODUCTION
Kinetics play a key role in the flotation process, particularly in the collection zone (Yianatos et al.,
2005, Yianatos et al., 2008). A measurement related directly with flotation kinetics, is the bubble size
distribution in the collection zone, below the froth interface (Finch and Dobby, 1990). Nevertheless,
accurate and reliable measurements are complex to perform, validate and report. These limitations
are due to the presence of dark particles swarming among the bubble, making it complicated to
isolate and size individual bubbles for measurement.
A solution to this problem is a technique developed by McGill University (Chen et al., 2001). A
portable manual device has been used for cell characterization (Hernandez-Aguilar et al., 2002), and
to decouple the effect of chemistry and hydrodynamics in industrial flotation machines (Gomez and
Finch, 2002). The device (Figure 1) consists of a sampling tube, a viewing chamber, background
lighting system and a digital video camera.
Viewing
chamber
Camera
Diffuser

Inclined Background
window lighting

Froth with
collected
Sampling particles
Flotation tube
machine

Collection zone

Figure 1 McGill University bubble viewer (Hernandez-Aguilar et al., 2002)

The device operates manually and the computation of bubble size distribution is reported off line,
normally weeks after the site-testing period. This technology has been intensely used for more than
15 years without significant improvements to the well-known limitations. Some limitations are: (a)
manual operation that requires washing of the chamber to allow for subsequent measurements, (b)
the exclusion of small bubbles below 0.5 mm and irregularly shaped bubbles. Small bubbles are hard
to detect due to pixel per distance settings that diminish when a large area is taken. Irregularly shaped
bubbles are usually discriminated when the shape factor (SF) criteria is used for bubble selection. The
shape factor corresponds to the level of circularity of an object and it is calculated as indicated in
Equation 1. A shape factor of one indicates a circle, and as the value approaches zero, it indicates an
increasingly elongated polygon (Bailey, 2004).
4𝜋𝐴
𝑆𝐹 =
𝑃2
(1)

2
where, A and P are the area and the perimeter of the segment (bubble), respectively.
The reciprocal of the shape factor is named compactness (Grau and Heiskanen, 2003) or roundness
(Empix inc., 2002). A shape factor for circles is used to identify single bubbles from bubble clusters,
based on that bubbles under flotation conditions appear to be mostly spherical (Hernandez-Aguilar,
2004; Grau, 2006; Bailey, 2004; Bailey et al. 2005). However, evidence from over 600 measurements in
industrial cells (over 4 million bubbles analyzed), using the McGill bubble sizing technique and the
enhanced image processing algorithm described above (convex area criteria), indicates that the
average aspect ratio is 0.79, and that bubbles appear to be generally elliptical, as in Figure 2. These
results show that a shape factor criterion fails for elliptical bubbles. From Figure 2, 80% of the isolated
bubble are not properly selected if only shape factor is used.

50
Frequency number (%)

40

30

20

10

0
0,5 0,6 0,7 0,8 0,9 1,0
Aspect ratio

Figure 2 Aspect ratio frequency distribution (613 industrial cell measurements)

As a consequence, a shape factor defined for circles does not appear to be the appropriate selection
criterion. This may affect the accuracy of bubble size distribution measurement and, importantly to
correlate bubble size with flotation kinetics.
Therefore, to overcome these shortcomings, enhancements were made to the bubble viewer to allow
for online bubble measurements and a more robust algorithm was developed to take into account
both small and irregular bubbles.

METHODOLOGY

McGill University Bubble Sizing Technique

Normally for the McGill bubble sizing technique, the shape factor value is set to 0.90 to classify
isolated bubbles. However, the shape factor for some bubble clusters can be accepted and some single
bubbles rejected, as is illustrated in Figure 3.

3
 overlapping single cluster cluster
4 selected
circles bubble bubble bubble
1 single bubble

SF = 0.91 SF = 0.91 SF = 0.91 SF = 0.91

touching single single single

circles bubble bubble bubble
4 rejected
3 single bubble

SF = 0.67 SF = 0.79 SF = 0.80 SF = 0.81

Figure 3 Low selectivity using shape factor criterion

Figure 4 shows shape factor values for two sets of data, in which bubbles were selected manually to
ensure single bubbles. In presence of high concentration of frother, bubbles tend to be spherical and
then the acceptance is high, while without frother the criterion is not selective.

Industrial testtest
Industrial (frother)
(frother) Laboratory test
Industrial
Laboratory (no
test frother)
(frother)
test (no frother) Laboratory test (no frother)
1,0
1.0 1.0 1.0 1.0
1,0
1.0 1.0

, 0.8
0,8
0.8 0.8 0.8
0,8
0.8 0.8
Shape Factor

Shape Factor
Shape Factor

, 0.6
0,6
0.6 ,
0.6 0.6
0,6
0.6 0.6

, 0.4
0,4
0.4 0.4 0.4
0,4
0.4 0.4
0 01 12 23 34 45 56 0 001 112 223 334 445 556 66 0 1 2 3 4 5 6
6
Bubble sizesize
(mm) Bubble size
Bubble (mm)
Bubblesize
size(mm)
(mm) Bubble size (mm)
Bubble (mm)

Figure 4 Shape factor values and the effect of frother

To avoid rejecting isolated bubbles and accepting bubble clusters, other criteria have been proposed.

4
Convex Area Criteria
Bailey et al. (2005) evaluated a selection criterion based on number of holes (bright spot indicating
bubble centre), hole size and shape factor. This enhanced the accuracy of isolating single bubbles, but
shape factor remained inappropriate to identify bubble from clusters.
Mena et al. (2005) proposed using the ratio of the area of the bubble to the convex area (concept
illustrated in Figure 5. According to Mena et al., a factor 0.99 appeared to be a "good criterion" for
ellipsoidal bubbles.

area bubble convex area

Figure 5 Actual bubble area and its corresponding convex area (95%)

However, if the same isolated bubbles used in Figure 4 are taken and the bubble area and convex
ratio are plotted (Figure 6), clearly this criterion improved significantly the identification of isolated
bubbles, regardless the shape factor. In this case, spherical and irregular isolated bubbles both are
properly identified.

Industrial testtest
Industrial (frother)
(frother) Laboratory testtest
Laboratory (no(no
frother)
frother)
1,0
1,0
1.01.0 1.01.0
area object / convex area
Factorarea

0,8
0,8
0.80.8 0.80.8
Shape Factor
/ convex
Shape
area object

0,6
0.60.6
0,6 0.60.6

0,4
0.40.4
0,4 0.40.4
0 1 2 3 4 5 6
00 011 122 233 344 455 566 6 0 0 1 Bubble
1 2 2 3size
3 4 (mm)
45 5 6 6
Bubble size (mm) Bubble size
Bubble (mm)
size (mm)
Bubble size
Bubble (mm)
size (mm)

Figure 6 Area bubble to convex area ratio selection criteria comparison

5
Online Bubble Size Analyzer

Sampling and Imaging Device

The new online bubble size distribution analyzer, developed by IMASERV® is showed in Figure 7.
The sampling tube is used to collect and direct bubbles into the viewing chamber, which has an
inclined window to spread bubbles into a single layer, reducing the occurrence of overlapping
bubbles. To control bubble entrance, a pinch valve is connected to the sampling tube. The valve
operates with a micro compressor and a purge valve, which are both electrically controlled. The
chamber operates with a diaphragm pump in order to extract air. In this, liquid remains in the
chamber, eliminating drag forces in the sampling tube.
The device automatically fills and washes the viewing chamber with process water using solenoid
valves. Once the chamber is full of water, the pinch valve is opened to allow bubbles to rise from the
collection zone to the inclined window. At this point, the image acquisition software, based on
Octave, collects and stores a sequence of grayscale digital image (8 bits). A high resolution camera
(10 megapixels) is used with a macro lens in a 20 mm by 30 mm, a focal distance of 100 mm, and a
shutter speed of 1 ms to reduce bubble blurring. Backlighting is applied to produce bubble contours,
differentiating individual bubble groups from the solid dispersion. To reduce spot light and
overheating, a LED (light emission diode) screen is mounted at a distance of 100 mm from the
viewing area. This distance was tested to be appropriate to provide adequate lighting for the
measurement. The sampling tube is immersed at 1 m from the top of the froth and a purge tube is
connected at the bottom of the chamber.

Figure 7 On-Line bubble size analyzer

6
RESULTS AND DISCUSSION
The identification of the individual bubbles is done systematically through the following steps:
1. Thresholding: The step differentiates blobs from background by creating a binary image.
2. Segmentation: This serves to isolate or separate blobs from a binary image.
3. Blob features measurement: This determines geometric properties for each segmented
blob, creating a data structure.
4. Blob classification: This selects isolated bubbles based on geometric characteristics while
the remaining bubble clusters are stored.
5. Bubble cluster separation: This separates touching or super-imposed bubbles from the
remaining non-selected bubbles; its properties are added to the data structure.
6. Repeat from Step 1, modifying the threshold value (typically 5 to 12 times).
The bubble classification includes a model or the ratio convex area over object area, as a function of
minor axis object size, which allows for selection of irregularly shaped bubbles in comparison with
the commonly used shape factor for circles (Hernandez-Aguilar et al., 2002; Grau et al., 2002; Gomez
and Finch, 2002). An example of irregular bubbles detection is shown in Figure 8, where the detected
bubbles are shaded in blue and numbered in red.

A) Using Shape factor only B) using convex area criteria

99 1111
1010
1212
44 44
2424
88 1515 88
1313
1616
77 77 1414
22 22
66 1717 66

11 55 11 55
33 1818
2121
2020 33
2222
1919
2323

Figure 8 Detection of small bubbles and irregularly shaped bubbles

Image Validation
To determine the accuracy of the image analysis software, the resulting detected bubbles are super-
imposed over the original grayscale image. This allows for visual inspection of the detected object,
increasing both reliability and traceability of the technique. An example of original and processed
image is shown in Figure 9.

7
 Original image Processed image

Figure 9 Original image vs. processed image

As shown in the picture not only the irregular bubbles can be detected, but also small bubbles (0.1 to
0.3 mm) are also identified in the picture. To determine the bubble diameter from a 2D image, an
oblate spheroid model is used. To compensate for uneven light distribution, the use of variable
thresholding (5 to 12 levels) and subtraction of detected object has been shown to be the most effective
technique in accuracy and running time in comparison with digital background removal (Acuña,
2008).
Therefore, by employing of the algorithm based on convex area and multiple thresholding for bubble
size measurement, the detection of bubbles is four times higher in comparison to the McGill Bubble
Viewer technique.

CONCLUSIONS
The improved on-line analyzer appears to be feasible for online industrial measurements of bubble
size in flotation machines. The proposed algorithms allow for increase in reliability of the image
analysis technique by including small bubbles and irregularly shaped bubbles usually omitted in a
bubble size analysis.
As a result, 80% of the sample bubbles are neglected and therefore significant bias is produced in
bubble size measurement technique if only the shape factor criterion is used.
The proposed algorithm, based on convex area and multiple thresholding, appears to significantly
improve the reliability of the McGill Bubble Viewer technique, and therefore predict kinetics with
better accuracy.

ACKNOWLEDGEMENTS
This work was made possible by funds provided by the following parties: CONICYT (Comisión
Nacional de Investigación Científica y Tecnológica) under Fondecyt Project 11090269. The authors

8
would like to thank the financial support of Universidad Técnica Federico Santa Maria project
116.27.2.

REFERENCES
Acuña, C.A. (2008). Measurement Techniques to characterize bubble motion in swarms. Ph.D. Thesis.
Department of Mining and Materials Engineering, McGill University.
Bailey, M. (2004). Analysis of bubble size distributions using the McGill bubble size analyzer. Master
Thesis. Department of Mining, Metals and Materials Engineering, McGill University.
Bailey, M., Gomez, C.O., and Finch J.A. (2005). “A method of bubble diameter assignment”. Minerals
Engineering 18: 119-123.
Chen, F., Gomez, C.O. and Finch, J.A. (2001). Technical Note: Bubble Size Measurement in Flotation
Machines. Minerals Engineering, 14(4), 427-432.
Empix, Inc. (2002). Northern Eclipse help reference manual, version 6.0, www.empix.com
Finch, J.A., Dobby, G.S. (1990). Column flotation. Pergamon Press: Elmsford, New York, pp. 23-24.
Gomez, C.O., Finch, J.A. (2002). Gas dispersion measurements in flotation machines, Technical paper,
CIM Bulletin, Vol. 95 (1066), 73-78.
Grau, R.A., Heiskanen, K. (2002). Visual technique for measuring bubble size in flotation machines.
Minerals Engineering, 15, 507-513.
Grau, R.A., Heiskanen, K. (2003). Gas dispersion measurements in a flotation cell. Minerals
Engineering, Vol. 16, pp.1081-1089.
Grau, R.A. (2006). An investigation of the effect of physical and chemical variables on bubble
generation and coalescence in laboratory scale flotation cells. Dissertation for the degree of Doctor of
Science in Technology. Department of Material Science and Engineering, Helsinki University of
Technology.
Hernandez-Aguilar, J.R., Gomez, C.O. and Finch, J.A. (2002). A Technique for the Direct
Measurement of Bubble Size Distribution in Industrial Flotation Cells. Proceedings 34th Annual
Meeting of the Canadian Mineral Processors: CIM, 389-402.
Mena, P.C., Ruzicka M.C., Rocha, F.A., Teixeira J.A., Drahos, J. (2005). Effect of solids on
homogeneous-heterogeneous flow regime transition in bubble columns. Chemical Engineering
Science, Vol. 60, pp. 6013-6026.
Yianatos, J., Bucarey, R., Larenas, J., Henríquez, F., Torres L. (2005). Collection zone kinetic model for
industrial flotation columns. Minerals Engineering, 18 (15), 1373-1377.
Yianatos, J.B., Moys, M.H., Contreras, F., Villanueva A. (2008). Froth recovery of industrial flotation
cells. Minerals Engineering, 21 (12-14), November 2008, 817-825.

9
Enhancing the Performance of Cu-Mo Circuits Using
the Jameson Cell
Virginia Lawson
Glencore Technology, Australia

ABSTRACT
It has been argued that, molybdenite particles, due to their shape, may be more sensitive to
hydrodynamic effects than copper mineral particles. The platelet shaped molybdenite particles may
align along streamlines and thus have lower probability of collision with bubbles. Increasing
turbulence would increase collision frequency and efficiency, and therefore increase the rate of
particle collection.
In a Jameson Cell the high shear generated by the plunging jet, breaks the entrained air into a
multitude of very fine bubbles and provides increased bubble/particle collisions. The
hydrodynamic conditions for particle collection inside the downcomer and separation in the tank
are identical between laboratory, pilot plant and full scale Jameson Cells. Scale up is direct and
proven. This high intensity can provide the best conditions for improving the flotation of
molybdenite in bulk copper/molybdenite flotation.
Case study examples of pilot scale demonstration of improved molybdenite flotation in
copper/molybdenite flotation will be discussed.

1
INTRODUCTION
Flotation has been described by Araya et al (2013) as a complex multifaceted process and as a
flotation triangle of ore, chemistry and machine characteristics. Each plant and mineral system will
have more or less influence by each of these factors. When separating floatable mineral from non-
floatable minerals the use of machines with froth washing is important to minimise entrainment.
When recovering floatable valuable minerals such as chalcopyrite/molybdenite or floatable
minerals where only one is valuable such as chalcopyrite/pyrite, separation may rely on the ore and
the design of the process to treat it, chemistry and in particular machine features.
The most important factor affecting flotation devices is their ability to make small bubbles as this
dictates both the flotation kinetics and the carrying capacity. Ensuring effective bubble particle
collisions and maximising froth recovery will ensure high recoveries and effective use of installed
capacity. The Jameson Cell provides an excellent device to achieve moderate recovery at high
carrying capacity and excellent separation from non-floatable gangue. Its use where valuable
minerals are to be floated such as in the bulk flotation of copper minerals and molybdenite takes
advantage of its excellent collection efficiency and froth stability. The inclusion of wash water
maximises concentrate grade by eliminating entrained gangue minerals. Examples of the excellent
performance of the Jameson Cell for copper-molybdenum flotation and the supporting reasons are
outlined with examples from laboratory and pilot plant operations and the application of these
results to plant scale Jameson Cells.

THE JAMESON CELL

The Jameson Cell technology was invented in the late 1980s to overcome the design and operating
deficiencies of column and conventional flotation cells. From its first commercial installation in 1989
it has been continuously improved to make it more robust and easy to use. Over 340 have been
installed throughout the world in various flotation duties. The latest cells combine the original
advantages of small bubble size and small footprint with new low maintenance and operator-
friendly designs. The Jameson Cell is a flotation device driven by fluid mechanics. The advantages
of this innovative machine are:
 Consistent fine bubble generation with no external equipment or spargers.
 Intense mixing with small bubbles achieving rapid flotation without mechanical
agitation.
 High throughput in a small footprint.
 Froth washing maximizes concentrate grade in a single flotation stage.
 Fast response and easy control.
 Steady operation and performance irrespective of changes in feed flow.
 No moving parts, simple to install and maintain, excellent availability.
With conventional or column flotation technologies scale up factors are required when using
laboratory or pilot plant results for full scale design. These factors account for variations in cell
geometry, mixing patterns (short circuiting) and energy intensity between the different sized units.
Scale up factors may also change depending upon the duty, feed characteristics and flotation
kinetics.

2
For Jameson Cell design no scale up factors are required. This is because the jet velocity, air
entrainment and hydrodynamic conditions for mixing are identical across different sized cells from
laboratory to full scale. The operating principle and process parameters of the downcomer are
exactly the same irrespective of cell size. For large cell sizes simply more downcomers are used.
Direct scale up has been proven across different applications including coal, base and precious
metals, solvent extraction and industrial minerals. Two examples of the scale up from pilot plant to
full scale for base metals are given in Figure 1.

Figure 1 Examples of pilot plant and full scale demonstrating scale up

Continuous onsite pilot plant testing can provide risk mitigation for projects and provide the
metallurgical data required to justify the capital cost of mineral processing circuit modification. It is
not however necessary as the Jameson Cell can be also scaled up from laboratory scale work. The
laboratory work can be in a pilot L150 Jameson Cell or by performing dilution cleaning tests onsite,
as designed by Glencore Technology. These tests have been described by Huynh et al (2014) and
were effective as justification for several Jameson Cell installations to date. An example of the lab
dilution testing and pilot scale results is shown in Figure 2. These laboratory tests can be conducted
by plant operators and over time can demonstrate feed variability to better predict expected
improvements through entrainment elimination on different plant streams. Wash water addition in
pilot and full scale Jameson Cells is fundamental to the rejection of entrained gangue.

10 100
Pilot Plant Pilot Plant
8 80
NSG Recovery (%)
SiO2 Recovery(%)

Lab Dilution Test Lab Dilution Tests

6 60

4 40

2 20

0 0
0 20 40 60 80 100 0 20 40 60 80 100
Cu Recovery (%) Cu Recovery (%)

Figure 2 Comparison of pilot plant results and laboratory dilution flotation tests

3
Collision Efficiency
The flotation process comprises the collision between air bubbles and mineral particles with
mineral particles adhering or attaching to the air bubbles and subsequently being transported to the
froth phase. This whole process is called collection and the efficiency of collection, E, may be
expressed as:
(1)

where Ec is the collision efficiency, Ea is the attachment or adhesion efficiency and Es is the stability
efficiency. In the Jameson Cell there is a high mixing velocity and a large interfacial area and thus
there is rapid contact and capture of the hydrophobic particles by the bubbles. In effect, with high
voidage fraction, pulp is a thin film surrounding air bubbles. This mechanism where the flotation
tank acts as a disengagement vessel where froth is discharged from the bottom of the downcomer
into a quiescent environment enables high stability efficiency.
Particle recovery by flotation is sensitive to both particle size and contact angle (Crawford and
Ralston, 1988). The traditional recovery by size curve demonstrates that fine particles (<10 μm) and
coarse particles (> 100 μm) float poorly. The reason for the poor flotation of these particles differ;
fine particles are more likely to follow streamlines in a flotation cell and thus collision efficiency is
reduced, while coarse particles are more likely to be detached from bubbles due to the turbulent or
disruptive forces in the flotation cell. Improved flotation should be seen when both turbulence
increases bubble particle collection and a quiescent zone is provided to decrease particle
detachment and the froth recovery is maximised. The Jameson Cell provides this environment.

Carrying Capacity
Finch and Dobby (1990) described a model developed by Espinosa-Gomez et al (1988) to represent
the carrying capacity in flotation; which has been widely used across flotation technologies and
applications. The model was based on the particle size dp and the particle density p.

(2)

Pilot and full scale testwork determined a linear relationship for fine particles in pilot and small
industrial columns. Further work by Patwardhan and Honaker (2000) refined the model using data
from numerous laboratory, pilot and full scale operations of the Jameson Cell. The new model was
determined as;

(3)

In which

(4)
√

and

(5)

4
where d50 is the mean particle size, σ is the size distribution modulus, a, b, c, d, and k are constants,
db is the bubble diameter and Jg the superficial gas velocity in cm/s. bc is a constant with three
different values for laboratory, pilot and full scale columns; due to observed decreases in carrying
capacity due to increases in column cross sectional area and were determined experimentally. For
additional information on the key parameters refer Patwardhan and Honaker (2000).
Jameson Cells are renowned for high carrying capacities and have been reported to be higher than
columns and conventional flotation cells. Based on the model above the reason is due to the small
bubble size, high contact efficiency and high froth recoveries due to a quiescent zone for froth
recovery and small tank volumes and residence time. From equation (3) bubble size is a key
determinant of carrying capacity that distinguishes Jameson Cells. Evidence of small bubbles are
given in the work of McGill University and are shown in Figure 3.

3.5

3.0

2.5
Bubble Size (mm)

2.0
Column 1
Column 2
1.5 Mechanical Cell 1
Mechanical Cell 2
Mechanical Cell 3
1.0
Mechanical Cell 6
Mechanical Cell 5
0.5 Mechanical Cell 7
Jameson Cell

0.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0

Superficial Gas Velocity - Jg (cm/s)

Figure 3 Bubble size as a function of Jg for different flotation technologies

An example of demonstrated carrying capacity improvements can be seen in the example given in
copper flotation where the pilot Jameson Cell was operated on the column feed and achieved
significantly improved carrying capacities compared to the operating columns. Improvements up
to a factor of three were seen and are shown in Figure 4.

5
 10

Carrying Capacity (t/hr/m2)

L500 Pilot
Plant Column
6

0
0 20 40 60 80 100
Cu Recovery (%)

Figure 4 Carrying capacity for a pilot Jameson Cell compared to the operating column

Froth Recovery
Zanin, Grano and Ametov (2010) performed surveys around a copper/molybdenum cleaning
column to determine the relative contributions of the collection zone and the froth phase. To do
this they collected metallurgical samples from different points including a sample just below the
pulp/froth interface and performed size-by-size analysis. The copper behaviour was typical of that
seen in copper columns with the intermediate 30 micron particles having higher recovery than the
finer and coarser particles. The molybdenum recovery was very low with less than 10 per cent
across all size ranges. Their analysis determined that molybdenum grade was depleted across the
froth phase while copper was enriched. This led the authors to conclude that the froth recovery of
molybdenum is much lower across the column than the copper recovery.
Seaman, Franzidis and Manlapig (2004) suggested that the selective rejection of particles in the
froth may be the result of one of the following three mechanisms;
1. Detachment of less strongly attached particles as the aggregates arrive at the pulp/froth
interface. This would be caused by forces exerted on the aggregate as it rapidly changes
momentum on hitting the interface.
2. Selective detachment of particles from bubble surfaces during coalescence events
occurring within the froth and bubbles bursting on the froth surface. This is unlikely to
occur, it is expected that this process is non-selective due to the sudden rupture of bubble
lamellae.
3. Selective reattachment of particles that drop off bubbles during coalescence. This process
is not likely to occur in a highly loaded froth due to a lack of available sites for
reattachment.
Seaman, Franzidis and Manlapig (2006) propose that the froth recovery of attached particles in a
flotation cell is selective based on particle size, density and hydrophobicity. They showed the
selective transfer of attached particles across the froth phase. The pulp/froth interface was
responsible for a large degree of upgrading of the particles attached to bubble surfaces and also a
significant proportion of the recovery loss across the froth phase. Honaker and Ozsever (2003)

6
studied the detachment process for systems where the conditions converged from kinetic limiting
conditions to carrying capacity limiting conditions under conditions where entrainment was largely
eliminated. They noted a significant enrichment across the froth zone under high solids loading
conditions and concluded that the detachment process is selective, with particles having the
weakest bond with the bubble surface are preferentially detached.
Further work by Honaker, Ozsever and Parekh (2006) established that conditions where carrying
capacity limitations applied, resulted in improved flotation selectivity due to the selective
detachment as a result of bubble surface reduction and the reflux process that occurs between the
collection and froth zones. Plant based measurements of Rahman, Ata and Jameson (2015) using
the device described in Rahman, Ata and Jameson (2013) investigated the effect of plant variables
on the froth recovery. Not unexpectedly the air rate was found to have a significant positive effect
on the recovery of particles; in both the collection and the froth zones, although the decrease in
concentrate grade resulted as lower grade particles were recovered. The froth depth was also
found to have a significant effect where froth recovery decreased significantly as drop back
increased resulting in significant upgrading of particles. The plant results indicated a froth
recovery in the order of 75 to 85 per cent could be achieved with an appropriate choice of operating
parameters as might be seen in a continuous concentrator. As Jameson Cells operate at shallower
froth depths than columns this may contribute to improved froth recovery.
When trying to selectively sulphide minerals large circulating loads can be used to increase the
competition for bubble space to remove the less hydrophobic particles as circuits are operated close
to carrying capacity limitations. This has been reported in the separation of copper from nickel and
in the flotation of pentlandite from pyrrhotite (Lawson et al 2014). This same effect in
copper/molybdenum flotation is unintentional and undesired. Welsby (2014) determined that
molybdenum required greater froth stability to reach maximum recovery, likely due to competition
in the froth between highly hydrophobic copper minerals and moderately hydrophobic
molybdenite for the limited bubble surface area. Operating at levels of four to seven times the
critical coalescence concentration for the frother was required for molybdenum; whereas two to
three was adequate for copper. This would infer that the positive results seen in Jameson Cells may
include the contribution of improved froth stability.

The Flotation of Molybdenite

Molybdenite consists of a molybdenum atom surrounded by six sulphur atoms arranged as a
hexagonal layer structure. Although there is also a trigonal system it is less common (Castro et al,
2016). Within the layers the S-Mo-S are strong covalent bonds while weak Van der Waals forces
exist between the adjacent S-S sheets. As suggested by Triffett et al (2008) these weakly bonded
layers are easily pulled apart in grinding circuits. This structure and the resultant product of the
comminution circuit results in particles that may have strongly hydrophobic faces and inert edges.
It is these factors that are likely to be the main drivers of individual particle hydrophobicity.
Molybdenite is characterised as a hydrophobic anisotropic mineral (Laskowski, 2012). This group
also includes talc and graphite.
Ametov et al (2008) argued that this shape factor associated with molybdenite particles could be the
reason for the lower flotation recoveries in flotation circuits compared to the performance of copper
minerals in the same circuit. In their supporting work the use of increased turbulence was used by
decreasing the solids density in the pulp in both laboratory and plant circuits was intended to
demonstrate the increased turbulence at lower solids density would have a positive impact on the

7
bubble particle collision efficiency. The results of their work showed that although copper
recovery was unaffected by changes in flotation feed density molybdenite recovery was affected
and resulted in increased recoveries at lower flotation feed densities. This effect was demonstrated
in both the operating plant through surveys and in laboratory tests of the flotation feed. Copper
and molybdenite grades were improved at the lower feed density.
The suggestion from the work by Ametov et al (2008) is that both particle morphology as well as
cell hydrodynamics may be important in the recovery of molybdenite. The molybdenite particles
due to their shape are more likely to align along streamlines which would reduce their likelihood of
colliding with bubbles. The improvement in recovery from their work was from a significant
improvement from the coarser particles. This improvement in the coarse particles was also
hypothesized to be because of the high face-to-edge ratio compared to fine particles that have a
higher edge contribution to their surface behaviour.
Other work conducted in operating copper/molybdenite plants includes that by Hernandez-Aguilar
and Basi (2009) at Highland Valley Copper mine. This study focussed on the molybdenum circuit
and in particular the final two columns of the cleaning circuit. A decreased bubble size resulted in
a 4-5 fold increase in production rate production rate and recovery of the columns and a minor
increase in grade when a large proportion of bubbles less than 1mm in size were generated.
In addition to differences in flotation performance as a result of changing hydrodynamic conditions
Triffett and Bradshaw (2008) identified correlations between ore types that likely resulted in a
decreased molybdenite recovery when the ore types were blended. This decrease in performance
was more significant for molybdenite than copper minerals. Subsequent laboratory testing of these
hypotheses by Zanin et al (2009) confirmed that molybdenite is more sensitive to the operating
environment than copper sulphide minerals. Higher concentrations of Ca, Mg, Fe and K were
measured on the slow floating molybdenite which correlated to the presence of skarn ores.
Molybdenum flotation is affected by ore type as well as by its anisotropic nature. The flotation
conditions that were suggested in plant trials to improve its flotation can be correlated to similar
improvements in its flotation in Jameson Cells. Although hydrodynamic factors are significant
there will likely be occasion when ore type drives performance.

CURRENT PLANT CIRCUITS

Flotation circuits in both North and South America are generally designed to include a circulating
load across the column cleaner stage to enable high circuit recovery in spite of the generally low
molybdenum stage recovery in columns. This design feature has been reported by Zanin, Grano
and Ametov (2010) and Bulatovic, Wyslouzil & Kant (1998). Zanin, Grano and Ametov (2010)
discussed the performance of a typical copper/molybdenum flowsheet using conventional flotation
cells and flotation columns. In the example given the unit recovery across the cleaner column was
56 per cent for copper and only 7 per cent recovery of molybdenum. The overall plant recovery of
67 per cent was only achieved by a very high circulating load returning molybdenite to the column
feed. Through measurements of the solids located just below the pulp/froth interface Zanin, Grano
and Ametov (2010) concluded that the copper was enriched across the froth phase while the
molybdenum is depleted.
Some plant data are provided to demonstrate this behaviour in Table 1 where four sites operating
primary columns have copper recoveries of 55 to 65 per cent with molybdenum recoveries of 7 to 23

8
per cent. The cleaner scavenger circuits recover in excess of 95 per cent of the remaining copper
and molybdenum to generate high circulating loads to enable satisfactory overall circuit recoveries.

FEED Rougher FINAL TAIL

Regrinding

BULK CONC

Pri
Col

Cleaner
Scavenger

Figure 5 Typical bulk copper molybdenum flowsheet

Table 1 Plant stage recoveries

Cu Mo Cu Mo Cu:Mo
Recovery Recovery Upgrade Upgrade Selectivity
Site 1
Primary Col 59 17 1.4 0.4 3.5
Scavenger Col 84 76 1.3 1.3 1.1
Site 2
Cleaner 67 33 7.8 3.8 2.0
Cleaner Scavenger 94 92 1.1 2.8 1.0
Site 3
Primary Col 65 15 3.9 0.9 4.3
Cleaner Scavenger 98 92 2.0 2.0 1.1
Site 4*
Primary Col 56 7 2.3 0.3 8.0
Cleaner Scavenger 99 99 NA NA 1.0
Site 5
Primary Col 55 23 NA NA 2.4
* from Zanin, Grano and Ametov (2010)

9
JAMESON CELLS FOR PRIMARY CLEANING
From the discussion provided on molybdenum flotation and from the provided plant data
provided in Table 1, it can be seen that the recovery of molybdenum is low across the columns and
cleaner cells likely due to poor bubble size distribution, low collection efficiency, poor froth
recovery and potentially carrying capacity limitations. Given that these specific conditions are all
better in a Jameson Cell an improvement in performance would be expected.
Figure 6 compares laboratory and pilot L500 Jameson Cell results for four different operations. As
previously demonstrated the laboratory and pilot results agree indicating that laboratory data is
sufficient for demonstrating expected Jameson Cell performance. In two operations the existing
column circuits are also surveyed and a comparison is provided. In all cases the Cu:Mo selectivity
is improved in the Jameson Cell and high copper and molybdenum recoveries were possible at all
sites. The performance of each of the sites are slightly different and factors such as liberation,
chemistry and ore types are likely contributing factors.
100% 100%

80% 80%
Cu Recovery (%)
Cu Recovery (%)

60% 60%

40% 40%
Lab
Lab
20% L500 20%
Plant Column L500
0% 0%
0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
Mo Recovery (%) Mo Recovery (%)

100% 100%

80% 80%
Cu Recovery (%)

Cu Recovery (%)

60% 60%

40% 40%
Lab Lab
20% L500 20%
L500
Plant Column
0% 0%
0% 20% 40% 60% 80% 100% 0% 20% 40% 60% 80% 100%
Mo Recovery (%) Mo Recovery (%)

Figure 6 Jameson Cell performance for copper molybdenum primary cleaner

10
CONCLUSIONS
Evidence from testing at operating plants shows Jameson Cells are able to be scaled up directly
from laboratory to full scale. This is because the jet velocity, air entrainment and hydrodynamic
conditions for mixing are identical across different sized cells from the laboratory L150 cell to full
scale. In addition to the L150, traditional laboratory tests have been developed that simulate their
performance and can be done easily on site to determine the benefits of a Jameson Cell over a wide
range of plant operating conditions. Examples of scale up from laboratory L150 cell and from
dilution cleaning tests to L500 pilot tests and to full scale operation have been demonstrated.
The flotation of molybdenum in copper molybdenum circuits has often been seen in plant practice
to suffer low recovery in the primary cleaning circuit. To ensure acceptable plant recoveries a bank
of conventional cells is generally used as a cleaner scavenger to recover the molybdenum and
generate a high circulating loads back to the cleaner. The poor recovery of molybdenum compared
to copper can be the result of low collision efficiency, poor froth recovery and selective drop back
from the froth phase and possibly as the carrying capacity is exceeded. The use of a Jameson Cell in
this duty has several hydrodynamic advantages that have been demonstrated in pilot plant trials.
The smaller bubbles, intensive mixing and shallow froth depths enabling high froth recovery,
enable in some cases 1:1 copper to molybdenum recovery.

REFERENCES
Ametov, I., Grano, S.R., Zanin, M., Gredelj, S., Magnuson, R., Bolles, T., Triffett, B., (2008) ‘Copper
and molybdenite recovery in plant and batch laboratory cells in porphyry copper rougher flotation’
Proceedings of the XXIV International Mineral Processing Congress (IMPC), Beijing, China, Vol 1., pp.
1129-1137.
Araya, R., Huynh, L., Young, M., & Arburo, K. (2013) ‘Solving challenges in copper cleaning circuits
with the Jameson Cell’, Proceedings of Procemin 2013, Gecamin, Santiago. pp. 261-271.
Bulatovic, S. M., Wyslouzil, D. M., & Kant C. (1998) ‘Operating practices in the benefication of
major porphyry copper/molybdenum plants from Chile: Innovated technology and opportunities, a
review’, Minerals Engineering, vol. 11, No. 4, pp. 313-331.
Castro, S., Laskowski, J.S. (2015) ‘Depressing effects of flocculants on molybdenite flotation’,
Minerals Engineering, vol. 74, pp. 13-19.
Castro, S., Lopez-Valdivieso, A., Laskowski, J.S. (2008) ‘Review of the flotation of molybdenite. Part
I: surface properties and floatability’, International Journal of Mineral Processing, vol. 148, pp. 48-58.
Finch, J. A., & Dobby, G. S. (1990) ‘Column Flotation’, Oxford: Pergamon Press.
Hernandez-Aguilar, J. R. & Basi, J. (2009) ‘Improving Column Flotation Cell Operation in a
Copper/Molybdenum Separation Circuit’, Proceedings of the 41st Annual Meeting of the Canadian
Mineral Processors, Ottawa, Ontario, pp. 39-61.
Honaker, R. Q. & Ozsever, A. V. (2003) ‘Evaluation of the selective detachment process in flotation
froth’, Minerals Engineering, vol. 16, pp. 975-982.
Honaker, R. Q., Ozsever, A. V. & Parekh, B. K. (2006) ‘Selective detachment process in column
flotation froth’, Minerals Engineering, vol. 19, pp. 687-695.

11
Huynh, L., Araya, R., Seaman, D. R., Harbort, G., Munro, P. D. (2014). ‘Improved cleaner circuit
design for better performance using the Jameson Cell’, Proceedings Twelfth Mill Operators Conference,
Townsville, Queensland, pp. 141-152.
Kelebek, S., Yoruk, S., & Smith, G.W. (2001) ‘Wetting behaviour of molybdenite and talc in
lignosulphonate/MIBC solutions and their separation by flotation’, Separation Science and Technology,
vol. 36 (2), pp. 145-157.
Laskowski, J.S. (2012) ‘Anisotropic minerals in flotation circuits’, CIM Journal. Vol 3., No. 4. Pp.
203-213.
Lauder, D. W., Mavotoi, M., & Glatthaar, J. W. (2003) ‘Fluorine Removal from Ok Tedi
Copper/Gold concentrates’, Proceedings Eighth Mill Operators Conference, Townsville, Queensland,
pp. 203-209.
Lawson, V., Hill, G., Kormos, L., Marrs G. (2014) ‘The separation of pentlandite from chalcopyrite,
pyrrhotite and gangue in nickel projects throughout the world’, Proceedings Twelfth Mill Operators
Conference, Townsville, Queensland, pp. 153-162.
Patwardhan, A., Honaker, R. Q. (2000) ‘Development of carrying capacity model for column froth
flotation’, International Journal of Mineral Processing, vol. 59, pp. 273-293.
Rahman, R. M., Ata, S., & Jameson, G. J. (2013) ‘Froth recovery measurements in an industrial
flotation cell’, Minerals Engineering, vol. 53, pp. 193-202.
Rahman, R. M., Ata, S., & Jameson, G. J. (2013) ‘Study of froth behaviour in a controlled plant
environment – Part 1: Effect of air flow rate and froth depth’, Minerals Engineering, vol. 81, pp. 152-
160.
Seaman, D. R., Franzidis, J-P. & Manlapig, E. V. (2004) ‘Bubble load measurement in the pulp zone
of industrial flotation machines – a new device for determining the froth recovery of attached
particles’, International Journal of Mineral Processing, vol. 74, pp. 1-13.
Triffett, B., Veloo, C., Adair, B.J.I., & Bradshaw, D. (2008) ‘An investigation of the factors affecting
the recovery of molybdenite in the Kennecott Utah Copper bulk flotation circuit’, Minerals
Engineering, vol. 21, pp. 832-840.
Welsby, S. D. D. (2014) ‘Pilot scale frother testing at Highland Valley Copper’, Proceedings of the 46th
Annual Meeting of the Canadian Mineral Processors, Ottawa, Ontario, pp. 301-314.
Zanin, M., Ametov, I., Grano, S., Zhou, L. & Skinner, W. (2009) ‘A study of the mechanisms
affecting molybdenite recovery in a bulk copper/molybdenum flotation circuit’, International Journal
of Mineral Processing, vol. 93, pp. 256-266.
Zanin, M., Grano, S. & Ametov, I. (2010) ‘Technical challenges in the flotation of molybdenite from
porphyry copper ores’, Proceedings of the XXV International Mineral Processing Congress (IMPC),
Brisbane, Queensland, Australia, pp. 2651-2661.

12
Determination of Collector Affinity for Flotation of
Sulfide Minerals
Jocelyn Quinteros1, Robinson Constanzo1, Gabriela Gioia1, Eduardo Enriquez2 and Nefer
Valenzuela2
1. Escuela de Ingeniería Química, Pontificia Universidad Católica de Valparaíso, Chile
2. Clariant, Chile

ABSTRACT
One of the important factors affecting the performance of froth flotation is the type of reagent and
the doses applied. The reagents can be the type of frothers, collector, depressants, etc., and are
chosen according to the nature of the mineral particles desired to be collected. In the case of
collectors, they provide hydrophobic characteristics to the valuable minerals, allowing a more
selective separation of the valuable minerals if the correct collector is used. In order to choose the
correct suit of reagents it’s important to study the affinity between collectors and minerals.
This paper aims to assess the affinity of a suit of collector reagents used in mineral flotation with
different pure minerals. A future work will involve studying real ores. The affinity protocol
considers the following aspects: first use of microflotation tests for pure minerals, and a second
stage to study the different absorption levels of the collector into the mineral surface using cyclic
voltammetry measurements. The pure minerals studies were chalcopyrite, chalcocite, covellite and
pyrite using different collector types. These minerals were floated at three different pH´s in order to
find the best pH condition for each particular collector studied. Cyclic voltammetry was used to
correlate this information with the results of flotation performance.
The results of this research allow classifying the collectors according to the degree of affinity within
each mineral of interest and for a specific pH. This offers technical and scientific support to
recommend the use of particular reagents in the operation according with the nature of the
minerals and its mineralogy. Preliminary results from flotation indicate that the best affinity for the
pure minerals are: collector 1 for covellite at pH 8.5, collector 3 for chalcocite at pH 8.5 and for
pyrite at pH 10.5 and collector 2 for chalcopyrite at pH 8.5.

1
INTRODUCTION
The flotation process requires a lot of reagents used principally for the recovery of minerals and
cleaning the concentrate to obtain a desire product. In this process the physicochemical properties
of the surface particles take advantage to effectively selective collection of minerals in liquid by
adhering gas bubbles or air to provide selective levitation of the solid particles. Air bubbles have
the sufficient force to carry certain mineral particles to the surface where stable foam retains the
mineral and allows removal as concentrated. While minerals that not preferentially adhere to the
air bubbles remain submerged and sink to the bottom as tailings.
The additions of different reagents such as, frothers, depressants, activators, pH modifiers and
collectors made possible the flotation process to be selective. Specifically collectors are adsorbed on
the surface of the minerals of interest, conferring hydrophobic properties which allow flotation
species to take place in a cost-efficient manner. Collectors are complex molecules, not symmetrical
in structure called heteropolar, consisting of two parts: a polar and a non-polar. The non-polar
portion is hydrocarbons radical oriented towards the water and practically not react with this and
generates the properties of water repellency, while the polar part can react with water and is
oriented towards the surface of the mineral. This orientation makes mineral water repellent.
The affinity of collectors is a very complex process that has been extensively studied. This
phenomenon is measured as the ability of the collector adsorbed on the surface of minerals. In
general, several requirements must be met to achieve good adsorption. These include:
a. Relative immobilization of the collector species. This immobilization increases the
capacity to resist the movement of fluid along the plane adjacent to the solid without
displacement, thereby preventing complete wetting of the solids.
b. Development of sufficient hydrophobic character to withstand the mechanical and
dynamic effects of flotation particular system.
c. Collector-modifier selection system that allows adequate control of selectivity.
d. Proper selection of a collector-foaming system that provides satisfactory kinetics
separation to be achieved. (Bulatovic, 2007)
According to Gardner and Woods (1979), several mechanisms have been proposed to explain the
mineral-collector interaction; the most accepted is the electrochemical. The electrochemical process
involves that oxygen is reduced and the collector is oxidized simultaneously to form a mixed
potential, these processes are conditioned mainly by two reactions. The selectivity of the flotation
process depends on the relative rates of these reactions (Ortiz et al., 2000).
1
Reaction 1 : 𝑂2 + 𝐻2 𝑂 + 2𝑒 − → 2𝑂𝐻 −
2
Reaction 2: 2𝑋 − → 𝑋2 + 2𝑒 −

Where, X- is the ion collector and X2 represents ion oxidation.

Some researchers have suggested that the presence of oxygen is absolutely necessary for the
adsorption of sulfides collectors, while others have found that the presence of large amounts of
oxygen to be damaging to the adsorption of xanthate (Bulatovic, 2007).
Walker et al., (1984) and O'Dell et al., (1986), have calculated the potential of transformed the
mineral surface into a hydrophobic mineral surface. Hepel et al., (1977) indicated that for chalcocite
(Cu2S) the calculated potential for the formation of the species becomes hydrophobic mineral
surface is -0.11 V at pH 8 and -0 24 V at pH 11. Ortiz et al., (2000), states that the potential at which

2
begins reduction of chalcocite is -0.25 V and is still stable to -0.5 V as shown in Figure 1, this
research was done a basic pH in the absence of collector.

Figure 1 Cyclic voltammetry of Chalcocite (Ortiz et al., 2000)

In the case of pyrite (FeS2) López Valdivieso (2005) studied the floatability of pyrite in the presence
of ethyl xanthate as a function of pH. He demonstrated that at low concentrations of collector, the
floatability of it decreases in the range of pH 5 to 9. Voltammetry made by Ortiz et al., (2000), shown
in Figure 2, indicates that at 0.1 V , in the sense cathodic, the surface is coated with a layer formed
by Fe(OH)3 and S, which is modified by reduction of the S at -0.1 V. When the potential becomes
more negative, the beginning of the reduction of Fe (III) to Fe (II) is at -0.45 V. Trahar (1984), found
that the optimum potential for pyrite flotation in the presence of butyl xanthate, is in the region
where the hydrophobic species is formed, 0.1 V, but also note that the float starts at -0.15 V. Ralston
(1988) demonstrated that pyrite present low floatability in the presence of butyl xanthate collector
in a range of potential between -2.5 V to 0.1 V. Whether the potential is increased to 0.35 V it is
possible to reach a recovery of approximately 75% of pyrite.

Figure 1 Cyclic voltammetry of Pyrite (Ortiz et al., 2000)

Tejeda (2005) studied the natural hydrophobicity of chalcopyrite (CuFeS2) at different potentials.
The results of this research are illustrate in Figure 3, showing that at low pH chalcopyrite present
natural hydrophobicity in a wide range of potential, which decreases as the pH becomes more
alkaline. The reason of increasing the pH is because the formation of certain hydroxides on the
mineral surface. Additionally, it can be inferred that the floatability of the chalcopyrite in the

3
absence of collector starts near 0 V. Different authors have shown that chalcopyrite present
floatability in a range of pH 4 to pH 10 with the presence of collector. The region of
maximum recovery obtained was between 0.1 V and 0.3 V, however flotation begins at -
0.25 V increasing with recovery.
According to Vargas in 2005, recovery is increased when the oxidation of collector is
favored. As the oxidation of the collector and reduction ore compete in mixed potential,
we can infer that a lower current response in the reduction zone is due to a high current
response in the oxidation zone.

Figure 2 Chalcopyrite natural floatability depending on the oxidation potential and pH (Tejada, 2005)

Figure 4 illustrates the results from Ross (1990), who studied the covellite (CuS) under
microflotation test with xanthate. It can be concluded that in the first region of potential the
covellite can be subjected to reduction, producing elemental copper or Cu2S. If the kinetics of
reduction of covellite in this region of potential it is sufficiently fast, then is expected that the
floatability increases due to the reaction between elemental copper and xanthate ion producing
species Cu-EtX. In fact, an average recovery of 10% in this region is obtaining. This suggests that the
reduction of covellite it is not significant enough to influence their floatability.
The start of floatability is -200 mV vs SCE which can be reasonably correlated to the direct reaction
between covellite ions and xanthate. It is interesting to note that a recovery of 62% represents the
maximum floatability. This can be achieved due to the possibility of dixanthogen to participate in
the flotation of covellite (Ross, 1990).
This project aims to study and establish the affinity between different species of collectors with
different types of minerals of interest, being an key knowledge in order to establish a criterion to
recommend a specific collector to the mineral industry.

4
 Figure 3 Microfiltration results of covellite in an electrochemical cell (Ross, 1990)

METHODOLOGY
To develop this research different pure minerals were used namely chalcopyrite, chalcocite,
covellite and pyrite, were tested in microflotation test using four different collectors provided by
Clariant. To perform the microflotation tests an experimental design was developed in order to
tested the adsorption of collectors in the pure minerals.

MICROFLOTATION TESTS
Microflotation tests were used to carry out the determination of the affinity with certain collectors
with pure minerals, synthetic and natural. These tests were performed by flotation research group
at PUCV. The total experimental design indicates that 48 experiments need to be carried out in
order to complete the experiments. For each mineral 12 experiments were performing, testing four
different of collectors at three different pH’s. Figure 5 shows the microflotation test of one
experiment.

Figure 5 Microflotation apparatus

5
To perform the microflotation test a flotation pulp must be prepared before the test started with 2%
in solid and 30 g/t of collector. This pulp was conditioned for 10 minutes. Whether the pH needs to
be modified lime was used to get the desired pH. An aliquot of 50 ml was immersed into the pulp
to reduce the error assuming a perfect mixing and then it was introduced in the Hallimond tube
together with a magnetic stirrer, 100 ml of distilled water was added and MIBC was used as frother.
Then, the air was turn to start the flotation process. The concentrate was collected during 10
minutes. After that, the air was tur off and the concentrate and the tailing was filtered. After each
flotation test the electrochemical experiments were performed.

ELECTROCHEMICAL TESTS
The absorptivity of collector on the mineral particles is evaluating by linear voltammetry technique.
This technique involves the generation of kinetic curves by a potential sweep in time, generating a
response current in the system.
To perform this test Dropsen electrodes DRP-110 were used with a Potentiostat Voltalab PGZ 301
(Figure 6) and a mineral paste of the concentrate was produced. To produce the electrode paste the
mineral and the graphite were mixed to produce a homogeneous paste. The mixture was
emplacement on the dropsen and was ready to perform the voltammetry tests. To carry out the
voltammetry test, the equipment is turned on and the parameters were set. The potential sweep is
carried out in cathodic direction to observe ore reduction. Linear voltammetry’s were done and the
measurement for each mineral was done in an open circuit measured the potential until -900 V,
with a scan rate of 20 (mV/s) at room temperature.

Figure 6 Electrochemical test system

6
RESULTS AND DISCUSSION

CHALCOPYRITE

Microflotation Test
The results of the microflotation test for chalcopyrite are shown in Figure 7, the percent of mass
recovery is presented as a function of four collectors studied. The graph does not indicate a clear
trend in the optimal pH for flotation of this pure mineral. However, it can be observed that
Collector 2 has the highest recovery for pure chalcopyrite with over 86% mass recovery in the range
of pH studied. This collector reaches 93.7% of mass recovery of chalcopyrite, indicating that it is the
most suitable collector for recovering CuFeS2 at pH 8.5. Also, it can be observed that pH 10.5 had
higher recovery for almost all collectors under study except for collector 3, which at this pH had
obtained the lower recovery. This pH is appropriate for the most industrial companies and even
collector 2 at pH 10.5 got the highest mass recovery of the all collectors with 86.7%of mass recovery.
The less selective collector for chalcopyrite is collector 1 at pH 8.5 with a recovery of 23.1% of mass
recovery.

Figure 7 Recovery of Chalcopyrite collectors standard dose of 30 g/t ore

Electrochemical Results
Linear voltammetry was applied to electrodes prepared from chalcopyrite concentrate´s at
pH 8.5, in order to compare the behavior of the collectors at the same pH in which the
highest mass recovery was obtained. Figure 8 indicates that Collector 2 has the lowest
response current in much of the reduction zone from -0.8 V to -0.4 V and from -0.1 V to 0
V, which indicates that oxidation of collector it is greater, thus obtaining the greatest
recovery at this pH. Collector 3 can be used in a range of potential between -0.4 and -0.1 V

7
of the flotation pulp to get some mass recoveries. While the collector 4 and 1 had lower
recoveries, which are not recommended for stable potential intervals.

Figure 8 Linear voltammetry’s Chalcopyrite

CHALCOCITE

Microflotation Test
The results of microflotation (Figure 9) show that pH 8.5 is the most suitable pH to recover
chalcocite in all the reagents except for collector 2. Collector 3 presented the highest mass
recovery with 58.3%, then collector 1 and then collector 4 with recoveries fairly similar
over 40%. Something interesting is that higher pH´s i.e., 9.5 and 10.5 get the lowest mass
recovery of chalcocite below 20%. Collector 2 is not the proper collector to be used with
pure chalcocite.

8
 Figure 9 Recovery of chalcocite vs standard dose of 30 g/t ore

Electrochemical Results
From the literature review, as Basilio (1985) and Heyes and Trahar (1979) they showed that
the chalcocite presents floatability at pH 8 from -0,3V/SHE to 0.4 V/SHE, which is in
accordance with the flotation pH of pure chalcopyrite. Therefore, if we set a range for the
flotation of chalcocite this ought to be between -0.3 V/SHE, in this area according to the
literature low recovery values are obtained, which increases as increasing the potential to
the formation of hydrophobic entity at -0.1 V and then to 0.4 V/SHE.
Figure 10 shown the voltammetry’s applied to all collectors at pH 8.5. It can be observed
that collector 3 is oxidized in large quantities and quickly in the range of -0.4 to -0.3 V, this
can be explained due to high current response showing in graph and high recovery
showed in microflotation test. With respect of collector 1 and collector 4 shows a low
current response resulting in a low recovery, showed a low current response is due and
the same can be applied for collector 2, which does not show any current response to the
voltammetry test.

9
 Figure 10 Chalcocite linear voltammetry’s

COVELLITE

Microflotation test
Figure 11 shows the mass recovery results for all the pH range under study and indicates
that as the pH decreases gradually the mass recovery decreases. The best pH for
recovering covellite is pH 8.5 for collector 1 with 47.1% of mass recovery. Collector 3
shows the lowest flotation affinity with pure covellite.

Figure 11 Recovery of covellite vs standard dose of 30 g/t ore

10
Electrochemical Results
From literature it is known that anodic covellite reduction is thermodynamically stable in
alkaline solutions from -0.35 V/SHE to 0 V/SHE, so the start of flotation is 0 V/SHE to 0.4
V/SHE. Figure 12 shown the voltammetry’s applied to all collectors at pH 8.5. It can be
observed that flotation start between 0 V/SHE to 0.4 V/SHE, in that range of potential
collector oxidation occurs to a mixed potential. In this range, the Collector 1 has the lowest
current response with respect to the rest of the collectors, thus obtaining a greater
recovery. In the range of potential from -0.02 to 0.03 V, the recommended pulp collector is
the Collector 1, which has values lower current, having a higher oxidation collector. While,
in potential ranges from -0.02 V to negative values and from 0.03 V to positive values the
recommended Collector 4.

Figure 12 Voltammetries linear covellite

PYRITE
Figure 13 shows the results of microflotation with pyrite. For this, mineral tests show very
low recovery rates. This is because it could not be done magnetic stirring at the time of
testing as it is a paramagnetic mineral and adheres to the agitator. Without stirring, it was
observed that the input air is not able to raise the pyrite leaving part of it in the bottom of
the tube.

11
 Figure 13 Recovery of Pyrite vs dose standard of collector of 30g/ton of ore

As this happened with all tests, the same associated error is generated. Therefore, from the
graph it can be seen that the collector greater recovery was obtained for Collector 3, at pH
10.5 with a recovery rate of 30%.Also, it is important to mention that the collector studied
does not represent a strong affinity with pyrite.

Electrochemical Results
Linear voltammetry was applied to this concentrate. Figure 14 shown the voltammetry’s
applied to all collectors at pH 10.5. From the literature review, we know that at pH
between 5 and 9, the floatability decreases due to the formation of iron oxyhydroxide on
the surface. We also know that the optimum potential for pyrite flotation is in the range
from 0.1 V / SHE. Since this value, we can see that has a low current response as oxidation
collector is greater than ore reduction, generating more hydrophobic entities achieving
greater recovery also this response current is in the cathodic area.

12
 Figure 14 Pyrite linear voltammetries

CONCLUSION
From the results, it can be seem that there is an affinity interaction between some collectors
and the pure mineral species, being the hypothesis confirmed. The development of this
protocol allows supporting the sales decisions providing a robust tool and application-
level to the industrial supplier. The combination of the microflotation test and linear
voltammetry’s and using the literature available it was possible to verified that results of
obtained are in accordance with the scientific background. Finally, defining the collectors
and optimum pH for each mineral are presented in Table 1, representing the collector
affinity with each of the ores studies in this project:

Table 1 Summary of the best collector performance for each pure mineral

Mineral Collector pH
Chalcopyrite Collector 2 8,5
Chalcocite Collector 3 8,5
Covellite Collector 1 8,5
Pyrite Collector 3 10,5

REFERENCES
Basilio, C., Pritzker, M.D., and Yoon, R.H., (1985).Thermodynamics electrochemistry of the chalcocite –
potassium ethyl xanthate system. 114th Annual Meeting, New York, Reprint No. 85-86.

Bulatovic, S. M., (2007). Handbook of flotation reagents. Elsevier Science & Technology Books.

Feasby, D. G., (1996). Investigation of Hallimond-tube flotation of low grade phosphate material.

13
Gardner J.R., and Woods R., (1979). An Electrochemical Investigation of the Natural Flotability of
Chalcopyrite, International Journal of Mineral Processing, 6, 1-16.

Hepel, T. and Pomianowski, A., (1977). Diagrams of electrochemical equilibria of the system copper-
potassium ethyl xanthate-water at 25°C. Int. 1. Miner. Process, 4: 345- 361
López Valdivieso A., L. S., (2005). On the cathodic reaction coupled with the oxidation of xanthates at the
pyrite/aqueous solution interface. International Journal of Mineral Processing , 77, 154-164.

Maung A., Min, A. V., (2013). An exponential decay relationship between microflotation rate and back-
calculated induction time for potential flow and mobile bubble surface. Minerals Engineering , Vol. 40, pp.
67-80

O'Dell, C. S., Walker, G. W., Richardson, P. E., (1986). Electrochemistry of the chalcocite xanthate
system.1. Appl. Electrochem., 16: 544- 554

Ortiz, J., Barbato, S., and Gautier, J.L., (2000). Reducción del oxigeno sobre calsosita, pirita y galena en
medio alcalino. Boletín de la sociedad Chilena de Química, Santiago, Chile. V. 45 n.3.

Ralston, R. A. (1988). The collectorless flotation and separaton of sulphide minerals by Eh control.
International Journal of Mineral Processing (23), 55-84.

Roos, J. P. (1990). Electrochemical control of chacocite and covellite-xantahte flotation. International

Journal of Mineral Processing , 17-30.

Tejada, J. C. (2005). Adsorción de los colectores xantatos e isopropil tionocarbamato y del depresor dextrina
en la interfase calcopirita/solución acuosa. Efecto de la adsorción en la flotabilidad de calcopirita.

Trahar, W. J. (1984). The influence of pulp potential in sulphide flotation. In: Principles of mineral
flotation. Victoria, Australia.

Walker, G. W., Stout, 1. V. III, Richardson, P. E., 1984. Electrochemical flotation of sulphides: reaction of
chalcopyrite in aqueous solution. Inter. 1. Miner. Process, 12: 55- 72

14
Slag Flotation at El Soldado
Miguel Becerra and Jaime Zúñiga
El Soldado, Anglo American, Chile

ABSTRACT
Historically El Soldado operation has processed slag from Chagres smelter, whose average total
copper grades were in a range between 2 % and 4 %. Since the incorporation of the Flash and
refining process these grades began to decline gradually due to greater circuit efficiency beside
with an increase in melting capacity, generating lower grades than 2 %.
From 2013 formally was incorporated the low grade slag processing (1.1 % total copper) in the
production plans of El Soldado due to specific contingencies in the mining operation. The definition
of the circuit to be used, as to metallurgical performance, were made from a series of laboratory
tests carried out internally and externally in the operation. With these parameters defined, it began
with the processing of slag massively to October 2013. From this date have made several
modifications to the operational strategy in order to improve metallurgical results of the circuit.
During 2014 and 2015, additional slag laboratory tests were done and columns cleaning technology
used on the flotation process was incorporated in the analysis. Tests and mineralogical analyzes
were performed and hydrodynamic analysis were incorporated into the cells used to enrich the
knowledge of the operation. At the same time, the characterization of the dump source of the slag
was improved to include variations thereof on the flotation operation.
After two year and a half of operation, has established that the slag recovery is more sensitive to the
feed size distribution and copper feed grade compared to copper sulfides. Copper grade 1.1 %
(with slow cooling slag processl) and a size D80
same time, has confirmed the essential role of the regrind stage to achieve copper concentrate
grades near to 20 %. Without this step, the concentrate grades to achieve no more than 15 %.

There is no full article associated with this abstract.

1
Advances in Flotation Chemistry: Xanthate
Alternative Products
Nefer Valenzuela1, Juan Román1 and Jacques Collin2
1. Clariant, Chile
2. Clariant, Germany

ABSTRACT
In the concentration of sulfide ores using flotation, collectors play a major role. Without them, the
limited natural floatability of minerals leads to disappointing outcomes.
Among the various chemical families of existing collectors, xanthates are known for its wide use in
the industry, either as a single collector or together with other sulfide-ore collectors to maximize
synergies.
Handling and use of xanthates pose several risks. Specially designed warehouses are necessary due
to its spontaneously combustible character. In the presence of water or moisture, xanthates degrade
quite rapidly, as a function of temperature, to form carbon sulfide which is in turn classified as a
toxic, highly flammable compound. Additionally, shipping companies are increasingly reluctant to
accept it; moreover, once empty, packaging becomes hazardous waste material, thus turning disposal
or re-use very challenging.
An alternative to xanthates offering similar metallurgic results but without the associated risks would
be highly desirable.
This paper discusses laboratory and plant-scale flotation results using alternatives to sodium isobutyl
and isopropyl xanthates. Trial ores were supplied by operations both in Chile and Australia.
Successful results have led to the approval and/or use of alternative collectors to xanthate.

1
INTRODUCTION
In the flotation of copper sulfide minerals, flotation reagents are mainly used to control the surface
chemistry of minerals in order to obtain a better affinity of the element of interest with the air bubble,
which is the main transport medium to its further beneficiation. Reagents used in flotation are
classified in several categories, such as modifiers, collectors, depressants, frothers, flocculants,
dispersants, etc.
Traditionally, reagents are commercialized as either commodities or specialty chemicals. Several
suppliers are able to manufacture commodities and these are sold purely on price basis, without a
technical support associated with its application in the mining operation. On the other side, specialty
chemicals (collectors, frothers, depressing agents) are sold based on their metallurgical performance,
and they require specialized technical support. These chemicals are protected by patents and
generally, they have higher prices.

Chemical components
After reviewing the development of the flotation reagents used these days, it can be said that the
majority of these products were introduced to the market more than 50 years ago. Only a small share
of reagent sales (less than 10%) can be classified as specialty chemicals, based upon traditional
criteria.
Before the 80’s, several research organizations were involved in the development and
commercialization of mining reagents: chemical companies, industrial research institutions the U.S.
Bureau of Mine and Universities, among others. In the majority of cases, a combination of
organizations participated in the R&D process.
In addition, the expiration of patents eliminated the natural constraints to manufacture “me-too”
products, among which xanthate can be found. The sales of these products are only price-driven,
which generates very low profit margins. Therefore, the development of new chemistries that create
both benefit and a competitive position in the constantly growing reagent market is required.

Flotation Collectors
As it is well known, collectors’ main purpose is to give hydrophobicity to mineral surfaces. Collectors
are heteropolar organic compounds, whose active part is the polar group in the molecule, which
binds them to the mineral surface, through physical-chemical adsorption. This mechanism can be
explained by partial elimination of the hydrated layer, and the formation of a solid-liquid-gas contact
interaction.
The following main groups can be distinguished among collectors:

Xanthates

2
 Dithiophosphates

Dithiocarbamates

Xanthates are the salts of xanthic acids, which are produced through reaction of an alcohol with
carbon disulfide.

This family of collectors has been extensively used thanks to their low cost, strong collecting
properties and a good selectivity. However, xanthates´ application is complex because of its
impurities and decomposition products. In addition, since xanthates are in solid state, their
preparation generates difficulties and additional costs to the operation. Furthermore, the added risk
of spontaneous combustion makes xanthates difficult to handle.

METHODOLOGY
The present work shows the methodology that lead to the development of a viable and attractive
alternative to replace xanthates. This process’s goal is to identify the customers’ basic needs in order
to define the minimum requirements that must fulfill the selected candidate that will be tested
initially at bench-scale, then at a plant trial.
The background of our work is based upon the following working scheme, as seen on fig. 1:
Workshops: several meetings were held with the main customers that use xanthate, in order to define
the objective and scope of the project.
Problem definition: the problem identification is key in the search of alternatives. For this project,
we wanted to “Reduce or eliminate the risks that are associated to the use of xanthates”.
Proposed solution: to find an alternative to xanthate, which is effective in terms of metallurgical
performance, with a competitive cost and that reduces the risks associated to the use of xanthates
onsite.

3
 Customer's Product
Idea identification
unmet needs specification

Succesful Hypothesis and

Products Experiments

Figure 1 Scheme of idea identification

The results obtained in the meetings with customers, allowed defining a matrix of unmet needs,
which according to their complexity, were classified in three groups, see fig. 2:

Handling and Environment and Metallurgical

Preparation Safety Performance

Decomposition and
Health and safety risks
Delivered in powder or caking of the product
during cleaning of
pellets when exposed to
tanks and sumps
moisture

Storage of products
Decompostion of the
with risk of
Xanthate preparation solution with time and
spontaneous
high temperatures
combustion

Contaminated Potential risk due to

packaging must be decomposition
properly eliminated (releasing of CS2)

Figure 2 Scheme of unmet customer needs

Once the customer’s needs were identified, it was required to define the specifications of the new
product that allowed it, in some way, fig. 3, to be more attractive than its current competitor. Then,
determine through a simple design of experiments, the components that will be tested in flotation
trials that would lead to the final development of the new product.

4
 Non flammable liquid Soluble in water
• Delivery in trucks • Easily dissolved in water
• No powder/pellet handling • Flexible addition point to the process.
• Avoid storage of contaminated packaging • No filter-clogging solid residues.
• Simplification or elimination of any reagent
preparation procedures.

Product Specifications

Equal or better metallurgical performance Effective replacement as either primary or

• Achieve a similar recovery and grade. secondary collector
• Storage time or temperature must not affect • Good performance as a primary collector
recovery. • Delivers performance and allows for flexibility
• Changes in pH must not affect recovery. when used with other collectors

Figure 3 Scheme of product specifications

COLLECTION MECHANISM
The new selected products for the experimental design belongs to the families of DTP (fig. 4), which
are classified into ionic reagents and presented as a water soluble salt at alkaline pH.

Figure 4 Scheme of classification of colectors

Their solubility decreases as the pH does, due to the conversion of its acidic form, is also insoluble in
water.
DTP form insoluble salts and/or organic complex through the electron interaction of sulfur atoms
and they can be adsorb on the surface of some sulfides only by a chemical mechanism, in the case of

5
chalcopyrite and pyrite, the adsorption leads quickly to the conversion of DTP to mixed disulfide by
an electrochemical reaction.
This disulfide species (dithiophosphatogen) insoluble in water, is extremely hydrophobic and it is
adsorbed on the surface of both minerals, stimulating the flotation. Something similar happens when
xanthates transform into dixanthogen.

EXPERIMENTAL DESIGN
To formulate the new product, a design of experiments with four variables was used to identify the
component combinations that generate positive impact in the main results of flotation (Cu, Mo and
Fe recovery, mass pull), see table 1, in order to develop the product in the R&D laboratory and, finally,
perform the tests with the formulations that fulfill the unmet needs established by each customer.
The formulations were tested in three different ores using the same approach. The following table
shows the obtained combinations:

Table1 Scheme of DOE

Pure Product
Test P1 P2 P3 P4
1 X
2 X
3 X
4 X
5 X X
6 X X
7 X X
8 X X
9 X X
10 X X
11 X X X
12 X X X
13 X X X
14 X X X
15 X X X X

The relative magnitude and statistical significance, both in the main effects and in the interactions
between components were identified with the help of Pareto diagrams. In this particular example,
the effects are arranged in decreasing order, where the reference line in the diagram indicates which
effects are significant, fig 5.
The interactions delivered by the different combinations along the performed tests delivered the
following results:

6
 Pareto from combined efects Pareto from combined efects
(Rec Cu. α = 0,05) (Rec Fe. α = 0,05)
Término 0,216 Término 0,182
BCD Factor Name C Factor Name
BC A P1 D A P1
C B P2 B B P2
C P3 C P3
AC CD
D P4 D P4
AD BCD
CD A
ABC ACD
ABCD ABCD
ACD ABC
D BD
B AD
BD ABD
ABD AC
AB BC
A AB

0,0 0,1 0,2 0,3 0,4 0,0 0,5 1,0 1,5 2,0 2,5
Efects Efect
PSE de Lenth = 0,20625 PSE de Lenth = 1,25625

Figure 5 Interaction graph

In the case of Cu recovery, there are four significant effects (α = 0.005). There is one main effect
(component C) while the other three are combined effects. In the case of Fe recovery, three main
components are identified as significant effects (C, D and B). Finally, for the mass recovery, only the
component C is significant, see fig. 6. With these preliminary results, it is possible to evaluate the
interactions between tested components quantitatively. For this purpose, interaction graphs (of two
components) are used, and interpreted as follows:
 If the lines are parallel, there is no interaction.
 The less parallel the lines look like, the interaction force is stronger.
It is important to point that these graphs reveal trends that will help direct the work towards the
variables that generate positive or negative responses to the requested results.

7
 Figure 6 Main interaction graph

The graphs show that, for Cu recovery, negative effects are observed when combining P3 with either
P1 or P2. In absence of P3, the natural interaction between P1 and P2 is positive. P4 delivers a better
response in Cu recovery, nearly in all the interactions.
In the case of Fe recovery, the results are similar with almost all the elements, except for the
interactions of P2 with P3 or P4, where the response in Fe recovery decreased. This could imply that
component P2 adds in some way selectivity to the product.
When looking to the mass recovery, the observed response is low when using P2, P3 and P4 in any
combination between them, which delivers interesting properties to the product, as the mass pull
will not excessively increase. Therefore, the product performance during a plant trial will not be
affected. This effect is also relevant since at plant scale, it is always desired to maximize the
treatments.

RESULTS AND CONCLUSION

Once the components that delivered good results, both individually and combined were successfully
identified, the candidate products aiming to replace xanthate were formulated at CMS Application
and Development Laboratory, table 2, taking into consideration the specifications defined during
workshops with customers. The following products were obtained:

8
Table 2 products that aimed the original specification based on the customer needs

Product Xanthate HOSTAFLOT® HOSTAFLOT®

Specification 7257 (2419*) 7222

Physical Pellet / Powder Liquid Liquid

appearance

Water solubility Yes Yes Yes

Flash point -30 °C NA NA

Decomposition 25 °C > 100 °C > 100 °C

temperature

Target price STD Comparable Comparable

(*): commercial name up to date

HOSTAFLOT® 7257 (2419)

SIPX replacement as primary collector SIPX replacement as secondary collector

SIBX replacement as secondary collector

9
CONCLUSION
Using HOSTAFLOT® 7257 or 7222 as xanthate replacement delivers the following advantages:
Better, or at least equal a) 1% increase in Cu recovery achieved when HOSTAFLOT 7257 was
metallurgical performance used replacing SIPX as secondary collector.

b) Recoveries at lower pH improved significantly

Safety, health and a) It is not necessary to periodically clean sumps or tank bottoms due
environmental to accumulation of solids.
improvements
b) It is not required to elminate contaminated packaging or filters

c) No need to store solids who have risk of spontaneous combustion

Handling and storage a) Exposure to gases generated by xanthate decomposition (CS 2 &
improvements COS) is supressed

b) No need to prepare a liquid solution of xanthate

c) Excess of reagent volume due to decomposition and aging during

storage is eliminated

REFERENCES
Alexander Sutulov, AS. (1963) Flotación de minerales, Primera edición, Instituo de Investigación
Tecnológica, Concepción.
Nelson Martinez, NM. (2004) Flotación de minerales, Primera edición, Universidad Arturo Prat,
Iquique.
Deepak Malhotra, DM. (1963) Reagents in the mining industry, commodities or specialty chemicals?,
Primera edición, Resource Development Inc., Wheat Ridge, CO.
Fuerstenaud, D. W. and Herrera-Urbina. (1989) Recents Advances in Coal/Mineral Flotation, Primera
edición, Chander and R. R. Klimpel, SME, Littleton, CO.

10
Humic Substances for the Flotation of Sulphide
Minerals: Chronology and Recent Developments
César Sáez-Navarrete 1 and Rosanna Ginocchio2,3
1. Chemical Engineering and Bioprocess Department-Engineering Faculty, Pontificia Universidad
Católica de Chile
2. Center of Applied Ecology and Sustainability (CAPES), Pontificia Universidad Católica de
Chile
3. Departamento de Ecosistemas y Medio Ambiente, Facultad de Agronomía e Ingeniería Forestal,
Pontificia Universidad Católica de Chile

ABSTRACT
Chemical reagents used in froth flotation of sulphide minerals tend to have characteristics and/or
properties that make them dangerous to both the staff that handles them, and the environment,
where they often persist because of their physico-chemical properties and xenobiotic character.
For almost a decade now, our group has worked on the research and development of naturally
occuring collector and foaming reagents, aimed to replace the traditional ones. The center of
interest has been those compounds derived from organic waste and/or their degradation products.
These renewable, natural reagents have shown rougher flotation recoveries comparable to those
obtained with conventional reagents. Naturally, there are additional benefits, associated with
assisted phytostabilization of tailings and lower environmental risks.
The first promising study involved the use of biosolids (treated sludge from domestic wastewater)
for flotation, but the numbers didn´t fit after evaluating the required dosage and the travelling
distance to the concentration operations.
Further research was aimed at the identification of the active agents responsible for the collecting
capacity for copper and molybdenum sulphides. In recent years, we identified humic substances as
the active species responsible for the observed effects. Further work has shown that not all humic
acids have the same selectivity as collector reagents. We have developed a specific extract that use
these substances in doses comparable with traditional processes, improving its cost-effectiveness
and maintainig the environmental benefits initially detected.
This work shows the chronology of this development and presents the latest developments in the
obtaintion of new flotation reagents that will become an important contribution to a more
sustainable flotation operation in copper mining.

1
INTRODUCTION
Alternatives for the recovery and recycling of massively produced organic residues (i.e. biosolids,
pig slurries and sludges) in Chile has being one of the main tasks of our research group in the last
decade. Several research lines were first developed at the Phytotoxicity and Phytorremediation Unit
of the Chilean Research Center for Mining and Metallurgy (CIMM), all of them related to the
mining industry (i.e. assisted phytostabilization of chemically degraded soils and post-operative
tailings storage facilities). One of these research lines was focused on the use of biosolids (sewage
sludge generated from municipal water treatment plants) as reagents (frothers and/or collectors) for
the recovery of mineralogical species through froth flotation of sulfide copper ores. Here we focus
on the chronology and recent developments about the use of biosolids and humic substances for the
flotation of sulphide minerals.

METHODOLOGY
The timeline associated with the development of this research line is summarized in Figure 1. The
most important milestones in this chronology will be detailed in the following sectios.

Figure 1 Timeline associated with the research development of biosolids as collector and frother agent for the
recovery of mineralogical species through froth flotation of sulfide copper ores

RESULTS AND DISCUSSION

Preliminary technical results showed to be very promising (Reyes-Bozo et al., 2011a, 2011b, 2011c);
however, the economic feasibility assessment of the model for central Chile (best possible scenario
both in terms of biosolids availability and presence of large scale copper mine operations) was
negative. This was mainly due to inadequate availability of biosolids for the large volume of ores
being daily processed by large-scale mine operations and the high transportation costs of biosolids
from municipal water treatment plants to copper mine operations. A different approach was
needed to solve this bottle neck. Further research performed in the Laboratory of Renewable
Energy and Wastes at Pontificia Universidad Católica de Chile showed that only certain humic

2
substances, contained in organic wastes and natural organic materials, could be used as collectors
and frothers in froth flotation processes instead of bulk organic wastes (Ginocchio et al., 2015a,
2015b). All patent applications generated (Chile, Canada, South Africa, and Peru) considered,
therefore, the use of either bulk organic substances or humic substances as collectors and frothers.
Subsequent comparative studies performed in the laboratory of a flotation plant in a copper mine of
central Chile showed very similar recovery efficiencies and copper concentrate grades for froth
flotation with humic substances and froth flotation with regular collectors (approximate rate of 40
g/ton).

Humic Substances
Humic substances (HS) are stable (resistant) organic compounds resulting from degradation of
organic matter. Soils, peat, Leonardite, compost, biosolids, and other organic substances derived
from aerobic or anaerobic biodegradation processes have variable amounts of these substances.
They have been frequently used as organic amendments and fertilizers in croplands due to their
high content of macronutrients (i.e. Nitrogen), organic matter, and several biocompounds that
improves physical (i.e. texture, water holding capacity) and chemical soil properties (i.e. cation
exchange capacity, buffering capacity).
HS are divided into humic acids, fulvic acids, and humins, according to their different pH
solubilities (Figure 2).

Figure 2 Humic substances classified according Stevenson (1982)

HS are not commonly defined in terms of their functional groups or chemical composition. This
lack of scientific knowledge is very relevant when trying to understand the mechanisms behind
their use as collectors and frothers in froth flotation processes of sulphide minerals, particularly
when only some HS can be effectively used for this industrial application. Content of aminoacids
peptides, polifenols, and lipids in different HS are shown in Figure 3.

3
 Figure 3 Humic acids (HA)-acadian derives from select species of marine plants Ascophyllum nodosum; HA-
alginite, it originates from the biomass of fossilized unicellular algae during several millions of years in
volcanic craters. HA-lignite, it is between coal and peat. HA-lignohumate it is made of lignin-containing raw
material (Pospíšilová, 2015)

The research group is currently working to elucidate the mechanistic factors that explain the
efficacy of HS as collectors and frothers of mineralogical species from sulfide copper ores. Indeed,
Carbon-13 Nuclear Magnetic Resonance (13C NMR) can be used for estimating carbon distribution
in carbonyl, carboxyl, aromatic, acetal, aliphatic and heteroaliphatic groups. Functional group
distributions, aminoacidic composition and elemental composition are also relevant to quantify. All
these information will allow us to define relevant mechanisms behind their efficacy as collectors
and frothers.

Recent developments
A series of froth flotation experiments were performed with collaboration of Dr. Jaime Álvarez of
the Department of Metallurgy at the Universidad de Concepción. While many previous studies
have been developed with this group, we only present here results of recent trials.
The main objective of these trials was to quantify the recovery and copper/iron/molybdenum
concentrate grades of HS when used as collector agent (named UC-2) in standard froth flotation
experiments. Additional trials were performed in order to study the effect of both conditioning time
and concentration of frother agents. Experiments were performed in an automatized EDEMET cell
considering a froth flotation time of 12 minutes (N=3). A batch of 35 kg of copper sulfide ore
obtained from a mine located in central Chile was conditioned according to standard protocols. The
experimental design of the laboratory experiments for rougher flotation trials, according to reagent
concentrations and conditioning times, is shown in Table 1.

4
Table 1 Experimental design of laboratory experiments for rougher flotation of copper sulfide ores.
UC-2, humic substances to test; D-101/NP-107, collector mix reagents; F-810/MIBC/F-1012, mix
foaming agents

Trial Main collector Frother Lime Time (s)

kg/ton
N° g/ton g/ton (conditioning)
(mill)
(mill) (conditioning)

1 38 D-101/NP-107 12 F-810/MIBC/F-1012 0.52 30

2 38 UC-2 24 F-810/MIBC/F-1012 0.53 30

3 38 UC-2 36 F-810/MIBC/F-1012 0.53 30

4 38 D-101/NP-107 12 F-810/MIBC/F-1012 0.52 60

5 38 UC-2 24 F-810/MIBC/F-1012 0.53 60

6 38 UC-2 36 F-810/MIBC/F-1012 0.53 60

Ore sample preparation

Downsizing of sulfide copper ore is the first step of sample preparation with sequential stages of
crushing, sorting, and homogenization, up to a product with 100% particle size below mesh10. This
product (-10#) is then homogenized with a roller, and a sorting method allows a head sample. The
rest of the sample is separated in identical 1-kg weight aliquots. Finally, the head sample is finely
grinded in a ring grinder up to a product with 100% particle size below mesh 200. This sample is
chemically analyzed for the head grade. The water content and the apparent density of the head ore
sample are also determined. Table 2 shows the main characteristics of the head ore sample used for
the trials.

Table 2 Main characteristics of the head ore sample used for the trials al

Head ore grade

Cu Mo Fe
% % %
0.73 0.015 3.38
Ore characteristics
Humidity (%) Density (g/mL)
0.22 2.91

5
Rougher froth flotation
The ore is ground up to 65% of solids (weight) with a laboratory Marcy mill. During the grinding
process, the ore is conditioned with the different reagents to test, according the rates defined in
Table 1. After the grinding and conditioning, the pulp is drained into a Denver flotation cell and
water is added up to the standard level. Frothers are then added following type and doses given in
Table 1.
After adjusting the operating parameters (conditioning and flotation time, impeller speed, air flow,
sampling interval, sampling rotation speed, pH value, and level control) on the control panel of the
rougher froth flotation cell (Table 3), these are automatically controlled throughout the trial. The
pulp conditioning is followed by the air flow and bubble generation into the cell, thus starting the
froth flotation process.

Table 3 Operating parameters of rougher froth flotation trials

Rougher flotation
Parameters Conditions
Air flow [L/min] 10
Impeller speed [rpm] 900
Sampling frequency [seg] 5
pH 9.5
Volumen of cell [L] 2.7

Concentrate and tailings obtained after each froth flotation processes were filtered and dried in an
oven. A representative sample of both materials is taken and delivered to chemical characterization;
a sub-sample of them is ground below mesh 200,
Table 4 shows metallurgic recoveries obtained for value minerals (copper, molybdenum and iron)
after the treatment of sulfide copper ores with selected tests of Table 1.

6
Table 4 Metallurgic recoveries of value minerals (copper, molybdenum and iron) after the
treatment of sulfide copper ores with selected tests detailed in Table 1

Trial Main collector Frother agent Concentrate grade Recovery (%)

N° (mill) (conditioning)

Cu Mo Fe Cu Mo Fe

(%) (ppm) (%) (%) (%) (%)

1 38 g/ton D-101/NP- 12 g/ton F-810/MIBC/F- 7.3 1,200 13 86 76 24

107 1012

2 38 g/ton UC-2 24 g/ton F-810/MIBC/F- 7.4 1,450 12 77 66 33

1012

3 38 g/ton UC-2 36 g/ton F-810/MIBC/F- 7.7 1,300 13 76 56 24

1012

4 38 g/ton D-101/NP- 12 g/ton F-810/MIBC/F- 7.5 1,100 13 86 54 28

107 1012

5 38 g/ton UC-2 24 g/ton F-810/MIBC/F- 7.6 1,100 12 83 76 27

1012

6 38 g/ton UC-2 36 g/ton F-810/MIBC/F- 7.6 1,250 13 84 53 26

1012

Table 5 Standard deviations concentrate grade and recovery

Trial St.Dv. Concentrate grade St. Dv. Recovery (%)

N° Cu (%) Mo Fe (%) Cu Mo Fe
(ppm)
1 0,25 --- 0,35 0,69 0,05 0,00

2 1,29 0,010 0,64 2,74 2,17 4,69

3 0,52 --- 0,07 1,96 0,13 0,00

4 0,37 --- 0,28 0,74 0,06 4,69

5 0,18 --- 0,14 0,35 0,94 8,77

6 0,28 0,010 0,21 1,20 0,96 4,50

7
Highest copper recoveries are obtained under trails 1 and 4, corresponding to standard mine trials,
thus showing that variation of conditioning time has no effect for copper recovery under standard
mine trials (trials 1 and 4 has conditioning times of 30 s and 60 s, respectively).
The change of the standard mine-used collectors by UC-2 (the HS agent; trial 2) results in a copper
recovery 9 points below to the standard trial. However, when the frother concentration is increased
(trial 3) or the conditioning time is increased (60 s; trial 5), copper recovery increases to either a
value similar of the standard test or 5 point above the standard value, respectively. If both
conditioning time and frother concentration are increased (trial 6), there is no further improvement
in copper recovery when compared to the standard trial 4.

CONCLUSIONS
We can conclude that the use of HS as collector agents is feasible in froth flotation processes of
copper and molybdenum (trial 5). However, the most suitable froth flotation conditions for this
new reagent need to be defined yet. The research team is currently planning some pilot scale trials
in an important copper mine of central Chile in order to scale up the use of this new collector agent.

REFERENCES
Ginocchio, R. (developer), Sáez-Navarrete, C. (developer), Herrera, M., (collaborator), Reyes-Bozo,
L. (collaborator). (2015). Southafrican Patent 2012/05536. Collector and frothing agent for flotation
based on organic residues to recover metals from frothing flotation minerals, collector and frothing
agent recovery process and frothing flotation process that uses the collector and frothing agent.
Ginocchio, R. (developer), Sáez-Navarrete, C. (developer), Herrera, M., (collaborator), Reyes-Bozo,
L. (collaborator). (2015). Canadian Patent 2,787,724 (21/01/2011). Collector and frothing agent for
flotation based on organic residues to recover metals from frothing flotation minerals, collector and
frothing agent recovery process and frothing flotation process that uses the collector and frothing
agent.
Lubica Pospíšilováa, Markéta Komínková, Ondřej Zítkab, René Kizekb, Gabriela Barančíková,
Tadeáš Litavec, Tomáš Lošák, Jaroslav Hlušek, Anna Martensson and Tibor Liptaj. (2015) Fate of
humic acids isolated from natural humic substances. Acta Agriculturae Scandinavica, 65, 6, 517 -
528.
Reyes-Bozo L., Herrera-Urbina R., Godoy-Faúndez A., Sáez-Navarrete C., Herrera M., Ginocchio R.
(2011) Use of biosolids as froth reagent in flotation process: Chemical and physical characterization.
Dyna, 78, No. 167, pp. 7-16.
Reyes-Bozo L., Herrera-Urbina R., Escudey M., Godoy-Faúndez A., Sáez-Navarrete C., Herrera M.
and Ginocchio R. (2011) Role of biosolids on hydrophobic properties of sulfide ores. International
Journal of Mineral Processing, 100, 3-4, 124-129.
Reyes-Bozo L., Herrera-Urbina R., Sáez-Navarrete C., Otero A., Godoy-Faúndez A. and Ginocchio
R. (2011) Rougher flotation of copper sulphide ore using biosolids and humic acids. Minerals
Engineering, 24, 14, 1603 - 1608.
Reyes-Bozo L., Higueras P., Godoy-Faúndez A., Sobarzo F., Sáez-Navarrete C., Vásquez-Bestagno
J., Herrera-Urbina R. (2014) Assessment of the flotability of chacopyrite, molybdenite and pyrite

8
using biosolids and their main components as collectors for greening the flotation of copper
sulphide ores. Minerals Engineering, 64, 38 - 43.
Stevenson FJ. 1982. Humus chemistry – genesis, composition, reactions. New York (NY): Wiley
Inter-science; 445 p.

9
Interactions between Kaolinite and Molybdenite in
Fresh and Seawater
Gastón Avilés, Andrés Ramírez and Leopoldo Gutiérrez
Universidad de Concepción, Chile

ABSTRACT
Scarcity of fresh water, particularly for operations situated in desert areas, has forced the copper-
molybdenum mining companies to look for new sources of water such as seawater. The use of this
saline aqueous medium in the process of flotation leads to some important effects on the efficiency
of this process. Besides, copper grades of large copper deposits have constantly decreased during the
last decades, which is typically associated to high concentrations of clay minerals in the ores which
is known to negatively affect the flotation process. The objective of this research was to evaluate the
effect of kaolinite on the floatability of molybdenite at different pH conditions, using two types of
solutions (0.01 M NaCl solution and seawater). Floatability of molybdenite particles was assessed
through micro-flotation tests in a 150 cm3 Partridge-Smith glass cell, and using nitrogen (N2) at a
flowrate of 20 cm3/min, for a total flotation time of 10 minutes. Flotation feed was prepared mixing
known amounts of molybdenite with particles of kaolinite. The results of this study indicate that
molybdenite floatability slightly decreased under the presence of fine particles of kaolinite when
conventional water was used in the tests. However, in seawater the negative effect of kaolinite was
more relevant, in particular at pH>9, which suggests that the presence of cations in solution intensify
the phenomenon of heterocoagulation between fine kaolinite particles and molybdenite (slime
coating). It is assumed that the stronger depressing effect of kaolinite at pH>9 could be related to the
formation of hydroxy-complexes of magnesium and calcium.

1
INTRODUCTION
Molybdenite (MoS2) is the primary source of molybdenum. From a structural point of view,
molybdenite particles are formed by crystal layers of S-Mo-S which are held together by the action of
strong covalent bonds. The interaction between each of these S-Mo-S crystals occurs through the S-S
planes (Figure 1), and the process of breakage of the molybdenite particles occurs along these planes.
As a result, a non-polar and highly hydrophobic surface which is usually referred to as face, is
created. Consequently, molybdenite particles are usually described as anisotropic entities with
hydrophobic faces and hydrophilic edges (Chander and Fuerstenau, 1972). It is known that the zeta
potential of molybdenite particles is the lowest at pH values ranging between 5.5 and 6.5, which is in
agreement with the range of maximum floatability (Chander and Fuerstenau, 1972). Regarding the
interaction between molybdenite and gangue particles, previous studies showed that the presence of
Ca2+ ions in solution induces heterocoagulation between silica and molybdenite which explains the
low molybdenum recoveries observed under these conditions (Tabares et al., 2006; Raghavan and
Hsu, 1984). The negative effect of silica on the floatability of molybdenite was assumed to be related
to a phenomenon of slime coatings (Raghavan and Hsu, 1984).

Figure 1 Crystalline structure of molybdenite (Castro et al., 2016)

Molybdenite is commonly associate with copper. The grades of copper-molybdenum deposits in the
world have declined, which is typically associated with high concentrations of phyllosilicates, with
the clay minerals being the most abundant group. These types of mineral species are usually
dispersed in mineral slurries as fine particles, which generates the conditions for non-Newtonian
behavior (Bakker et al., 2009). These characteristics affect the entire mineral processing chain, i.e.,
milling, froth flotation, thickening, and dewatering (Ndlovu et al., 2014; Ndlovu et al., 2013; Ralston
et al. 2007; He et al., 2004; Liu et al., 2002; McFarlane et al., 2005; Melipichun and Gutierrez, 2014).
The efficiency of the froth flotation process can be severely affected by the presence of high contents
of phyllosilicates, and in many cases this problem can be dealt with successfully by diluting the
slurry, blending different ore types, and sometimes even disposing of the problematic ores which
inevitably leads to losses of valuable minerals (Connelly, 2011). In particular, clay minerals are the
most common phyllosilicates found in copper ores and kaolinite is one of the most abundant species
(Kelm et al., 2013). From the surface chemistry point of view, kaolinite has two different surfaces
generated during particle breakage, i.e., the basal silica-like and the alumina-like faces (Laskowski,

2
2010; Laskowski, 2012). The silica-like faces carry a negative charge while the alumina-like displays
a pH dependent surface charge resulting from protonation/deprotonation of exposed hydroxyl
groups with IEP values around 7.0 (Johnson et al., 1998). Some studies also suggested that the silica-
like faces display hydrophobicity and that dispersion of kaolinite particles in water is explained by
the presence of surface imperfections on the faces which create polar micro-edges (Sadda et al., 1995;
Miller et al., 2007). The anisotropic behavior of clay minerals promotes the attachment of these fine
particles to minerals (slime coating), and their interactions with reagent and bubbles. All these
phenomena reduce the copper and molybdenum floatability and hence recovery (Ralston et al., 1999;
Connelly, 2011; Forbes et al., 2014; Ndlovu et al., 2014), and cause major changes in the rheological
behavior of mineral slurries (Uribe et al., 2016).
Nowadays, water is a critical and scarce resource for the mining industry and the use of seawater to
process copper-molybdenum ores by flotation seems to be a sustainable solution in zones with
restricted resources of fresh water such as northern Chile (Castro et al., 2012). Seawater contains
significant concentrations of different ions (Na+, K+, Mg2+, Ca2+, SO42-, Cl-) which can potentially adsorb
on the solid surfaces, and induce heterocoagulation between molybdenite and clay minerals. Earlier
publications from the CRHIAM Center of the University of Concepcion, on flotation of Cu-Mo sulfide
ores in seawater (Castro et al., 2012; Castro et al., 2014) indicated that molybdenite flotation is strongly
affected in seawater when pH is raised with lime to depress pyrite, which is related to the depressing
effect of Mg2+ and Ca2+ hydrolysis products on the hydrophobicity of molybdenite. These studies
showed that molybdenite floatability can be severely depressed by the action of the products of
magnesium hydrolysis, in particular Mg(OH)2, and that in the case of Ca2+ the depressing action can
be related to the action of Ca(OH)+. Recently Uribe et al. (2016) showed that the depressing effect of
kaolinite on chalcopyrite may be related to the formation of hydrolysed species of calcium and
magnesium which can induce heterocoagulation between kaolinite and chalcopyrite (slime coating).
Liu et al. (2002, 2004) also showed that calcium enhanced heterocoagulation between montmorillonite
and bitumen can lead to to low bitumen recoveries.
The objective of this investigation was to evaluate the depressing effect of kaolinite on the floatability
of molybdenite using two types of solutions, i.e., a 0.01 M NaCl solution, seawater, and solutions
containing Ca2+, Mg2+, Na+, and K+ at the concentrations at which they appear in seawater, over the
pH range from 8 to 11.

METHODOLOGY

Samples and reagents

The kaolinite sample was provided by the Clay Minerals Society Source clays. XRD analyses (Bruker
D4-Endeavor operated with Ni-filtered Cu radiation) indicated that the sample had 96% kaolinite,
3% anatase, and 1% of other traces. Cation exchange capacity (CEC) and Brunauer–Emmett–Teller
(BET) surface area were 0.7 cmol/kg and 180 m2/g respectively. The molybdenite sample used in this
work was obtained from an industrial molybdenum concentrate, and further purified through 3
stages of flotation without collector. Samples of two particle size distributions were used in this
research, i.e., coarse and fine samples. The coarse sample with a P80 of 100 microns corresponded to
the original molybdenite samples obtained from the process of purification, while the fine sample
with a P80 of 50 microns was obtained by hand grinding the coarse sample. Figure 2 shows the
particle size distributions. The samples were then washed and cleaned following the procedure
described by Uribe et al. (2016) which considers 3 stages of washing with sodium hydrosulfide

3
(NaSH) and acetone in order to remove any organic reagents such as collectors present in the sample.
Table 1 shows the chemical analysis of the samples. In order to keep molybdenite oxidation at
minimum, the samples were stored in a freezer in sealed plastic bags previously blown with nitrogen,
at a temperature of about 1 °C (Ekmekci and Demirel, 1997; Zhang et al., 1997; Ansari and Pawlik,
2007).

100
Passing cumulative, %

10

1
1 10 100 1000
Size, microns

Fine molybdenite Coarse molybdenite

Figure 2. Particle size distribution of molybdenite samples

Table 1 Chemical analyses of molybdenite samples

Chemical analyses in XRD

Sample Cu(%) Fe(%) S(%) Mo(%) Ins(%)

Coarse 0,55 0,49 37,9 55,4 5,3

Fine 0,53 0,39 38,0 55,8 5,1

Methyl Isobutyl Carbinol (MIBC) of analytical grade was used as frother at 15 ppm, and potassium
amyl xanthate (PAX) as collector at 25 ppm for the micro-flotation tests. It has to be noted that PAX
was purified by dissolution in acetone, and precipitated with ether. Analytical grade NaCl, KCl,
MgCl2 and CaCl2 (Merck) were used to prepare the solutions. Distilled water with an electrical
conductivity of 21.5 µS/cm was used in all the experiments. Seawater samples were taken from the
Lenga Coast in the VIII Region of Bio Bio in Chile. The pH was adjusted with sodium hydroxide and
hydrochloric acid (Merck).

4
Micro-flotation
The floatability of molybdenite was evaluated through micro-flotation tests that were carried out in
a 150 mL Patridge and Smith glass cell using nitrogen (N2) at a flowrate of 80 mL/min (Figure 3). The
flotation feed was prepared by mixing 1 g of molybdenite (coarse and fine) with 0.15 g of kaolinite to
then prepare a suspension using 100 mL of solution (0.01 M NaCl or seawater). After 3 minutes of
conditioning at the required pH, MIBC and PAX were added at a concentration of 15 ppm and 25
ppm respectively (based on the 150 mL of the micro-flotation cell), and conditioned for 5 more
minutes at the previously established pH value. Additional solution was then added to complete the
150 mL required for the micro-flotation glass cell. Finally, the gas valve was opened to start the
process of flotation which was carried out for 3 minutes, pulling froth every 10 seconds. Pulp level
in the micro-flotation cell was kept constant by adding a background solution prepared at the same
composition and pH of the original solution. Finally, in order to separate molybdenite and kaolinite
particles, the concentrates and tailings were sieved using a 400# screen (38 microns) and dried in an
oven at 105°C for 5 hours. This way it was possible to measure the mass of molybdenite in the
concentrate, and the solids flotation yield was calculated as the ratio between the mass of
molybdenite in the concentrate and the mass of molybdenite in the concentrate tailings. All testing
was done in duplicate.

Induction time measurements

Induction time measurements were conducted using the procedure described by Castro et al. (2014)
and the induction time meter built at the University of Alberta. According to this procedure, a fresh
bubble held on a capillary is pushed downward through a solution so that it is kept in contact with a
bed of particles for a measured and controlled contact time span. Then, the bubble is lifted upwards
to its initial position and observed through a microscope in order to determine whether attachment
occurred during the contact period or not. The process is repeated 10 times at different positions on
the particle bed and the number/percentage of successful contacts is determined. Then, the contact
time is changed to different values and the procedure is repeated again in each case. The final result
of the test is a plot of percentages of successful contacts (Nsc) versus the measured contact time (tc).
In this work, the induction time was obtained by determining the contact time at which 50 % of the
contacts were successful. The experimental data were fitted using Equation 1.
The beds of particles were prepared by mixing 3 grams of molybdenite (for size) with 30 mL of
solution containing 40 ppm of PAX. Immediately after, fine particles of kaolinite were added at a
concentration of 13 % (solid base). Then, the pH was adjusted to the required value and the slurry
was conditioned for 12 minutes, after which it was transferred to a 15 mL cell in order to start the
induction time measurements. It has to be pointed out that no frother was added in these experiments
and that air was used to generate the bubbles. All these experiments were conducted in duplicate
thus the results presented in this paper are average values. It has to be noted that these types of
measurements were previously used to evaluate the effect of slime coating on the floatability of
molybdenite in the presence of clay minerals (Oats et al., 2010) and different ions.

5
RESULTS AND DISCUSSION

Micro-flotation
Figure 3, shows the recovery of coarse molybdenite in 0,01 M NaCl solution and in seawater, in the
presence and absence of kaolinite. The results indicate that the effect of kaolinite on the floatability
of molybdenite at the whole range of pH is minor when the process of flotation is carried out using
a 0.01 M NaCl solution. In contrast, in seawater the effect of kaolinite on molybdenite recovery seems
to be stronger, in particular at pH > 9. In other words, the depressing effect of kaolinite seems to be
important only in an aqueous media of high ionic strength such as seawater.
The theory of stability of colloidal systems developed by Derjaguin-Landau-Verwey-Overbeek
(DLVO) can be used to better understand the interactions between kaolinite and molybdenite. The
DLVO theory considers the existence of two forces that affect the process of aggregation of colloidal
particles, i.e., van der Waals forces of attraction and repulsive forces generated by the interactions of
the electrical double layers. Thus, to explain the results presented in Figure 3, it is necessary to
analyze the effect of pH on the surface charge of both minerals. The surfaces of the faces and edges
of molybdenite particles are expected to be negatively charged at pH values between 7 and 11,
(Chander and Fuerstenau, 1972). Besides, previous studies reported that the surface charge of the
alumina-like faces of kaolinite becomes close to zero at pH values around 7-8 (Laskowski, 2010;
Laskowski, 2012). Then, according to this information heterocoagulation between kaolinite and
molybdenite should be more likely at lower pH, however the data presented in Figure 3 indicate that
there is not a clear correlation between pH and the depressing effect of kaolinite on the floatability of
molybdenite which suggests no heterocoagulation.
When looking at the results for seawater in Figure 3, it can be seen that the effect of kaolinite is more
significant in seawater than in fresh water at the whole range of pH, but at pH >9 the depressing
effect of kaolinite becomes more important. The fact that the pH value at which the trend changes is
9, suggests that the stronger depressing effect of kaolinite observed at this pH range is related to the
formation of hydroxy-complexes of magnesium and calcium and their influence on the interactions
between kaolinite and molybdenite. Previous studies showed that the first hydroxy-complex of
magnesium Mg(OH)+ starts forming at pH 9, reaching a maximum concentration at pH 10.5. At this
pH, positively charged precipitates of Mg(OH)2(s) can also form which can affect the properties of the
air/liquid and solid/liquid interfaces. In the case of calcium, the first hydroxy-complex Ca(OH)+ forms
around pH 9.5-10 with the maximum concentration obtained at pH values around 13-13.5. The
presence of these hydroxy-complexes seems to enhance the interactions between clay particles and
molybdenite, phenomenon referred to as “slime coating”.

6
 100

90

80

70
Recovery, %
60

50

40

30
Coarse sample
20

10

0
6 7 8 9 10 11 12
pH

0,01 M NaCl 0,01 M NaCl, with kaolinite Seawater Seawater, with kaolinite

Figure 3 Recovery of coarse molybdenite using conventional and seawater with and without kaolinite

The induction time is the time required to break the liquid film existing between a bubble and a
particle after the two collide. Therefore high induction time values indicate that the particle-bubble
adhesion is favorable and therefore, better molybdenite recoveries should be expected. Figure 4
shows the induction time data obtained using coarse molybdenite in conventional and seawater, and
with and without kaolinite. The results of induction time are in good agreement with the micro-
flotation data. Firstly, it can be seen that in general the induction time values are higher in the
presence of kaolinite in both fresh and seawater. Additionally, the depressing effect of kaolinite is
more important in seawater, in particular at pH> 9, which is again in perfect agreement with the data
of micro-flotation. Therefore, the induction time results validate the conclusions obtained in the
micro-flotation testing.
Figure 5 shows the recovery of fine molybdenite using conventional and seawater, with and without
kaolinite. The first observation that can be made is that in general the recovery values obtained in a
0.01 NaCl solution using the fine molybdenite sample tend to be lower than those obtained with the
coarse sample. This is also observed in the results obtained using seawater at pH 7-9. This can be
related to the fact that the face/edge ratio decreases with particle size, which has an effect on
hydrophobicity leading to low values of recoveries. However, when comparing the recovery of
molybdenite at pH>9, it can be seen that the depressing effect of kaolinite is more important on coarse
particles of molybdenite. As was previously explained, formation of hydroxy-complexes of
magnesium and calcium at pH>9 seems to enhance heterocoagulation between kaolinite and
molybdenite and the results presented in Figures 3 and 5 indicate that this phenomenon is more
relevant on coarse particles. The presence of hydroxy-complexes of magnesium and calcium can
induce attachment of kaolinite to the faces and edges of molybdenite through a mechanism of surface
charge neutralization. As the face/edge ratio is higher for coarse particles, heterocoagulation between
kaolinite and the molybdenite faces is more likely in this case, and that could explained why kaolinite
affects more the recovery of coarse molybdenite at pH>9.

7
 10000,0

Induction time, ms
1000,0

100,0

10,0
6 7 8 9 10
pH

0,01 M NaCl 0,01 M NaCl, with kaolinite Seawater Seawater, with kaolinite

Figure 4 Induction time of fine molybdenite using conventional and seawater with and without kaolinite

100

90

80

70
Recovery, %

60

50

40

30
Fine sample
20

10

0
6 7 8 9 10 11 12
pH

0,01 M NaCl 0,01 M NaCl, with kaolinite Seawater Seawater, with kaolinite

Figure 5. Recovery of fine molybdenite using conventional and seawater with and without kaolinite

8
CONCLUSIONS
The depressing effect of kaolinite on the floatability of molybdenite at pH values between 7 and 11 is
less important when the process of flotation is carried out using a 0.01 M NaCl. In seawater the
depressing effect of kaolinite seems to be stronger, in particular at pH > 9. The fact that the pH value
at which the trend changes is 9, suggests that the stronger depressing effect of kaolinite observed at
this pH range can be related to the formation of hydroxy-complexes of magnesium and calcium and
their influence on the interactions between kaolinite and molybdenite. Hydroxy-complexes of
magnesium and calcium can induce attachment of kaolinite to the faces and edges of molybdenite
through a mechanism of surface charge neutralization. As the face/edge ratio increases with particle
size, heterocoagulation between kaolinite and molybdenite faces is more likely for coarse sizes, and
that could explained why kaolinite affects more the recovery of coarse molybdenite at pH>9.
These conclusions were validated through induction time measurements.

ACKNOWLEDGEMENTS
The support from the National Fund for Scientific and Technological Development (FONDECYT) of
Chile under the projects initiation into research no. 11140184, and Centro CRHIAM Proyecto Conicyt
Fondap no.15130015, is gratefully acknowledged.

REFERENCES
Ansari, A. and Pawlik M. (2007) ‘Floatability of chalcopyrite and molybdenite in the presence of
lignosulfonates. Part I. Adsorption studies’, Minerals Engineering, Volume 20, Issue 6, 600-608.
Bakker, C. W., Meyer, C. J. and Deglon, D. A. (2009) ‘Numerical modelling of non-Newtonian slurry
in a mechanical flotation cell´, Minerals Engineering, Volume 22, Issue 11, 944–950.
Castro, S., Rioseco, P. and Laskowski, J.S. (2012) ‘Depression of molybdenite in sea water’, XXVI
International Mineral Processing Congress-IMPC 2012, Proceedings, New Delhi, India, September 24-
28, 737-752
Castro, S., Uribe, L. and Laskowski, J.S. (2014) ‘Depression of inherently hydrophobic minerals by
hydrolysable metal cations: molybdenite depression in seawater’, XXVII International Mineral
Processing Congress-IMPC 2014, Flotation Chemistry Chapter, Santiago, Chile, October 20-24, 207-217.
Castro, S., Lopez-Valdivieso, A. and Laskowski, J.S. (2016) ‘Review of the flotation of molybdenite.
Part I: Surface properties and floatability’, International Journal of Mineral Processing, Volume 148, 48–
58.
Chander, S., Fuerstenau, D.W., (1972) ‘On the natural floatability of molybdenite’, Transactions of
AIME, Volume 252, 62–69.
Connelly, D. (2011) ‘High clay ores – a mineral processing nightmare Part 2. In: Australian’ Bulk
Handling Review: September/October edition, 78–81.
Ekmekci, Z. and Demirel, H. (1997) ‘Effects of galvanic interaction on collectorless flotation behaviour
of chalcopyrite and pyrite’, International Journal of Mineral Processing, Volume 52, Issue 1, 31-48.

9
Forbes, E., Davey, K. J. and Smith, L. (2014) ‘Decoupling rehology and slime coatings effect on the
natural flotability of chalcopyrite in a clay-rich flotation pulp’, Minerals Engineering, Volume 56,136–
144.
He, M., Wang, Y. and Forssberg, E. (2004) ‘Slurry rheology in wet ultrafine grinding of industrial
minerals: a review’, Powder Technol, Volume 147, Issues 1–3, 94–112.
Johnson, S.B., Russell, A.S. and Scales, P.J. (1998) ‘Volume Fraction Effects in Shear Rheology and
Electroacoustic Studies of Concentrated Alumina and Kaolin Suspensions’, Colloids and Surfaces A:
Physicochem and Engineering Aspects, Volume 141, Issue 1, 119-130.
Kelm, U., Helle, S., Jerez, O. and Pincheira, M. (2013) ‘What are copper clays? Geometallurgical
implications’, Copper international Conference (COPPER 2013), Santiago, Chile, December 1-4, 257-266.
Laskowski, J.S. (2010) ‘Rheological properties of aqueous suspensions of anisotropic minerals In:
Rheology and Processing of Fine Particles’, Proceedings of the 8thUBC-McGill-UA International
Symposium on Fundamentals of Mineral Processing, Met. Soc., Vancouver, BC, Canada, August 3-6, 137-
157.
Laskowski, J. S. (2012) ‘Anisotropic Minerals in Flotation Circuits’, Canadian mineral processing CIM
Journal, Volume 3, issue 4, 203-213.
Liu, J., Zhou, Z., Xu, Z. and Masliyah, J. (2002) ‘Bitumen-clay interactions in aqueous media studied
by zeta potential distribution measurement’, Journal of Colloid and Interface Science, Volume 252, Issue
2, 409–418.
Liu, J., Xu, Z. and Masliyah, J. (2004) ‘Role of fine clays in bitumen extraction from oil sands’, AIChE
Journal, Volume 50, Issue 8, 1917–1927.
McFarlane, A.J., Addai-Mensah, J. and Bremmell, K. (2005) ‘Microstructure, rheology and dewatering
behavior of smectite dispersions during ortho-kinetic flocculation’, Minerals Engineering, Volume 18,
Issue 12, 1173–1182
Melipichun, C. and Gutierrez, L. (2014) ‘Influence of clays on induction time, floatability and foam
stability in the process of flotation of copper sulphide minerals’, XXVII International Mineral Processing
Congress, 2014, Santiago, Chile , October 20-24, 1-9.
Miller, J.D., Nalaskowski, J., Abdul, B. and Du, H. (2007) ‘Surface Characteristics of Kaolinite and
Other Selected Two Layer Silicate Minerals’, The Canadian Journal of Chemical Engineering, Volume 85,
Issue 5, 617-624.
Ndlovu, B., Farrokhpay, S. and Bradshaw, D. (2013) ‘The effect of phyllosilicate minerals on mineral
processing industry’, International Journal of Mineral Processing, Volume 125, 149–156.
Ndlovu, B., Forbes, E., Farrokhpay, S., Becker, M., Bradshaw, D. and Deglon, D. (2014) ‘A preliminary
rheological classification of phyllosilicate group minerals’, Minerals Engineering, Volume 55, 190–200.
Oats, W. J., Ozdemir, O., and Nguyen A. V., (2010) ‘Effect of mechanical and chemical clay removals
by hydrocyclone and dispersants on coal flotation’, Minerals Engineering, Volume 23, Issue 5, 413–
419.
Raghavan, S. and Hsu, L.L. (1984) ‘Factors affecting the flotation recovery of molybdenite from
porphyry copper ores’, International Journal of Mineral Processing, Volume 12, Issues 1–3, 145–162.

10
Ralston, J., Fornasiero, D., Grano, S., Duan, J. and Akroyd, T. (2007) ‘Reducing uncertainty in mineral
flotation—flotation rate constant prediction for particles in an operating plant ore’, International
Journal of Mineral Processing. Volume 84, Issues 1–4, 89–98.
Ralston, J, Fornasiero, D. and Hayes, R. (1999) ‘Bubble–particle attachment and detachment in
flotation’, International Journal of Mineral Processing, Volume 56, Issues 1–4 133–164.
Saada, A., Siffert, B. and Papier, E. (1995) ‘Comparison of the Hydrophilicity/Hydrophobicity of Illites
and Kaolinites’, Journal of Colloid and Interface Science, Volume 174, Issue 1, 185-190.
Tabares, J., Madrid, I., Reyes, J., Sánchez, A., Valdez, D., López, A., 2006, ‘Surface properties and
floatability of molybdenite’, Proceedings of 2006 China-Mexico Workshop on Minerals Particle Technology,
San Luis Potosí, Mexico, 115-124.
Uribe, L., Gutierrez, L. and Jerez, O. (2016) ‘The depressing effect of clay minerals on the floatability
of chalcopyrite’, Mineral Processing and Extractive Metallurgy Review, Volume 37, Issue 4, 227-235.
Zhang, Q., Xu, Z., Bozkurt, V. and Finch, J., A. (1997) ‘Pyrite flotation in the presence of metal ions
and sphalerite’, International Journal of Mineral Processing, Volume 52, Issues 2–3, 187-201.

11
Use of Blending to Optimize the Flotation Process of
High Clays Ores
Jaime Álvarez1, Leopoldo Gutiérrez1*, Sonia Helle2, Ursula Kelm2, Marcos Pincheira2 and
Oscar Jerez2
1. Universidad de Concepción, Chile
2. Institute for Applied Economic Geology (GEA), Universidad de Concepción, Chile

ABSTRACT
The aging of major copper deposits in Chile has caused the continuous diminution of mineral
resources resulting in increased processing of low grades ores rich in problematic gangue species
such as clay minerals. The problems associated with the presence of clays during the process of
froth flotation can be attenuate by several ways. One of the most typical practices is to do is referred
as to “Blending”, which consists in mixing low with high clay ores so that the overall recovery of
valuables obtained by processing the ores mixture is better than the recovery that would be
obtained when the ores are individually treated.
Although, Blending is use in many operations in Chile, it is generally base on empiric results tests
but not in a systematic application of mineralogical tools and knowledge. Moreover, its efficiency
and the variables that define what ore type can or cannot be blended is something that is far from
being understood.
The objective of this work was to study the effect of Blending on the recovery of copper and
molybdenum from eight (8) different ores samples, and to obtain a correlation with the
mineralogical and textural data. To achieve this objective, flotation tests were carried out using feed
samples prepared by Blending low and high clay ores at different mass ratios (100:0, 80:20, etc).
Then, the effect of Blending on the flotation recovery was evaluated by comparing the calculated
overall flotation result obtained from the proportional contribution of each type of ore, with the real
values obtained from the flotation tests. The difference between both values indicated the extent of
the effect of Blending. Ore samples were characterized using image analysis and QEMSCAN.
The results indicate that Blending is very good option to improve the overall recovery of copper
and molybdenum, but it only works well when the ores samples blended display very different
mineralogical and textural properties. Another important conclusion is that Blending also has an
important effect on the grinding process that should be taken into account in the mineral processing
operations.

There is no full article associated with this abstract due to ongoing patenting processes.

1
Hematite Flotation Using a Crude Biosurfactant
Extracted from a Gram Positive Bacteria
Jhonatan Puelles, Antonio Merma, Carlos Castañeda and Maurício Torem
Pontificia Universidade Católica do Rio de Janeiro, Brazil

ABSTRACT
Bioflotation is defined as a separation process by which the mineral of interest is floated or
depressed selectively using reagents of biologic origin, also known as bioreagents. These substances
are characterized by their green chemistry, selectivity and potential to treat fine particles. In that
sense, the research’s principal objective is the assessment of the hematite floatability using a crude
biosurfactant extracted from a Gram positive bacteria and consequently determine its potential as
an alternative against synthetic reagents or the bacteria itself.
Characterization by FTIR spectrophotometry identified alcohol (-OH), ketone (C=O) groups, and
saturated and unsaturated carbon chains, which may compose the mycolates that form the cellular
wall of the bacteria. Surface tension measurements of the crude biosurfactant determined a CMC of
92 ppm lowering the water surface tension from 72 to 52.5 mN.m-1.
Finally it was tested the crude biosurfactant against the bacteria itself in microflotation tests,
resulting in a faster process and improved hematite recovery. The maximum hematite floatability
obtained with the bacteria was around 43.5% at neutral pH whilst the optimum recovery with the
crude biosurfactant was 95% at pH 3. The results showed a high affinity of the crude biosurfactant
for hematite particles and relatively low reagent consumption.

1
INTRODUCTION
Bioflotation has been extensively studied these late years, because it is an attractive alternative to
replace the common reagents by environmental friendly bioreagents (Dwyer et al., 2012). They are
characterized by having low toxicity and degrading easily, once discharged to the environment. In
addition, the raw material for its production is low cost, renewable and available (Salehizadeh et al.,
2014). Finally, they have potential to treat low grade ores, giving new alternatives for uneconomic
deposits (Mesquita et al., 2003).
On the other hand, even though there is plenty of evidence that bioflotation processes have good
recovery and selectivity (Natarajan, 2006; Mesquita et al., 2003; Merma et al., 2013; Sharma, 2001),
problems such as the low technology including a poor understanding of the mechanisms, kinetics
and thermodynamics of the process (Fomina et al., 2014) hinder a successful scale up. Because the
bacteria is a heterogeneous mixture of several compounds, it is complicated to know the specific
mechanism by which they selectively turn the minerals hydrophobic. Furthermore, it is interesting
to highlight that the theoretical models that describe the adhesion between the mineral and the
bacteria do not account for biological factors (Hermansson, 1999). The inclusion of a biological
frame in the field of bioflotation will have a great significance in understanding what occurs behind
the process.
Some biomolecules that are responsible of the adhesion and selective floatability of the minerals are
surface active substances which are excreted or cell bounded on the microorganism surface
(Kuyumcu et al., 2009). They are known as biosurfactants and have several functions such as
facilitate the growth of their producers by increasing the substrate availability, transporting
nutrients, and acting as biocide agents (Rodrigues et al., 2006). Therefore, in order to narrow the
heterogeneity and reduce the complexity of the bioflotation process; this study focuses on the
extraction, characterization, modeling and use of the biosurfactants as potential flotation reagents.

METHODOLOGY

Microorganism and culture media

A Gram positive bacteria obtained from the Brazilian Collection of Environmental and Industrial
Microorganisms (CBMAI-UNICAMP) was used in this study. The maintenance solid medium
consisted of 10.0 g dm-3 of glucose, 5.0 g dm-3 of peptone, 3.0 g dm-3 of malt extract, 3.0 g dm-3 of
yeast extract, 2.0 g dm-3 of CaCO3, and 12 g dm-3 of agar. In addition, the composition of the liquid
medium was the same as the solid media with the absence of agar.

Crude biosurfactant extraction and characterization by FTIR

The extraction method using hot ethanol was based on the work of Moreau et al. 2003. In order to
estimate the concentration of the crude biosurfactant the equation 3.1 was used. The crude
biosurfactant concentrate was stored at 4°C for a maximum of 5 days. This assumption was based
on the biosurfactant biodegradability studies conducted by Pei et al. 2009.
In addition, in order to identify the functional groups presented in the crude biosurfactant, infrared
absorption spectra was carried out on a Nicolet FTIR 2000 spectrophotometer; a KBr matrix was
used as reference and a deuterated triglycine sulphate (DTGS) as detector. The sample was dried at
75 °C and it was properly mixed with the KBr.

2
Surface tension measurements
The effect of crude biosurfactant concentration on the surface tension of distillated water was
studied at neutral pH and varying the biosurfactant concentration from zero until two hundred
fifty ppms, with intervals of fifty ppms. The surface tension measurements were performed using
the ring method in a Kruss K10 digital tensiometer.
In addition, to estimate the critical micelle concentration of the crude biosurfactant two tangents at
the minimum and maximum surface tension points were drew, corresponding the intersection to
the CMC. This estimation was based on França et al. (2015).

Microflotation studies
The flotation tests were carried out in a modified Hallimond tube, with 10-3 M NaCl as background
electrolyte, air rate of 35 dm3 min-1, hematite size fraction +75 -150 µm, conditioning time 2 min. and
flotation time 1 min. A variable concentration of bioreagents was added, the flotability was then
calculated as the ratio of floated and the total weighed mineral.

RESULTS AND DISCUSSION

Crude biosurfactant extraction

The bacteria was incubated in a rotatory shaker at 125 rpm by 3 days. After the incubation time, the
biomass was centrifuged and washed two times with deionized water to remove the broth. The
crude biosurfactant recovery was 0.3 g per liter of broth. It is necessary highlight that the crude
biosurfactant may be composed of phospholipids, fatty acids, aminolipids in addition to the
glycolipids, due to the non-selectivity of the ethanol and the cell lysis as result of the relatively high
temperatures and pressures in the autoclaving.

Comparison of the crude biosurfactant and gram positive FTIR spectra

It can be seen in the fingerprint region of bacteria, below 1500 cm-1, a large number of absorptions
due to a variety of C-C, C-O and C-N single-bond vibrations may be occur; this region is unique for
every substance (McCurry, 2012). Additionally, it was found an intense peak between 1750 and
1620 cm-1 characteristic of aromatics, aldehydes, ketones and esters; Sharma (2001) reported the
same peak. The mycolates that form part of the cell envelope (Stratton et al., 1999) and are
responsible of the bacteria hydrophobicity (Sutcliffe, 1998) may be reflected by the alkane, ketone
and aldehyde peaks. The presence of amino groups and aromatic compounds, which may be part
of aromatic amino acids, indicate proteic substances which where reported to play a determinant
role in flocculation and flotation processes (Patra et al., 2008).
Regarding the crude biosurfactant, Table 1 shows the possible functional found in the crude
biosurfactant. The alcohol, alkane, alkene, and ketone groups at 3398, 2929 and 1629 cm -1
respectively, may indicate the presence of mycolic acids (Nishiuchi et al., 2000). The identification of
aromatic groups as well as amine groups at 1400, 1548 and 3350 cm -1, could indicate the presence of
polar amino acids such as tyrosine (Berg et al., 2012).

3
Table 1 Possible functional groups identified by the FTIR spectra of the crude biosurfactant

Wavenumber(cm-1) Intensity Possible functional group Structures*

3397.94 High Alcohols

3350.00 Medium Amines

2929.27 Medium Alkanes

1629.44 High Alkenes, ketones ,

1400.41 Medium Aromatic

1047.35 Medium Alkanes

Finally, the similarity between both spectra as is show in Figure 1 indicate that the biosurfactant
was contained in the biomass, which also was used in microflotation tests. Although, it was
reported bacterial proteins play a determinant role in flocculation and flotation processes because
of its amphiphilic character (Patra, 2008). In the case of this bacteria evidence, suggest that proteic
compounds are not the active substances responsible of conferring mineral hydrophobicity, because
they would be denaturized by the temperature and pressure conditions as well as the solvent itself.

Figure 1 FTIR spectra of the bacteria (above) and the dried crude biosurfactant extracted by hot ethanol

Surface tension measurements

Figure 2 shows the surface tension in function of the crude BS concentration at neutral pH. The
estimated CMC is around 92 ppm and the surface tension decreases until 50.5 mN m-1. On the other
hand, most of the biosurfactants extracted from bacteria were reported to low the surface tension of
water from 72 mN m-1 to values between 19 and 43 at CMC values between 17 and 37 ppm after
refination processes (Christova et al., 2014).

4
Figure 2 Effect of the crude biosurfactant concentration on the surface tension of deionized water at 23°C and
neutral pH

Merma et al. (2013) studied the surface tension of a gram positive bacteria, which also had a
hydrophobic character. It was reported that at 300 ppm and pH 3, the surface tension of water was
lowered to 52.5 mN m-1. Therefore, the crude biosurfactant extracted has a moderate capacity to low
the surface tension of water but it is not as efficient as a pure biosurfactant. In addition, the extract
may be composed of polymeric biosurfactants, which do not necessarily reduce the surface tension,
but they may effectively reduce the interfacial tension between immiscible liquids and form stable
emulsions (Dhanarajan et al., 2014).

Hematite microflotation tests

Figure 3 shows the hematite floatability using the crude biosurfactant extracted from bacteria. The
optimum is reached at pH 3, around 96%. Vidyadhar et al. (2014) reported a hematite flotability of
90 % at 304 ppm of sodium oleate in microflotation tests carried out in a Hallimond tube. Literature
review suggest that most of the non-toxic biosurfactants are anionic (Christova et al., 2014); in
addition, based on electrophoretic studies reported by Mesquita et al. (2013), the hematite isoelectric
point was reported around 5.1. It is easy to see the correlation between the pH and the BS adhesion.
At acid pH there is going to be electrostatic attraction between the mineral surface and the anionic
biosurfactant, resulting in maximum adhesion and therefore maximum hematite recovery. On the
other hand at basic pH there is going to be minimal adhesion because of the electrostatic repulsion.

5
 Figure 3 Hematite flotation test, NaCl 10-3 M as background electrolyte, size fraction +75-150, conditioning
time 2 min, flotation time 1 min and air rate of 35 dm3 min-1

Figure 4 shows the mineralized froth using the crude biosurfactant to float 1 g of hematite at pH 3.
The affinity of the biosurfactant for hematite is significant.

Figure 4 Biosurfactant froth with hematite particles (+75 -150 µm), NaCl 10-3 M, and concentration of 125 ppm
and pH 3

Figure 5 shows a bar diagram comparing the hematite floatabilities using the biosurfactant and
bacteria itself. In order to evaluate this contrasting, both experiments were conducted at similar
conditions, airflow, conditioning time, flotation time, hematite sample, size fraction and
background electrolyte concentration.
It is worth to note that the hematite recovery using the biosurfactant as collector is much greater
than the bacteria itself. In addition, while the maximum hematite recovery using the bacteria
resulted at pH 7, the biosurfactant had an optimum recovery at pH 3. The high performance of the
bioreagent at acid pH is characteristic of most biosurfactants which are stable, even at extreme
temperatures, pH and salinity (Kosaric, 2001).

6
 Figure 5 Bar diagram comparing hematite floatabilities using the crude biosurfactant (green) and bacteria
(blue)

CONCLUSIONS
The extraction method allowed the recovery of the cell associated and intracellular substances that
are responsible of conferring hydrophobicity to the hematite surface. The crude biosurfactant
recovery was around 0.3 g.dm-3.
The functional groups identified in the FTIR spectra of the crude biosurfactant suggest that it is
composed of mycolyc acids and polar aminoacids. The former is responsible of the bacteria
hydrophobicity. The similarity of both spectra, the bacteria and the crude biosurfactant, indicate
that the biosurfactant was contained in the biomass. Finally, it was discarded the presence of
proteins in the crude biosurfactant, because of its denaturation along the extraction process.
Throughout surface tension measurements of the crude biosurfactant, it was estimated a CMC
around 92 ppm, lowering the water surface tension from 72 to 52.5 mN.m -1.
Regarding hematite flotation tests with the bisurfactant showed that it required a relatively shorter
flotation time, requiring of 1 min to achieve its highest hematite recovery of 95%, at acid pH of 3
and crude biosurfactant concentration of 125 ppm. Finally, it was concluded a better floatability of
hematite in the presence of the crude biosurfactant than the bacteria itself. Such results show the
potential use of the crude biosurfactant as a low toxic and environmental friendly alternative
against synthetic reagents in hematite flotation.

7
ACKNOWLEDGEMENTS
The authors acknowledge PUC-Rio (Pontifical Catholic University of Rio de Janeiro), CNPq
(Conselho Nacional de Desenvolvimento Científico e Tecnológico), Fundação de Amparo à
Pesquisa do Estado do Rio de Janeiro-FAPERJ and ITV-VALE for the financial support.

REFERENCES
Berg, J. M., Tymoczko J. L., Stryer, L. (2012) Biochemistry. 7ed., W. H. Freeman and Company, New
York.
Christova, Nelly; Stoineva, Ivanka (2014) Trehalose Biosurfactants; Biosurfactants Research Trends
and Applicactions. Ed. By Mulligan, Catherine N.; Sharma, Sanjay K.; Mudhoo, Ackmez; CRC
Press; Boca Raton.
Dhanarajan, G.; Sen, R. (2014) Amphiphilic molecules of microbial origin; Biosurfactants Research
Trends and Applicactions; Ed. By Mulligan, Catherine N.; Sharma, Sanjay K.; Mudhoo, Ackmez;
CRC Press; Boca Raton.
Dwyer, R., Bruckard, W. J., Rea, S., & Holmes, R. J. (2012). Bioflotation and bioflocculation review:
microorganisms relevant for mineral beneficiation. Mineral Processing and Extractive Metallurgy,
121(2), 65–71. http://doi.org/10.1179/1743285512Y.0000000005
França, Í. W. L.; Lima, A. P.; Lemos, J. A. M.; Lemos, C. G. F.; Melo, V. M. M.; Santana, H. B.;
Gonçalves, L. R. B. (2015). Production of a biosurfactant by Bacillus subtilis ICA56 aiming
bioremediation of impacted soils; Catalysis Today, 255, 10–15.
http://doi.org/10.1016/j.cattod.2015.01.046
Fomina, M.; Gadd, G. M. (2014); Biosorption: Current perspectives on concept, definition and
application; Bioresource Technology, 160, 3–14; http://doi.org/10.1016/j.biortech.2013.12.102
Hermansson, M. (1999) The DLVO theory in microbial adhesion. Colloids and Surfaces, 14, 105–119.
Kosaric, N., (2001) Biosurfactants and their application for soil bioremediation. Food Technology
and Biotechnology, 39(4), pp.295-304.Vancouver
Kuyumcu, H. Z., Bielig, T., Vilinska, A., & Rao, K. H. (2009). Biocoagulation and its Application
Potentials for Mineral Bioprocessing. The Open Mineral Processing Journal, 2(1), 1–11.
http://doi.org/10.2174/1874841400902010001
Lang, S.; Philp, J. C. (1998). Surface-active lipids in Rhodococci. Antonie van Leeuwenhoek, 74:59–
70.
Mcmurry, John (2012) Organic Chemistry; Brooks/Cole 8 ed.; Canada.
Merma, A. G., Torem, M. L., Morán, J. J. V, & Monte, M. B. M. (2013). On the fundamental aspects of
apatite and quartz flotation using a Gram positive strain as a bioreagent. Minerals Engineering, 48,
61–67. http://doi.org/10.1016/j.mineng.2012.10.018
Mesquita, L.M., Lins, F.F. & Torem, M.L. (2003). Interaction of a hydrophobic bacterium strain in a
hematite–quartz floatation system. International Journal of Mineral Processing, vol. 71, pp. 31–44.

8
Moreau, R., Powell, M. J., & Singh, V. (2003). Pressurized liquid extraction of polar and nonpolar
lipids in corn and oats with hexane, methylene chloride, isopropanol, and ethanol. Journal of the
American Oil Chemists’ Society, vol. 80, pp. 1063–1067. http://doi.org/10.1007/s11746-003-0821-y
Natarajan, K.A., (2006). Microbially-induced mineral flotation and flocculation: prospects and
challenges, Proceedings of XXIII International Mineral Processing Congress 487–498.
Nishiuchi Y, Baba T, Yano I (2000) Mycolic acids from Rhodococcus, Gordonia, and Dietzia. J
Microbiol Methods 40:1–9
Patra, P., & Natarajan, K. a. (2008). Role of mineral specific bacterial proteins in selective
flocculation and flotation. International Journal of Mineral Processing, 88(1-2), 53–58.
http://doi.org/10.1016/j.minpro.2008.06.005
Pei, X.; Zhan, X.; Zhou, L. (2009). Effect of biosurfactant on the sorption of phenanthrene onto
original and H2O2-treated soils. Journal of Environmental Sciences, vol. 21, pp. 1378–1385.
Rodrigues, L., Banat, I.M., Teixeira, J. & Oliveira, R. (2006). Biosurfactants: Potential applications in
medicine. Journal of Antimicrobial Chemotherapy, vol. 57, pp. 609-618.
Salehizadeh, H.; Yan, N. (2014); Recent advances in extracellular biopolymer flocculants;
Biotechnology Advances, 32(8), 1506–1522. http://doi.org/10.1016/j.biotechadv.2014.10.004
Sharma, P. K. (2001) Surface studies relevant to microbial adhesion and biofloatation of sulphide
minerals. Doctoral Thesis, Lulea University of Technology; Sweden; 2001.
Stratton, H. M.; Brooks P. R.; Seviour R. J. (1999) Analysis of the structural diversity of mycolic acids
of Rhodococcus and Gordonia [correction of Gordonla] isolates from activated sludge foams by
selective ion monitoring gas chromatography-mass spectrometry (SIM GC-MS); Journal of
Microbiological Methods, 35(1):53-63.
Sutcliffe, I. C. (1998). Cell envelope composition and organisation in the genus Rhodococcus.
Antonie van Leeuwenhoek, 74(1-3), 49-58.
Vidyadhar, A.; Kumari, N.; Bhagat, R. P. (2014) Adsorption Mechanism of Mixed Cationic/Anionic
Collectors in Quartz–Hematite Flotation System; Mineral Processing and Extractive Metallurgy
Review, 35(2), 117–125; http://doi.org/10.1080/08827508.2012.723649

9
Improved Flotation Performance through
Modernization with Outotec Forced Air Technology
Chris Cruickshank1, Koen Aberkrom1, Miika Tirkkonen1 and Alejandro Yáñez2
1. Outotec, Finland
2. Outotec, Chile

ABSTRACT
In the current tough mining environment with lower commodity prices, operators are placing
greater emphasis on asset management and operational profitability. The efficiency of the installed
process equipment plays a crucial role and can often be a determining factor in the overall circuit
performance. Inferior/outdated process equipment can result in lower metallurgical performance
and poor plant availability, reducing operating margins. In addition, external factors (such as
improved energy consumption, tighter regulations on health and safety standards and more
advanced process control strategies) can further highlight inefficiencies in previously installed
process equipment.
Operators facing these challenges have the difficult decision whether to invest in new equipment or
to modernise their existing equipment. In many cases, the modernisation route is preferred due to
the lower capital cost and shorter installation times. New equipment is usually only preferred when
there is a risk of mechanical failure or when operators are targeting a significant tonnage increase.
Recently Outotec has retrofitted multiple self-aspired and forced-air, non-Outotec flotation cells in
various locations around the world, converting the existing design to the Outotec forced air
technology with FloatForce mixing mechanism. This paper summarizes some of these
modernization cases and details observed benefits and discusses about challenges in these projects.

There is no full article associated with this abstract.

1
In-Situ Gas Holdup Sensor for Industrial Flotation
Machines
Miguel Maldonado1*, Pablo Olivares1 and César Gomez2
1. Universidad de Santiago de Chile
2. McGill University, Canada

ABSTRACT
A submersible sensor for real-time gas holdup measurement in the collection zone of industrial
flotation machines is introduced. The working principle is based on the generation and
measurement of a pulp flow using a gas-exclusion cell immersed in the aerated slurry. The gas
exclusion cell consists of a variable-area open vertical tube designed to be filled only with pulp by
preventing gas bubbles entrance. The higher density of the fluid column within the cell generates a
downward flow of pulp; the rate of this flow is proportional to the gas holdup outside the cell. The
gas holdup measurement is achieved by tracking the velocity of the pulp flow leaving the gas-
exclusion cell, using a magnetic flowmeter. Construction details of a prototype and its testing in
two- and three-phases are discussed.

1
INTRODUCTION
Flotation process optimization consists of finding and maintaining operating conditions that
maximize flotation performance measured in terms of recovery and grade. The selective separation
of mineral particles proceeds in two interactive zones where different and simultaneous
phenomena take place: in the collection zone, the valuable particles are separated by collection on
the surface of bubbles, while in the froth zone, the bubble-particle aggregates are gathered, drained
to remove entrained particles, and taken into the concentrate stream. The cell overall performance,
which can be determined from proper stream sampling, is a function of the recoveries in the
individual zones. These zone recoveries cannot be independently measured nor calculated from
operating variables. Progress in the assessment of these recoveries has been slow because on-line
and reliable sensors to measure properties of the drastically different dispersions existing in both
zones are not available.
In the case of the collection zone, it has been demonstrated that the recovery of hydrophobic
particles on the surface of bubbles, considered as a process with first-order kinetics, correlates with
the bubble surface area flux (Finch et al., 2000; Hernandez, et al., 2003; Lopez-Saucedo et al., 2012).
Bubble surface area flux (bubble area generated per unit time and cell cross-sectional area), is a
variable affected by gas flow rate, bubble size distribution, and frother type and concentration. Gas
holdup (volumetric fraction of gas dispersed in the slurry) is also affected by the same variables,
and as such, carries information on the interacting variables that define the performance of the
process. Therefore, a correlation between gas holdup and bubble surface area flux was expected
and demonstrated (Finch et al, 2000), and since gas holdup is easier to measure as no bubble size is
required, it has the potential to be the basis of a cell control and optimization strategy if a
continuous sensor is available.
Several methods have been proposed to measure gas holdup in flotation cells. Efforts to develop
continuous methods are based on the pressure difference measured at different depths (Finch and
Dobby, 1990), fiber optics (Schweitzer et al., 2001), sound propagation through an aerated fluid
(O´Keefe et al., 2007), tomography (Patel and Thorat, 2008), and changes in the electrical
conductivity of a liquid with a non-conducting dispersed phase (Sigrist et al., 1980; Uribe Salas et
al., 1994; Tavera et al., 1996; Gomez et al., 2003). These methods, however, have found little
application at industrial scale except for the conductivity-based gas holdup sensor, which has been
widely used at the lab and industrial scale for diagnostic purposes (Gomez and Finch, 2007).
The objective of this communication is to introduce a new submersible on-line sensor to measure
gas holdup in the collection zone of industrial flotation machines. The working principle, its main
components, and the theory behind the measurement, are described. Design and construction
details of a prototype are discussed, and the results of its lab testing in two- and three-phases are
included.

A NEW GAS HOLDUP SENSOR

Components
The proposed gas holdup sensor comprises a vertical gas-exclusion cell, connected to a magnetic or
ultrasonic flow sensing element to measure the velocity of the slurry leaving the gas-exclusion cell,
a flow transmitter, and a data processing unit, as shown in Figure 1.

2
The component that makes the measurement possible is the gas-exclusion cell, which consists of a
variable-area open vertical tube designed to be filled only with pulp by preventing the entrance of
rising gas bubbles. The higher density of the fluid column thus created within the cell generates a
downward flow of pulp; the rate of this flow is proportional to the gas holdup outside the cell.
The flow sensing element and the flow transmitter, preferably installed remotely outside the
flotation machine, provides a signal indicative of the velocity of the slurry leaving the gas-exclusion
cell, which is further processed to provide a signal indicative of the gas holdup.

(a) (b)

Figure 1 Schematic representation (a), and photograph (b), of the gas holdup sensor components

Gas-exclusion Cell: Working Principle and Theoretical Considerations

The gas-exclusion cell is a variable-area open vertical tube with a larger diameter at the top than at
the bottom. It is designed to be filled only with pulp by initially reducing the entrance of rising gas
bubbles; the higher density of the fluid column within the cell generates a downward flow of pulp,
which completes the exclusion of bubbles. Dimensions of the gas-exclusion cell must be selected to
have a downward flow of pulp that results in a velocity large enough to be reliably measured after
leaving the cell and, at the same time, a velocity entering the top sufficiently small for not
entraining small bubbles.
A schematic of the gas-exclusion cell is shown in Figure 2 indicating relevant dimensions: a top
section of diameter D, a bottom section of diameter d, joined by a truncated inverted cone, with a
total height L. Application of an energy balance to the flow of an incompressible fluid between the
top and bottom planes results in the following equation:

dE Pt vt2 Pb vb2 (1)

= Ft ( + gzt + ) − Fb ( + gzb + ) − PLoss
dt ρt 2 ρb 2

3
where
E : energy of the slurry in the defined system
Ft , Fb : mass flow rate of slurry at planes t and b, respectively
Pt , Pb : pressure at planes t and b, respectively
zt , zb : elevations at planes t and b, respectively
vt , vb : slurry velocity at planes t and b, respectively
ρt , ρb : slurry density at planes t and b, respectively.
PLoss : power loss due to friction between planes t and b
g : acceleration due to gravity

At steady state, the energy change between planes t and b is constant, and the pulp mass flows
crossing the planes t and b are the same (Ft=Fb). In the ideal case when there are no friction losses:

Pt vt2 Pb vb2
+ zt + = + zb + (2)
ρg 2g ρg 2g

Figure 2 X-ray view of the gas-exclusion cell and its relevant dimensions

The pressure difference between planes b and t can be expressed as a function of the total
length L of the gas-exclusion cell, which is the elevation difference between planes (z t - zb), and the
gas holdup 𝜀𝑔 in the surrounding aerated pulp:
𝜀𝑔
Pb Pt 𝜌 (1 − 100) gL 𝜀𝑔 (3)
− = = (1 − )L
𝜌g 𝜌g 𝜌g 100

For an incompressible fluid, volumetric flows at the planes t and b are the same as the pulp
density is constant, and the velocities are related as follows:

4
 vt A t = vb A b (4)

Replacing Equations 3 and 4 in Equation 2 we have:

100 vb 2
𝜀𝑔 (%) = ( ) [1 − 𝛽 4 ] (5)
2gL Cd

This equation indicates that the gas holdup between the planes t and b outside the gas-exclusion
cell is related to the pulp velocity through its bottom section and, therefore, measurement of this
velocity allows the gas holdup to be calculated. The parameter 𝛽 is the contraction coefficient
defined as the ratio of the bottom to the top section diameters (d/D), and a discharge coefficient (Cd )
is introduced to account for friction losses.

Prototype

Design

A prototype was designed and constructed around a plastic gas-exclusion cell and a
magnetic flowmeter (Siemens, model SITRANS FM Transmag2) with a sensor 911/E-IP68 (Figure
3a). The gas-exclusion cell was made of a long top section, 101.6 mm in diameter, and a bottom
section, 12.5 mm in diameter, joined by a 90o truncated cone. The gas-exclusion cell and the
magnetic flowmeter sensor were flanged together; the flowmeter transmitter cable allowed remote
location of this unit. The magnetic flowmeter requires a fluid with an electrical conductivity of at
least 5 µS/cm, a value well below mineral pulp conductivities.
The design considered a gas holdup measurement range between 5 and 30%, and was
based on the following gas-exclusion cell specifications:
1. A minimum bottom section diameter of 1inch to prevent clogging;
2. A minimum bottom section length of five internal diameters with no flow restrictions;
3. A minimum length of three internal diameters downstream the flowmeter sensor with no flow
restrictions;
4. A maximum downward pulp velocity at the top section of 3 cm/s to maintain size of entrained
bubbles below 0.3 mm (Clift et al., 2005).
5. A minimum pulp velocity at the bottom section of 0.6 m/s to ensure an acceptable measurement
accuracy.

5
 (a) (b)

Figure 3 - (a) Dimensions of gas holdup sensor prototype; and (b), experimental setup to determine
the discharge coefficient

Determination of Discharge Coefficient

The discharge coefficient is defined, in the case of a gas-exclusion cell, as the ratio of the bottom
section flowing velocity to that calculated by applying Bernoulli´s streamline energy equation
assuming there are no friction losses (vbideal ).
vb
Cd = ideal (6)
vb

The characterization of the discharge coefficient for the constructed prototype was accomplished
using the same procedure previously used to determine that of the gas-exclusion cells (Córtes-
López, 1998). The prototype was partially immersed in a tank containing water (or pulp) as
illustrated in Figure 3b. A centrifuge pump suctions water from the bottom of the tank and delivers
a manually controlled flow (using a valve) into the top section of the gas-exclusion cell. For a series
of water flow rates, the level difference (H) between the water in the top section of the prototype
and that in the tank is determined, once steady state is reached.
The discharge coefficient can then be calculated as follows:
vb
Cd =
2gH (7)
√
1 − 𝛽4

where vb is the fluid velocity measured by the flowmeter and H is the level difference between the
water in the top section of the prototype and that in the tank.

6
 For turbulent flow, the discharge coefficient of Venturi tubes, flow nozzles and orifices can
be modelled as follows (Miller, 1996):

b
Cd = C∞ + (8)
(𝛽Reb )n

where C∞ is the discharge coefficient at an infinite Reynolds number, and b and n are parameters
that must be determined experimentally. The Reynolds number in the bottom section Reb is given
by:

𝜌vb d vb d
Reb = = (9)
𝜇 𝜈

where 𝜇 and 𝜈 are, respectively, the viscosity and kinematic viscosity (𝜇 ⁄𝜌) of the fluid.
Calculated discharge coefficients vs. Reynolds numbers obtained in the characterization of
the prototype are included in Figure 4. The model proposed by Miller was found to fit the data
with the values indicated in the following equation and represented in Figure 4 by the dashed line:

0.99
Cd = 0.972 − (10)
(𝛽Reb )0.12

Figure 4 Discharge coefficients vs. Reynolds numbers obtained in the prototype characterization

Figure 5 shows a block diagram representation that summarizes the calculations required to
transform the velocity measurement into a gas holdup measurement.

7
 Figure 5 Block diagram representation of calculations

Laboratory Testing in Two- and Three-phase Suspensions

The prototype performance was assessed in a rectangular column (Figure 6) that was operated with
water (two-phase) and with a diluted pulp (three-phase). The prototype was fully immersed in the
aerated water or pulp. Bubbles were produced by using a SS porous sparger installed at the bottom
of the column and gas holdup was varied by changing the gas flow rate. In both cases, MIBC at a
concentration of 30 ppm was used as a frother. True gas holdup was determined in air-water
testing from differential pressure readings using a transmitter (ABB, model 266DSH), and from
collection zone expansions in three-phase measurements.

Figure 6 Experimental setup to assess performance of proposed gas holdup measuring system

8
The results of the assessment are included in Figures 7 and 8 for the two- and three-phase testing,
respectively. In the case of two-phase testing, prototype gas holdups were higher than those
measured by pressure difference (the dashed line represents perfect match between measurements),
with a bias of around 1%. Similar results were obtained in three-phase testing, run in pulps
containing 5 and 10% solids by weight, which were prepared with ore particles (P80 120 µm).

Figure 7 Comparison of gas holdups measured by the prototype with those determined from pressure signals
in two-phase testing

Figure 8 Comparison of gas holdups measured by the prototype with those determined from collection zone
expansion in three-phase testing

9
DISCUSSION
The on-line application of the sensor requires knowledge of the discharge coefficient (Figure 5),
which depends on the Reynolds number, a parameter that can only be estimated as the pulp
kinematic viscosity is not measured on-line. Therefore, it is necessary to assess how sensitive the
accuracy of the gas holdup measurement is to variations of the selected kinematic viscosity value.
To this end, the sensitivity function between gas holdup and the kinematic viscosity, which
represents the factor by which the relative errors in the kinematic viscosity will be propagated to
𝜀𝑔
the gas holdup measurement, needs to be determined. The sensitivity function (S𝜈 ) between gas
holdup and the kinematic viscosity is defined as:
∆𝜀𝑔
𝜀𝑔 𝜀𝑔 𝜕𝜀𝑔 𝜈
S𝜈 = lim = ∙ (11)
∆ν→0 ∆𝜈 𝜕𝜈 𝜀𝑔
𝜈

The partial derivative in the above equation can be determined by use of the chain rule:

𝜕𝜀𝑔 ∂𝜀𝑔 ∂C𝑑 ∂Reb (12)

=
𝜕𝜈 ∂Cd ∂Reb ∂𝜈

as the involved partial derivatives can be obtained from Equations 5, 8 and 9:

𝜕𝜀𝑔 vb2 −2 (13)

= (1 − 𝛽 4 ) 3
𝜕Cd 2gL Cd

∂Cd
= −b𝛽 −n nReb−n−1 (14)
∂Reb

∂Reb −vb d
= 2 (15)
∂𝜈 𝜈

Replacing Equations 13, 14, and 15 into Equation 11:

𝜀𝑔 𝜕𝜀𝑔 𝜈 −2n(𝛽 ∙ Reb )−n

S𝜈 = = (16)
𝜕𝜈 𝜀𝑔 Cd

and Equation 5 into 16, we finally obtain:

𝜀𝑔 (C∞ − Cd )
S𝜈 = 2n (17)
Cd

This equation indicates that to reduce kinematic viscosity effects on the measurement of gas
holdup, the sensor must be designed to have a high discharge coefficient. Considering an average
value of the discharge coefficient of 0.59 (from Figure 4), the sensitivity function takes the value:
𝜀𝑔 (0.972 − 0.59)
S̅𝜈 = 2 · 0.12 = 0.16 (18)
0.59

10
This result means that a relative error of 10 % in the selected value of kinematic viscosity (in this
case that of water at 25oC), will produce a 1.6 % relative error in the gas holdup measurement. This
value is considered low implying that gas holdup measured is largely insensitive to differences
between the selected and real kinematic viscosity values.

CONCLUSIONS
A new approach to measure gas holdup in the collection zone of flotation machines was introduced
and demonstrated. The technique is based on a submersible unit which generates a localized flow
of pulp using a gas-exclusion flow cell. Theoretical considerations demonstrated that the flow of
pulp (and corresponding velocity) was proportional to the gas holdup in the external aerated pulp.
A prototype was designed and constructed using a magnetic flowmeter to measure the pulp exit
velocity leaving the gas-exclusion cell.
The gas holdup calculation requires the discharge coefficient of the gas-exclusion cell which was
characterized and modeled as a function of the Reynolds number. Testing in two- and three-phase
systems demonstrated reliable measurements with standard deviations of 0.8% and 0.5%
respectively. Long term testing in industrial units is being programmed.

ACKNOWLEDGEMENTS
The authors acknowledge financial support from the Chilean Council for Scientific and
Technological Research (CONICYT) through project FONDEF CA13I10320 and CORFO project
15VEIID-45786.

REFERENCES
Clift, R., Grace, J.R., Weber, M.E. (2005). Bubbles, Drops and Particles. Dover Publications, 1st Edition.
Córtes-López, F. (1998). Design of a gas holdup sensor for flotation diagnosis. MSc Thesis, McGill
University.
Finch, J.A. and Dobby, G.S. (1990). Flotation Column. Pergamon Press.
Finch, J.A., Xiao, J., Hardie, C., & Gomez, C.O. (2000). Gas dispersion properties: bubble surface
area flux and gas holdup. Minerals Engineering, 13, No. 4, 365-372.
Gomez, C.O. & Finch, J.A. (2007). Gas dispersion measurements in flotation cells. International
Journal of Mineral Processing, 84, 51-58.
Gomez, C.O., Cortés-López, F., & Finch, J.A. (2003). Industrial testing of a gas holdup sensor for
flotation systems. Minerals Engineering. 16, 6, 493-501
Hernandez, H., Gomez, C.O., & Finch, J.A. Gas dispersion and de-inking in a flotation column.
Minerals Engineering, 16, 2003, 739-744.
Lopez-Saucedo, F., Uribe-Salas, A., Pérez-Garibay, R., & Magallanes-Hernandez, L. (2012). Gas
dispersion in column flotation and its effect on recovery and grade. Canadian Metallurgical Quarterly,
Vol. 51, No.2, 111-117.
Miller, R.W. (1996). Flow Measurement Engineering Handbook. McGraw-Hill Education, 3 Edition.

11
O’Keefe C, Viega J., & Fernald. M (2007). Application of passive sonar technology to mineral
processing and oil sands application. Proceedings of the 39th Annual Meeting of the Canadian
Mineral Processors, CIM, Ottawa (Canada), 429–457.
Patel, A.K., & Thorat, B. (2008). Gamma ray tomography – An experimental analysis of fractional
gas hold-up in bubble columns. Chemical Engineering Science, 137, 376-385.
Schweitzer, J.M., Bayle, J., & Gauthier, T. (2001). Local gas hold-up measurements in fluidized bed
and slurry bubble column. Chemical Engineering Science, 56, 1103-1110.
Sigrist, L., Dossenbach, O., & IBL, N. On the conductivity and void fraction of gas dispersions in
electrolyte solution. Journal of Applied Electrochemistry, 10, 1980, pp. 223-228.
Tavera, F.J., Gomez, C.O., & Finch, J.A. (1996). Novel gas hold-up probe and application in flotation
columns. Trans. Instn. Min. Met., Vol. 105, C99-C104.
Uribe-Salas, A., Gomez, C.O., & Finch, J.A. (1994). A conductivity technique for gas and solids
holdup determination in three-phase reactors. Chemical Engineering Science. Vol. 49, 1, 1-10.

12
Use of Depressant BK-511 in Selective Flotation of
Molibdenite, Andina Division, Codelco
Mauricio Bustamante and Héctor Tobar
Andina Division, Codelco, Chile

ABSTRACT
This summary presents the industrial evaluation of copper sulfide depressant reagent BK-511 in the
process of selective flotation of molybdenite in Codelco Chile Andina Division. This reagent used
for the first time worldwide is developed by the department of mineral processing in Chinese
company BGRIMM.
Sodium Sulfhydrate is a reagent used extensively in selective flotation plants due to its excellent
depression of copper sulfides. However, the required high consumption involves handling of large
volumes, high production costs, health risks of workers by generate hydrogen gas type. The BK-511
reagent has been developed as a non-toxic depressant for copper minerals that can work in low
doses, giving less pollution, lower treatment costs, low toxicity and high selectivity.
During 2015, Andina Division conducted an industrial tests with reagent BK-511. The industrial test
was developed during the months of December 2015 to January 2016. During this period, 18 tons of
reagent were used, considering the replacement of 50 % of sodium hydrosulfide in the preparation
for the addition of 120 liters of BK-511. The methodology consisted of weekly assessment of reagent
addition to exhaust the 18 tons.
Among the results obtained in industrial tests, the reagent BK-511 proved be able to reduce the
consumption of sodium hydrosulfide, without affecting efficiency and product quality.
Due to the obvious result obtained during the first week of reagent application BK-511, purchase
management starts to acquire an additional 36 tons of reagent to continue testing in order to
optimize the dose and metallurgical results.
The averaged consumption of sodium hydrosulfide on periods of use of the reagent BK-511 was 2,6
kilos per ton, which implies a reduction of 42 % compared to historical indicator 4,5 kilos per ton
and a 33,5 % compared with the indicator budgeted for 2016 (3,96 kilos per ton).

There is no full article associated with this abstract.

1
Modelling H2S Gas Generation in
Copper/Molybdenum Flotation Cells
Humberto Estay, Karina Gonzalez and Rodrigo Kong
Arcadis, Chile

ABSTRACT
Copper/Molybdenum flotation process uses sulfide and sulfuric acid as main reagents in order to
be feasible the separation of copper and molybdenum. Typical pH values of this process ranging
from 7.0 to 8.5, which promotes Hydrogen Sulfide (H2S) generation. In this context, the design of
the process must provide full control on gaseous H2S evolution to ensure personnel safety. Even
though gas extraction and scrubber systems play a fundamental role in the control and reduction of
potentially dangerous H2S levels, currently there are no formal methods to estimate the H 2S
generation in the Copper/Molybdenum flotation process in order to afford a robust design for the
extraction and scrubber systems.
The objective of the present work is to propose a general methodology to predict the generation of
H2S gas in the Copper/Molybdenum flotation process. For this purpose, a mathematical model was
developed, supported by the phenomenology of the process and considering the system’s
operational conditions. This model considers the chemical equilibrium of H 2S in solution, the
liquid-gas phase equilibrium and the mass transfer phenomena between the liquid and gas phases.
The results predicted by the model have indicated rates of H2S gas generation (0.5-32 g/hm3),
showing H2S concentrations ranging between 20 and 1000 ppm at the inlet of the scrubbing system
under typical operational conditions, when low pH (7.0) and high power/volume rate of the cell are
the most critical conditions.

1
INTRODUCTION
Copper/Molybdenum flotation process uses sulfide and sulfuric acid as main reagents in order to
be feasible the separation of copper and molybdenum. Typical pH values of this process ranging
from 7.0 to 8.5, which promotes Hydrogen Sulfide (H 2S) generation. In this context, the design of
the process must provide full control on gaseous H2S evolution to ensure personnel safety. Even
though gas extraction and scrubber systems play a fundamental role in the control and reduction of
potentially dangerous H2S levels, currently there are no formal methods to estimate the H 2S
generation in the Copper/Molybdenum flotation process in order to afford a robust design for the
extraction and scrubber systems and quantify H2S generation rate for environmental issues.
In this context, this study present a model which predict H2S gas generation in a flotation cell, based
on thermodynamic and mass transfer phenomenology. This model depends on operational
conditions (pH, agitation power and sulfide dosage), system variables (temperature and
atmospheric pressure) and flotation cell characteristics (airflow, volume, dimensions, among
others).

Description of Copper/Molybdenum flotation Process

Many copper ores contain molybdenite (MoS 2) and the flotation process consists mainly of
separating copper sulfide minerals and molybdenite from the waste rock and other minerals.
At first, the crushed copper-molybdenum ore is fed to the grinding circuit where is mixing with
water and ground in order to obtain a correct particle size. Lime, copper collectors (xanthates) and
diesel oil for collecting molybdenite are usually added at this point. Frothers and copper collectors
are added to the slurry previously to the rougher flotation cells stage. Additional copper collector is
often added at other points. The rougher-scavenger concentrate enriched in copper and
molybdenum may go to regrind and then into the cleaner flotation cells. These stages provide
additional cleaning and increase the concentrate grade. This material is called the collective Cu-Mo
concentrate.
The Cu-Mo concentrate is pumped into the molybdenum plant where molybdenite is separated to
the copper sulfides, which is thickened and conditioned with sulphuric acid (H2SO4), water, diesel
oil (molybdenite collector) and sodium hydrosulfide (copper minerals depressant). Sulphuric acid
and sodium hydrosulfide (NaHS) are added at other points of the process. The molybdenite
flotation process occurs at pH values ranging between 7.0 and 8.5. The tailings generated from the
molybdenum plant correspond to final copper concentrate. After rougher flotation and several
cleaning stages, a final molybdenum concentrate is produced, containing Mo grades typically
greater than 55%.
A simplified diagram of a typical molybdenite flotation process is shown in Figure 1.

2
 Figure 1 Typical molybdenite flotation process

In the conditioning tank and flotation cells where sulphuric acid and sodium hydrosulfide are
added H2S is generated according to Equation (1). The gaseous H 2S is collected by a gas extraction
system and transported to the scrubber.

2 NaHS  H 2 SO4  2H 2 S (aq)  Na2 SO4 (1)

METHODOLOGY

Modelling H2S gas generation in Copper/Molybdenum flotation process

Chemical equilibrium in solution

The sulfide anion in water forms acid according to the following dissociation reaction (Lewis, 2010):

 HS   H 
H2S  (2)

The logarithmic exponent of the dissociation constant pKa of H2S at 25ºC is 6.99 (Lewis, 2010). pKa
values higher than 6.99 increase the hydrosulfide anion content in comparison with the dissolved
H2S content while pKa values lower than 6.99 promote the presence of dissolved H2S in the solution.
A typical curve of H2S and hydrosulfide anion distribution in aqueous solution at 25ºC is shown in
Figure 2.

3
 1,0

Ionization Fraction, [H2S]/[HS-]

0,8
H2S
0,6 HS-

0,4

0,2

0,0
3 4 5 6 7 8 9 10
pH

Figure 2 H2S and hydrosulfide anion distribution in aqueous solution at 25ºC

Therefore the H2S and HS- contents in solution can be determined at a fixed temperature by the
equilibrium constant (Ka) described by Equation (3):

[ H  ][ HS  ]
Ka  (3)
[H 2 S ]

On the other hand, the dissolved H2S content depends on temperature, since the chemical
equilibrium of Equation (3) depends on temperature. This relationship is described by the Van’t
Hoff’s equation as:

dLnKa h
 (4)
dT RT 2

where Δh is the standard enthalpy change of reaction and R is the universal gas constant (8.314
J/molK). The value of Δh is 20.6 kJ/mol at 25ºC (Poling et al., 2000).

Liquid-gas phase equilibrium of H2S

As mentioned above the Copper/Molybdenum flotation process includes a gas extraction and
scrubbing system connected to the process equipment where air used in the flotation process flows
into scrubbing system. The air bubbles in contact with slurry in the flotation cell generates a liquid-
gas interface, where dissolved H2S is volatilized into the air as H2S gas. At the interface the equal
chemical potentials between both phases generates a liquid-gas phase equilibrium which can be
determined by Henry’s law for H2S (Poling et al., 2000):

P y H 2S  H H 2S xH 2S (5)

4
where P is the pressure of the system (atm), yH2S is the molar fraction of H2S in the gas phase, HH2S is
the Henry’s law constant (atm/molar fraction) and xH2S is the molar fraction of H2S in the liquid
phase. The Henry’s law constant depends on temperature in a logarithmic relationship. Perry and
Green (1997) reported data of H2S phase equilibrium, which has been adjusted in this work, to
obtain the following Equation (6):

 1789.3
Ln H H 2 S   12.261 (6)
T

where HH2S is Henry’s law constant (atm/molar fraction) and T is absolute temperature (K).

Mass transfer of H2S in Copper/Molybdenum flotation process

Mass transfer of H2S from flotation slurry into the air bubbles is determined by the difference of
chemical potentials between liquid phase and gas phase. The H2S transfer rate for this system can
be estimated by a H2S concentration gradient between the bulk and the interface in the gas phase,
and a mass transfer coefficient, which is dependent on the geometry of the system, physical
properties of fluids and operational conditions. Figure 3 shows a schematic diagram for the H2S
mass transfer promoted by a concentration gradient from the bulk of the liquid phase (contained in
the slurry) to the bulk of the gas phase (contained in the bubble), H2S being volatilized at the
interface. Equation (7) shows the general equation to estimate the H2S transferred in the system.

N H 2S  K AT Cmlg (7)

In Equation (7) NH2S is the overall H2S transfer rate (mol/s), K is the overall mass transfer coefficient
based (m/s), AT is the total area of mass transfer (m2), and ΔCmlg is the logarithmic mean driving
force in the gas phase (mol/m3). The estimation of mass transfer using Equation (7) has been
developed in previous work to others application (Estay et al., 2012). The overall mass transfer
coefficient is strongly dependent by the individual mass transfer coefficient of the liquid (slurry)
phase (kL). The estimation of kL was performed using a correlation based on the effect of slurry
properties (Beenackers and Van Swaaij, 1993).

5
 Feed

Air Gas phase

H2S(g)
Froth
CH2Sg

CH2Sg < CH2Sg0

CH2Sg0
CH2Sl0

CH2Sl > CH2Sl0

Concentrate

CH2Sl
Liquid phase
Tailings

Bubbles
Slurry Agitator

Figure 3 Schematic diagram of the H2S mass transfer from the bulk of the liquid phase to the bulk of the gas
phase

Design criteria of Copper/Molybdenum flotation process

The H2S volatilization in Copper/Molybdenum flotation process depends on the site conditions
(temperature and pressure), operational parameters (pH and sulfide dosage) and sizing of the
flotation process (cells volume, power of agitators, air flow, among others). The
Copper/molybdenum flotation process must operate pH values ranging between 7.0 and 8.5 in
order to maximize the separation. Equation (7) shows the dependence of H2S transferred with area
of mass transfer. According to Gorain et al. (1999) the bubble surface area can be estimated as
shows the following equation.

c
Q
S b  aN   As d P80e
b
s
(8)
 A

Where Sb is the bubble surface area flux (m2/m2s), NS is the impeller peripheral speed (m/s), (Q/A) is
the air flow rate per unit cell cross-sectional area, (As) is the impeller aspect ratio (cm/s), (P80) is 80%
passing feed size (µm), a, b, c, d and e are the parameters for the model of cell reported by Gorain et
al. (1999).

RESULTS AND DISCUSSION

The simulation of H2S gas generation in the Copper/Molybdenum flotation process has permitted
to identify the key parameters in the system. In this context, operational parameters (pH and
power/volume ratio) have been analyzed.

6
Effect of pH in flotation cells
The solution pH in the flotation cell is one of the most important parameters in
Copper/Molybdenum flotation process since copper/molybdenum separation takes place when pH
is ranging between 7.0 and 8.5. As shown in Figure 4, by varying the pH in the flotation cell from
pH 8.5 to pH 7.0 an increase of about 1300% in the H2S transferred ratio was produced, due to H2S
distribution in solution (Figure 2) where H2S content can rise from 3 to 50%. This result
demonstrate the high impact of pH value in the system.

35 1200

30 1000
H2S gas generation rate, g/hm3

H2S concentration, ppm

25
800
20 H2S gen rate
600
H2S, ppm
15
400
10

5 200

0 0
7 7,5 8 8,5 9
pH

Figure 4 H2S gas generation rate and concentration as function of pH values, at power/volume ratio of 2.9

Effect of the power/volume ratio in flotation cells

The most incidental parameter in the gaseous H2S generation is the liquid mass transfer resistance,
which is impacted by the power/volume ratio in the flotation cells. In this connection an increase in
the agitation power reduces the liquid mass transfer resistance, thus affecting the H2S transferred
ratio as shown in Figure 5. Thus, the effect of agitation power on H2S gas generation should be
taken into account for the Copper/Molybdenum flotation process design.

7
 5 120
4,5
100
4

H2S gas generation rate, g/hm3

H2S concentration, ppm

3,5
80
3
2,5 60
2
H2S gen rate 40
1,5
1 H2S, ppm
20
0,5
0 0
2,2 2,3 2,4 2,5 2,6 2,7 2,8 2,9 3
Power/Volume, HP/m3

Figure 5 H2S gas generation rate and concentration as function of Power/Volume ratio values, at pH 8.5.

CONCLUSION
A general methodology has been constructed to predict H2S gas generation in
Copper/Molybdenum flotation process as a function of operational parameters. According to the
results obtained from the model simulations, pH and the power/volume ratio in the reactors are
relevant parameters in H2S gas generation were. The results predicted by the model indicate that
H2S gas generation is highly sensitive to operational conditions (pH and power/volume ratio)
obtaining values between 2-32 g/hm2 and that the H2S concentrations at the inlet of the scrubbing
system could be higher than 1000 ppm when pH is 7,0 and power/volume ratio is close to 3,0. For
pH values close to 8,5, H2S gas concentration is around 20 ppm, decreasing drastically due to H 2S
distribution in solution. These values indicate that under normal processing conditions the overall
H2S gas generation should present dangerous levels to personnel in case of leaks. The model and its
conclusions should be tested against the operational values for H2S contents in the gas streams of
Copper/Molybdenum flotation plants currently operating around the world. However, this fact
does not limit the results obtained in this work because the accuracy of the model is determined by
the liquid mass transfer resistance which under most conditions is the phase controlling H2S
transfer. In this context, the further studies in this matter must be experimental test work in order to
validate the results of this model. Finally, it is believed that the model developed in this study can
be applied as a useful tool to predict risk conditions in the Copper/Molybdenum process and to
guide a correct design of their scrubbing and extraction systems, ensuring safety in the process.

REFERENCES
Beenackers, A., Van Swaaij, W. (1993) ‘Mass Transfer in Gas-Liquid Slurry Reactors’, Chemical
Engineering Science, vol. 48, no. 18, pp. 3109-3199.
Estay, H., Becker, J., Carvajal, P., Arriagada, F. (2012) ‘Predicting HCN gas generation in the SART
process’, Hydrometallurgy, vol. 113-114, pp. 131-142.

8
Gorain, B.K., Franzidis, J.P, Manlapig, E.V. (1999) ‘The empirical prediction of bubble surface area
flux in mechanical flotation cells from cell design and operating data’, Minerals Engineering, vol. 12,
no. 3, pp. 309-322.
Lewis, A.E. (2010) ‘Review of metal sulphide precipitation’, Hydrometallurgy, vol. 104, pp. 222-234.
Perry, R., Green, D., (1997) Perry’s Chemical Engineers’ Handbook, Seventh edition. McGraw-Hill,
New York.,
Poling, B.E., Prausnitz, J.M, O’Conell, J.P. (2000) The Properties of Gases and Liquids, McGraw-Hill,
New York.

9
On the Interaction Effect between Xanthate Collector,
Frothers and Solids: A Laboratory Characterization of
Copper Sulfide Flotation
Ignacio Molina1, Paulo Martins2 and Álvaro Videla1
1. Pontificia Universidad Católica de Chile
2. The Dow Chemical Company, Brazil

ABSTRACT
Although there is a general consensus regarding the significant impact of frother type for mineral
flotation performance, there is still a lack of research in the area. It is well accepted that frother,
water quality, collector type and ore mineralogy interact on the particle surface to impact the
flotation performance. The interactions are complex and perhaps are the main reason for the lack of
research in comparison to the development observed for collectors. The present work looks to
improve the understanding of how common frothers influence froth quality as part of a complex
system during the copper sulphide flotation process. To do so, the level of interaction between
frother, collector and particle mineralogy are evaluated in a flotation cell specifically designed to
measure the effect in froth depth, collapse time and bubble size distribution.
During the experience MIBC, DOWFROTHTM 1012, DOWFROTHTM 250 and DOWFROTHTM 400
have been tested. The interaction has been evaluated in a frother-water system, a frother-water-
Xanthate system at two levels of concentration (high and low), and a frother-synthetic sea water
system specially prepared for this purpose. Additionally, the results are compared with 0%, 5%
and 10% solids to observe the effect of solids-frother-collector interaction. The preliminary results
are presented for this set of experiments and show a complex interaction due to frother types,
collector dosage, water quality type and solids that deserves further research.

1
INTRODUCTION
Frothers and froth quality has a strong impact on flotation performance. Although important, few
studies have been done to date to explore the effects of the complex interactions between frothers
and operative variables. In this paper, we intend to present the preliminary results on the
establishment of experimental capabilities in order to relate frother type and frother quality with
flotation performance in consideration of their interaction with collector dosage, water quality and
solids surface characteristics.
To do so, an Amira flotation cell type has been built to run experiments under controlled
conditions. Initial results are promising, and an on-going research project is under development to
characterize the frother performance and its impact on metallurgical recovery.

METHODOLOGY
A flotation column has been constructed to run the experiments. The flotation cell is a square-
shaped acrylic column 1.57 cm high with an empty chamber at the bottom to build enough air
pressure before the air passes through a porous glass material with pores 100-160 microns size.
Figure 1 shows a picture and diagram of the cell.

Figure 1 Amira Column. Left side shows a picture under operation. Right side shows a diagram

For each experiment the flotation cell was cleaned and later filled with 6 litres of tap water,
previously mixed with the respective dosage of frothers and collectors. Air was injected at the
bottom of the cell and a rota-meter located in between the cell air-feed and the air compressor
controlled the flow rate. Before collecting any data, the cell was run for at least 45 minutes or until a
stable froth column was achieved. The set of results presented herein were all evaluated under

2
same pH and air flow conditions, 7.0 and 2 L/m respectively. The former value assumed a Jg value
of around 1.5 cm/s.
Frothers under evaluation were MIBC, DOWFROTHTM 250, DOWFROTHTM 1012, and
DOWFROTHTM 400 at a dosage of 20 gr/ton. Each one was evaluated under the following
conditions:
 Using tap water only
 Using tap water and 30 gr/ton Isopropyl Xanthate (ISX)
 Using tap water and 50 gr/ton Isopropyl Xanthate (ISX)
 Using tap water with 5% solids
 Using tap water with 10% solids
 Using synthetic seawater
Each test was run between three and five times depending on variability. For each test, once
equilibrium has been achieved, the froth height and collapsing time was measured. A copper
sulphide ore ground to 100% <100 mesh was used as solid material.

RESULTS AND DISCUSSION

Figure 2 shows a comparison in the froth height obtained for each frother studied after the addition
of 20 gr/ton of MIBC, DOWFROTHTM 250, DOWFROTHTM 1012 and DOWFROTHTM 400 in tap
water. Under these conditions, the MIBC alcohol achieves a lower froth height than all three other
polyglycols. Also, it is interesting to observe that MIBC performance does not change significantly
at any ISX dosage level. In contrast, the froth height of DOWFROTHTM 1012 and DOWFROTHTM
400 decreases as a function of ISX dosage until achieving levels similar to than MIBC.
DOWFROTHTM 250 shows an opposite response, with froth height increasing at higher SXI dosage
than the reference case.
Figure 3 shows the impact of ions in solution. A comparison of the froth height is shown as
obtained from dissolution of frother in tap water and frother in synthetic seawater. As observed
from the plot, the presence of ions in water does not affect MIBC, but has a significant impact in the
polyglycols. All of them show a higher froth height.

3
 Figure 2 Froth height as function of frother and collector dosage

Figure 3 Froth height as function of water quality

4
Figure 4 shows the effect of solids on the froth behavior. The presence of solid particles increases
the froth height in the MIBC, DOWFROTHTM 250 and DOWFROTHTM 400 cases in comparison to
frother dissolved in tap water only. In contrast, the presence of same solid particles decreases the
froth height in the case of DOWFROTHTM 1012. It is again interesting to observe that MIBC is less
affected than polyglycols by the presence of solids particles in solution. Although variation is
observed, the impact of solids particles is less important than water quality and SXI dosage for
polyglycol frothers. In contrast, the presence of solids particles is more significant in the case of
MIBC.

Figure 4 Froth height as function of solids percentage in solution

Finally, Figure 5 shows the results for all conditions under evaluation. As observed, the MIBC
seems to have a very stable response under the experimental parameters, while polyglycol froth
height varies strongly depending on operative conditions. More research is on-going to
characterize the frother response in terms of bubble size and water recovery.

5
 Figure 5 Summary of data set with froth height as function of all variables

CONCLUSION
The method based on a column cell has proved to distinguish different froth behaviours in terms of
the operational conditions under evaluation. The column has been tested with water and frothers,
Isopropyl Xanthate, synthetic seawater and up to 10% solids showing a detectable difference in
froth quality. Further work is currently on going to relate the characteristics of the froth with the
metallurgical recovery.

REFERENCES
Castro S., Miranda, C., Toledo P. and Laskowski, J.S., (2013). Effect of frothers on bubble
coalescence and foaming in electrolyte solutions and seawater. Int. J. Miner. Process., vol. 124, 8-14.
Cho Y.S., Laskowski, J.S., (2002). Effect of flotation frothers on bubble size and foam stability. Int. J.
Miner. Process., vol. 64, 69-80.
Allendes, M., (2009). “Implementación de una metodología para selección de espumantes en
flotación”, Departamento de Química, Universidad de Magallanes.

6
Inferring Gangue Entrainment from Conductivity
Measurements at the Top of the Froth
Rodrigo Araya1, Dalton Major1, Brigitte Lacouture1, César Gómez2 and Kristian Waters2
1. Red Dog Mine, Teck, USA
2. McGill University, Canada

ABSTRACT
The separation of mineral particles by flotation is accomplished in machines operated to create a
collection (or pulp) and a cleaning (or froth) zone. The efficiency of the process is defined by the
recoveries achieved in these zones. The performance in the cleaning zone has been associated with
the volumetric rate of water in the concentrate stream overflowing the unit, which contains
valuable mineral particles captured on the surface of bubbles, and also gangue particles
hydraulically entrained in the water carried by the bubbles. An effective operation requires
establishing a concentrate stream with a water flow rate that involves a compromise: as larger
values result in high recoveries but low grades; while smaller values produce high grades at low
recoveries. A correlation between water recovery and entrainment of unwanted mineral particles
has been established previously. Finding the ideal recovery-grade combination at a particular
operation is complicated because there is currently no sensor to monitor on-line the water flow rate
in the concentrate stream. One alternative demonstrated in laboratory columns, is measurement of
liquid holdup at the top of the froth using a conductivity cell, which correlated with the water
overflowing the column. This communication presents data collected in the columns of the Zn
cleaning stage at the Teck Red Dog Mine in Alaska during the preliminary testing of a short flow
conductivity cell designed at McGill University. The results showed linear correlations between the
top of the froth conductivity and the fraction of silica in the concentrate, which provides a
continuous signal that can be used to complement an on-line stream analyzer’s (OSA) periodic
readings.

There is no full article associated with this abstract.

1
CHAPTER 3

Concentration Plant
Management and
Operation
A Process Mineralogy Approach to Gravity
Concentration of Tantalum Bearing Minerals
Yousef Ghorbani, Rob Fitzpatrick, Melanie Kinchington and Gavyn Rollinson
University of Exeter, UK

ABSTRACT

This study aims to characterise the nature of tantalum mineralogy of the Penouta Balsa Grande
tailings samples using QEMSCAN and XRF. The historic Penouta mine in north-west Spain is the
focus of efforts to extract tantalum from tin waste. This paper further identifies any other phases of
interest and determines the most effective method for extraction using gravity separation. The
gravity concentrates obtained through these tests were analysed using QEMSCAN and EMPA.
Tantalum was found to be present predominantly in the solid-solution columbite-tantalite mineral
group, along with minor Ta present as microlite and as impurities within cassiterite. Following
characterisation of the sample material to identify the key Ta-bearing mineral phases and assess
liberation, a series of gravity separation trials were conducted using Heavy Liquid Separation
(HLS), Mozley table and Knelson concentrator separation. Mass balance calculations were carried
out to calculate the contained metal content of the feed material and concentrate products in order
to assess recovery rates for Ta, Sn and Nb across a range of grains sizes. Research into the Penouta
tailings material has found that over 70% of the Ta is contained within the -125 µm fraction, with
the Ta-bearing minerals tantalite and microlite being closely associated with quartz. Mozley table
separation resulted in recoveries of 89% Ta, 82% Sn and 85% Nb for the -125 µm fraction. The
Knelson Concentrator trial was carried out on the -625 µm size fraction, thereby eliminating the low
grade coarser fraction. Size analysis of the recovery rate for each product, shows that the Knelson
concentrator is the most efficient for recovering of -180 µm particles.

1
INTRODUCTION

The historic Penouta mine in north-west Spain is the focus of efforts to extract tantalum from tin
waste. This study aims to characterise the nature of tantalum mineralogy at Penouta, identify any
other phases of interest and determine the most effective method for extraction using gravity
separation. Tantalum was found to be present predominantly in the solid-solution columbite-
tantalite mineral group, along with minor Ta present as microlite and as impurities within
cassiterite. Following characterisation of the sample material to identify the key Ta-bearing mineral
phases and assess liberation, a series of gravity separation trials were conducted using heavy liquid,
Mozley table and Knelson concentrator separation.

EXPERIMENTAL

Ore Sample
The Penouta deposit is located 1.5 km north of the town of Penouta in Orense Province, northwest
Spain (Figure 1). It is located on the northwest fringe of the Iberian Sn-W province, in the Central
Iberian Zone (Mangas & Arribas, 1991) of the Iberian Massif.

Figure 1 Location and geology map of Penouta, within the Iberian Sn-W province (Modified after:
Schermerhorn, 1981; Gonzalo Corral, 1985; Oosterom, 1988; D-Maps (2015); and Anon, 2015)

Gravity concentration

Sample Preparation

Sample Preparation for Heavy Liquid Separation and Shaking Table

A 1 kg representative sample of the Penouta tailings was split into two samples using a Jones riffle
Splitter. One sample of 528.3 g was screened on a √2 system, splitting the sample into nine fraction
sizes (Table 1). Sieving was undertaken for 25 minutes using a Pascal Sieve Shaker.

2
Table 1 Size fractions acquired through sieving of a representative sample from Penouta

Sieve Size Fraction Sample Size (g) Sample Size (wt. %) Separation Method
+2800 µm 4.0 0.76 N/A
-2800/+1400 µm 66.1 12.51
-1400/+1000 µm 75.8 14.35
-1000/+710 μm 61.8 11.70
-710/+500 μm 65.3 12.36 Heavy Liquid Separation
-500/+355 μm 62.6 11.85
-355/+250 μm 57.6 10.90
-250/+180 μm 46.0 8.71
-180/+125 μm 39.4 7.46
Mozley Table
-125 μm 49.7 9.41

The size fraction +2800 µm was not subjected to testing due to insufficient amounts of material and
the presence of clay in the sample. The size fractions -180+125 µm and -125 µm were not subjected
to heavy liquid separation due to the fine grained nature of the material. Instead, these size
fractions underwent separation by Mozley table.
Sample Preparation for Knelson Concentrator
Approximately 20 kg of bulk material from Penouta was screened at 625 µm to obtain 10 kg of
sample at -625 µm for centrifugal separation. The coarser fraction was removed because centrifugal
separation is designed for recovery of fine particles. A 10 kg sample was used because Laplante
(2000) states that for feed masses <10 kg an overestimation of gravity recoverable content can occur.
A bulk separation trial was undertaken using a Knelson 3” laboratory concentrator. A 10 kg sample
of -625 μm was prepared as the feed sample. The test was run under the following conditions: feed
rate of 0.49 kg/hr, water pressure of 6 PSI (40 kpa) and a g force of 60 (chosen as this is the industry
standard g force used in working plants). The test conditions were chosen based on the test
conditions used by Laplante (2000) during separation trials for gold. Six products were produced
from the trial: a concentrate and five tailings products. The tailings products were collected over set
time periods (Table 2). A feed rate of 0.49 kg/hr indicates a run time of two minutes per 1 kg of feed
material.

Table 2 Tailings collection times for Knelson concentrator trial, based on movement of feed
through the concentrator

Tailing Product Time Feed

1 0-2 mins 1 kg
2 2-6 mins 2 kg
3 6-10 mins 2 kg
4 10-14 mins 2 kg
14-20 Mins
5 3 kg

Following the trial, each product was split into three samples using a Jones riffle Splitter. One half
of the products were screened into eight size fractions using a √2 sieve system and analysed for Ta,
Sn and Nb using a portable XRF. The other half of the products were split repeatedly until
approximately 50-60 g were obtained; this split was sent for analysis by semi-Quantitative XRF.

3
Mineralogical analysis

Quantitative Evaluation of Minerals by Scanning Electron Microscopy (QEMSCAN®)

Quantitative mineralogical analyses were conducted on the samples using a QEMSCAN® 4300
which is based on a Zeiss EVO 50 series SEM and consists of four light elements: Bruker SDD
(Silicon Drift Droplet), Energy Dispersive X-ray Spectrometers (EDS) and an electron backscatter
detector (see Rollinson et al. 2011 for more details). Both the fieldscan and PMA measurement
modes were used on the samples depending on the sample particle sizes (see Gottlieb et al. 2000,
Pirrie et al. 2011, Goodall and Scales 2007). iMeasure v. 4.2 was used for data acquisition and
iDiscover v. 4.2 & 4.3 was used for the data processing. For this study the LCU5 SIP provided with
the QEMSCAN® system (containing common minerals) was modified as required to match the
samples. The QEMSCAN unit used in this study was located at the Camborne School of Mines
(CSM), University of Exeter, UK.
Quantitative X-ray Diffraction (QXRD)
Qualitative mineralogical analysis was also conducted to determine the amount of the different
major mineral phases using a Siemens D5000 XRD on powdered samples (<50 microns).
Measurements were performed at room temperature with Cu Ka radiation, 40 kV and 30 mA, over
2Θ angles of 2–70°, using 0.02° steps at 1 s, taking approximately 60 min per sample. Samples were
rotated during measurement to minimize effects of preferential orientation. The XRD profiles were
interpreted using the JCPDS PDF-2 2004 database.
X-ray Fluorescence (XRF)
Bulk chemical analyses of the samples were carried out using a Bruker S4 Pioneer WD-XRF
instrument. Samples were ground to a powder less than 50 microns and then prepared into pressed
discs, and analysed using semi-quantitative mode which is calibrated to internal standards. The
results from QEMSCAN were validated by comparison with XRF data and QXRD data.
Portable x-ray fluorescence analysis (pXRF) was carried out in order to provide a quick analysis of
samples and to help to determine the next course of action using a DP-6000C Portable XRF on the
different sized fraction samples of gravity tests products (Figure 2). The pXRF is calibrated with a
known standard (316 stainless steel disk) in a pass or fail test. Any peak drift during operation was
monitored with a known standard reference material (NIST 2710A). Measurements were taken at
the beginning of the sample run. All samples were analysed using the geochemistry setting, with
two beams; each beam was set to analyse for 45 seconds giving a total analysis run time of 1 minute,
30 seconds. Each sample was analysed three times, through the sample bag, with the sample being
moved between each analysis. The mean average and standard deviation of the three analyses were
then calculated. It worth noting that the direct contact of pXRF on powder samples is a suitable
method to increase the accuracy of the measurement.

4
 Figure 2 Analysis of separator products by handheld XRF.

Electron Microprobe Analysis (EMPA)

Detailed elemental analyses of minerals were carried out using a JEOL JXA-8200 EPMA. The EPMA
analyses were carried out using a 10 nA electron beam accelerated to 20 kV and focused to a 5
microns beam diameter using wavelength dispersive X-ray spectrometers only. The results were
quantified with reference to pure metal and natural oxide standards using a ZAF matrix correction
routine.

RESULTS AND DISCUSSIONS

Heavy Liquid Separation

Microscopy and QEMSCAN results showed that heavy liquid separation has worked well for
removing gangue material (except in the case of the mica phases). As presented in the Tables 3, 4
and 5, mass balance calculations show that Ta, Sn and Nb are concentrated in the finer fractions.
Comparison of this study to the work done by Montero-Ríos (2012) show that both studies have
found that the majority of contained metal is located in the finest size fractions: Montero-Ríos (2012)
found that 83% of Ta and Sn are concentrated in the -630 µm fraction, compared to 90% Ta and 62%
Sn in the -500 µm calculated from the mass balance.

5
 Table 3 Summary table of recovery of Ta by size fraction

Weight % of Size Size Fraction Weight Calculated Feed Recovery to

Size Fraction (µm) Distribution (%)
Fraction Sink % Float % Grade (ppm) sink(%)

+2800 0.76 0.38 98.49 1.53 0.03 100.00

-2800 12.51 0.38 98.49 1.53 0.43 100.00

-1400 14.35 0.26 98.68 1.07 0.34 100.00

-1000 11.70 0.31 98.59 11.21 2.94 11.09

-710 12.36 0.41 98.38 21.59 5.99 7.75

-500 11.85 2.40 95.69 17.10 4.55 100.00

-355 10.90 1.91 95.66 23.51 5.75 58.29

-250 8.71 1.09 95.87 7.85 1.53 100.00

-180 7.46 6.52 92.23 32.78 5.48 71.51

-125 9.41 4.73 94.47 345.69 72.96 88.98

Total 100.00 100.00

Table 4 Summary table of the recovery of Sn by size fraction

Weight % of Size Size Fraction Weight Calculated Feed Distribution Recovery to

Size Fraction (µm)
Fraction Sink % Float % Grade (ppm) (%) sink (%)
+2800 0.76 0.38 98.49 483.67 0.77 22.15

-2800 12.51 0.38 98.49 483.67 12.68 22.15

-1400 14.35 0.26 98.68 343.95 10.34 21.71

-1000 11.70 0.31 98.59 326.30 8.00 26.68

-710 12.36 0.41 98.38 256.61 6.65 45.67

-500 11.85 2.40 95.69 185.49 4.61 42.15

-355 10.90 1.91 95.66 199.89 4.57 31.33

-250 8.71 1.09 95.87 203.97 3.72 17.59

-180 7.46 6.52 92.23 292.71 4.57 36.18

-125 9.41 4.73 94.47 2236.84 44.10 82.25

Total 100.00 100.00

6
 Table 5 Recovery of Nb by size fraction

Size Fraction Weight

Weight % of Size Calculated Feed Distribution Recovery to
Size Fraction (µm)
Fraction Sink % Float % Grade (ppm) (%) sink(%)

+2800 0.76 0.38 98.49 62.30 0.54 2.46

-2800 12.51 0.38 98.49 62.30 8.88 2.46

-1400 14.35 0.26 98.68 66.89 10.93 1.59

-1000 11.70 0.31 98.59 62.05 8.27 2.00

-710 12.36 0.41 98.38 50.47 7.11 3.32

-500 11.85 2.40 95.69 61.01 8.24 36.04

-355 10.90 1.91 95.66 56.83 7.06 31.00

-250 8.71 1.09 95.87 49.64 4.92 20.33

-180 7.46 6.52 92.23 67.21 5.71 44.41

-125 9.41 4.73 94.47 357.83 38.35 84.83

Total 100.00 100.00

Knelson Concentrator
The Knelson Concentrator trials were carried out on the -625 µm size fraction, thereby eliminating
the low grade coarser fraction. The products were analysed by SQ-XRF and the metal grades used
in a series of mass balance calculations. The Knelson achieved good recovery rates in the first
minute, possibly due to the build-up of the particle bed within the concentrator cone. For both Ta
and Nb recovery was fairly constant before dropping off in the final stage. Sn recovery however
increased during the final stages. No mineralogical work was done on the products of this trial
however the chemistry of the products is dominated by SiO2, Al2O3 and K2O indicating that there is
still a large proportion of silicate minerals within the concentrate. The drop in recovery is believed
to be due to the replacement of lower density quartz particles (containing unliberated columbite-
tantalite and cassiterite) with higher density mica grains in the concentrate bed.
Size analysis of the recovery rate for each product, shows that the Knelson concentrator is the most
efficient for recovery of -180 µm particles. This may be due to the way the particles settle within the
concentrate bed. If smaller particles become trapped between larger particles they are less likely to
be replaced by later material entering the concentrate cone. Comparison of recovery rates from this
study to previous work by Montero-Ríos (2012) reveals very similar recovery rates (Table 6).

Table 6 Comparison of results of centrifugal separation

Product Recovery Recovery (Montero-Ríos, 2012)

Ta 63.6% 61.9%
Sn 40.0% 46.4%

The work carried out by Montero-Ríos (2012) only tested material -100 µm in size. For the recovery
to show no significant change, despite the change in particle size (material -625 µm was used in this
study), it is encouraging and it indicates that centrifuge separation could potentially be used to treat

7
a wider variety of grain sizes than just the fines, with no significant loss to in fine size fractions. The
products from the Knelson Concentrator trial were split into two; one half of the products were sent
for semi-quantitative XRF analysis of the whole product, the other half was screened into different
size fractions and analysed by portable XRF. The semi-quantitative results for the whole samples
(Table 7) show little variation in major element chemistry between the different tailings products
and the feed. However the tailings are depleted in Ta 2O5, Nb2O5 and SnO2 when compared to the
feed material.
These elements have been upgraded in the concentrate; Ta2O5 has upgraded from 55 ppm to 4320
ppm (a factor of 79), Nb2O5 has upgraded from 104 ppm to 4545 ppm (a factor of 44) and SnO 2 has
increased from 661 ppm to 2.9 % (also a factor of 44). TiO2, Cr2O3, MnO, Fe2O3, ZnO, As2O3, Y2O3
and ThO2 have also upgraded in the concentrate.
Quantification of the chemistry of the columbite-tantalite series
EMPA study of the columbite-tantalite grains allows accurate quantification of the chemistry of the
columbite-tantalite series. This is a broad series of minerals so knowing the accurate chemistry
would help:
a. Aids processing.
b. Determines whether the mineralogy is economic.
c. Allows an understanding of mineral paragenesis.
d. Increases understanding of distribution of Ta and Nb within minerals.
EMPA results are variable regarding both accuracy and reproducibility, 141 points were analysed
by spot analysis across 67 grains from one carbon coated polished block: FC1 (-1000 µm
concentrate). Of those, 60 spot analyses fall within ±1.5% of the expected 100% element total and are
summarised in Table 8. For the remaining 81 points, there is no clear reason for the low totals and,
as a result, these points have been excluded from further analysis. Standard deviations show the
variation on chemistry for each mineral. Low standard deviations for microlite indicate chemistry is
consistent across microlite grains. Cassiterite also has consistent chemistry except for SnO2 and
Ta2O5, where standard deviations indicate variable chemistry. This is most likely due to substitution
of Ta for Sn within the cassiterite mineral structure. Columbite and tantalite both show highly
variable chemistry (particularly regarding Ta2O5 and Nb2O5 content) which is expected as these
minerals are part of a solid solution series, where chemistry is spread between 0-86% for both Ta
and Nb. One point analysis returned results of 50% Ta2O5 with 36% SnO2 and only 4% Na2O5. Based
on the high SnO2 content it is possible this grain may be ixiolite or wodginite, however the
chemistry doesn’t quite fit for either mineral.

Table 7 Semi-quantitative XRF results for the Knelson Concentrator products

Sample Name Start Feed End Feed Tails 1 Tails 2 Tails 3 Tails 4 Tails 5 Conc.

CO2 2.04 2.08 2.09 1.98 2.02 2.04 2.04 2.05

Na2O 1.60 1.75 1.69 1.63 1.69 1.70 1.67 1.34

MgO 0.20 0.19 0.20 0.21 0.19 0.19 0.20 0.22

Al2O3 16.40 16.90 17.30 16.80 16.80 16.90 17.00 12.20

SiO2 71.09 70.60 69.96 70.67 70.78 70.61 70.40 71.35

P2O5 0.01 0.01 0.01 0.02 0.02 0.01 0.02 0.05

SO3 0.11 0.10 0.10 0.11 0.10 0.09 0.11 0.08

8
 K2 O 6.52 6.42 6.71 6.63 6.51 6.56 6.62 4.57

CaO 0.10 0.10 0.11 0.11 0.11 0.11 0.11 0.11

TiO2 0.12 0.11 0.12 0.12 0.11 0.11 0.12 0.22

Cr2O3 0.05 0.04 0.04 0.04 0.04 0.04 0.04 0.09

MnO 0.10 0.10 0.08 0.09 0.09 0.09 0.09 1.14

Fe2O3 1.38 1.33 1.32 1.32 1.32 1.32 1.34 2.19

ZnO 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.05

Ga2O3 0.0050 0.0050 0.0060 0.0050 0.0050 0.0050 0.0050 0.0024

As2O3 0.0030 0.0030 0.0030 0.0030 0.0020 0.0030 0.0030 0.0138

Rb2O 0.1705 0.1678 0.1741 0.1719 0.1702 0.1679 0.1713 0.1183

SrO 0.0030 0.0030 0.0030 0.0030 0.0030 0.0030 0.0030 <LOD

Y2O3 <LOD <LOD <LOD <LOD <LOD <LOD <LOD 0.0231

ZrO2 0.0106 0.0140 0.0095 0.0100 0.0109 0.0107 0.0111 0.2952

Nb2O5 0.0096 0.0111 0.0059 0.0070 0.0070 0.0066 0.0073 0.4545

SnO2 0.0581 0.0740 0.0366 0.0392 0.0421 0.0384 0.0372 2.9120

Ta2O5 0.0060 0.0050 0.0010 0.0020 0.0020 0.0020 0.0030 0.4320

ThO2 <LOD <LOD <LOD <LOD <LOD <LOD <LOD 0.0116

Total 100.00 100.02 99.99 99.96 100.03 100.03 99.99 99.92

Table 8 Summary table of EMPA results for point analysis

Mineral Na2O F Ta2O5 Nb2O5 MgO SnO2 MnO CaO FeO TiO2 Totals No. of Samples

Mangano- Mean b.d. b.d. 20.25 59.91 0.07 0.07 15.00 b.d. 3.84 0.12 99.14
13
columbite
Std. Dev - - 5.63 5.26 0.00 0.01 0.98 - 1.11 0.03 0.51

Mangano- Mean b.d. b.d. 50.23 32.11 b.d. 0.27 13.41 b.d. 3.01 0.13 99.03
4
tantalite
Std. Dev - - 8.86 8.67 - 0.11 1.03 - 1.26 0.03 0.35

Mean 5.98 3.96 72.56 4.35 b.d. 1.50 0.12 10.65 b.d. 0.17 99.21
Microlite 3
Std. Dev 0.07 0.16 0.15 0.12 - 0.03 0.00 0.07 - 0.02 0.34

Mean 0.12 0.24 2.17 0.58 0.07 98.32 0.15 b.d. 0.39 0.12 100.76
Cassiterite 39
Std. Dev 0.01 0.04 1.79 0.31 0.01 2.70 0.07 - 0.27 0.09 0.58
Other Mineralogy:
Ixiolite or Wodginite? b.d. b.d. 50.34 3.77 b.d. 36.31 5.74 b.d. 4.49 0.09 100.76 1

a. Columbite-Tantalite Grains
Quantitative analysis by WDS provided data on the chemical composition of the columbite-tantalite
grains. The Ta:Nb and Fe:Mn oxide ratios were calculated for each point and plotted on a
quadrilateral plot of mineral composition (Figure 3). The quadrilateral plot shows a bimodal
distribution, that spans the full Mn-end of the columbite mineral series. Each analysed point was
assigned a mineral name according to its Ta:Nb and Fe:Mn ratios. A Ta:Nb ratio >0.5 is classed as
tantalite. An Fe:Mn ratio >0.5 is classed as Mn-rich, and assigned the prefix mangano- (as opposed
to the Fe-rich end members which are assigned the prefix ferro-). Of the 76 points analysed across

9
32 columbite-tantalite grains, only 18 fell within the acceptable ±1.5% of 100% range. 13 of those 18
points were identified as mangano-columbite and five as mangano-tantalite. Only three analysed
points were identified as belonging to the Fe-rich end of the series and have been classed as ferro-
tantalite; however these points all had low totals so have been disregarded from analysis. The
EMPA analysis also allowed the opportunity to observe mineral textures and changes in
composition within single grains. As Figure 4 shows, the columbite-tantalite grains are highly
zoned, commonly with Nb-rich cores moving to Ta-rich rims. It must be noted that the point
analysis was carried out across these zoned minerals with several analyses per grain. Thus, any
bimodal distribution previously commented on is within the grains rather than between the grains.

Columbite-Tantalite
Ferro-tapiolite Mangano-tantalite
1,00
Ta/(Ta+Nb)

0,50

0,00
0,00 0,50 1,00
Ferro-columbite Mn/(Mn+Fe) Mangano-columbite

Points within 1.5% of 100% total

Figure 3 Quadrilateral plot of columbite-tantalite Figure 4 Columbite-tantalite grains analysed by EMPA, show
elemental composition. Plotted based on oxide weight per clear elemental zoning from Nb-rich centres to Ta-rich rims
cent rather than element weight percent

b. Microlite Grains
Three points, analysed across one grain were identified as microlite. Microlite is a sub-group of
minerals within the pyrochlore supergroup. The presence of F within the mineral grains (Table 8)
suggests this is the fluorcalciomicrolite species of the microlite group.

c. Cassiterite Grains
39 point analyses across 27 grains were identified as cassiterite. SnO2 content within the cassiterite
grains was variable from 91-100% (full chemical results are available in Appendix C). For grains
that reported low SnO2 content the totals were still within the acceptable ±1.5% of 100% limit.
Analysis of the data revealed that SnO2 was being substituted for by Ta2O5 (Figure 5). Ta2O5 within
cassiterite was observed in two forms: as inclusions of Ta-bearing minerals (Figure 6) and as
substitution within the cassiterite crystal structure.

10
 8
Ta2O5 Nb2O5 FeO MnO
6
Impurities wt.%

0
90 92 94 96 98 100 102
SnO wt.%
Figure 6 Cassiterite grains analysed by EMPA, show Ta-rich
Figure 5 Elemental impurities within compositions
cassiterite

CONCLUSION
The aim of this project was to determine the extractability of tantalum from the tailings material at
the historic Penouta Mine, northwest Spain, using gravity separation methods. Mangano-tantalite
and mangano-columbite are the main tantalum hosting phases in the feed material. These minerals
occur in solid solution together. Minor tantalum is present in the pyrochlore group mineral,
microlite. The tailings material contains 200-500 ppm Sn in cassiterite. The cassiterite averages 50-
150 µm in size and is well liberated in the +500 µm size fraction.
Mineralogy is dominated by silicate phases, particularly quartz and muscovite, which average 20-
225 µm in size across the samples analysed. All ore minerals are most abundant in the -125 µm size
fraction (73% Ta, 44% Sn and 38% Nb), which accounts for 9% of the feed material. Less than 30% of
the ore mineralogy is well liberated, although liberation increases as grain size decreases. Ore
minerals are closely associated with quartz and muscovite.
Assess recovery rates for each gravity separation technique indicated that the heavy liquid
separation worked best for +1000µm size fractions. Overall recovery of 51% Ta, 28% Sn and 12% Nb
was achieved. The Mozley table separation worked best for the -125 µm size fraction, achieving
overall recoveries of 88% Ta, 78% Sn and 80% Nb. The Knelson concentrator separation achieved
overall recoveries of 64% Ta, 40% Sn and 37% Nb. Recoveries were highest during the first minute
as the concentrating bed built up. The -180 µm size fraction responded but to centrifugal
separation.
For the optimum processing route the material should be screened to 1000 µm to remove low grade
material. Coarse material should be ground to 100 µm to liberate trapped ore minerals. Mozley
table separation is recommended as the most efficient separation process. Two stages (rougher and
cleaner) are recommended to remove gangue and upgrade the concentrate. Ta2O5 needs to be at
least 30% of the concentrate to be economic.

11
REFERENCES
Anon (2015) Chemical bonding and crystal chemistry [online]. Available from:
http://umanitoba.ca/faculties/science/geological_sciences/faculty/sherriff/website/277/05/Lecture%2
03%20Crystal%20chemistry.05.htm [Accessed 31st August 2015].
D-Maps (2015) espagne 30, galice 22 and galice 34 [online]. Available from: http://d-
maps.com/continent.php?num_con=17&lang=en [Accessed 29th August 2015].
Gonzalo Corral, F.J. & Gracia Plaza, A.S. (1985) Yacimientos de estaño del oeste de España: Ensayo
de caracterización y clasificación económicas [online]. Available from:
http://ruc.udc.es/bitstream/2183/5927/1/CA-9-16.pdf [Accessed 16th February 2015].
Laplante, A.R. (2000) A standardized test to determine gravity recoverable gold [online] Available
from: http://knelsongravity.xplorex.com/page449.htm [Accessed 2nd September 2015].
Mangas, J. & Arribas, A. (1991) Fluid inclusion study of tin-mineralized greisens and quartz veins
in the Penouta apogranite (Orense, Spain). Mineralogical Magazine, 55, pp.211-223.
Montero-Ríos, B.M. (2012) Proyecto de viabilidad de una planta de tratamiento para mineral de
estaño y tántalo [online]. Thesis, School of Mining Engineers, Polytechnic University of Madrid.
Available from: http://oa.upm.es/15454/1/PFC_Borja_Maortua_Montero_Rios.pdf [Accessed 16th
February 2015].
Oosterom, M.G. (1988) Europe (Iberian Peninsula): The geochemistry of granitiod-related deposits
of tin & tungsten in orogenic belts. In: Hutchison, C.S. ed. Geology of tin deposits in Asia & the
Pacific: selected papers from the Internation Symposium on the geology of tin deposits. Nanning,
China, October 26-30, 1984: United Nations, New York, pp.187-198.
Pirrie, D. & Rollinson, G.K. (2011). Unlocking the applications of automated mineral analysis.
Geology Today, 27(6), pp.235-244.
Rollinson, G.K., Stickland, R.J., Andersen, J.C., Fairhurst, R. & Boni, M. (2011). Characterisation of
Supergene Non-Sulphide Zinc Deposits using QEMSCAN®. Minerals Engineering, 24, pp.778-787.
Schermerhorn, L.J.G. (1981) Framework and evolution of Hercynian mineralization in the Iberian
Meseta. Leidse Geologische Mededelingen, 52(1), pp.23-56.

12
Metal Accounting: The Core Responsibility of
Process Engineers
Lorenzo Cappai1, Manuel González2, Stephane Brochot1 and Pierre Vix3
1. Caspeo, France
2. Caspeo, Chile
3. Glencore, Switzerland

ABSTRACT
Metal accounting is one of the main tools for financial and technical management of metal
production industry. It is based on measurements and has to manage the uncertainty inherent to
the measurement process generating financial risk. In 2005, the AMIRA P754 project established a
code of best practice for implementing efficient and reliable metal accounting systems.
The application of the AMIRA code is not only linked to the use of given software whatever
qualities it may get or functionalities it may provide. It must be underlined that implementing an
accurate metal accounting system, first relies on the involvement of the whole company, including
the top management, the financial division, the quality assurance team and the production and
laboratory teams, all actors being ready to participate at different levels to the implementation.
The starting point of an efficient metal accounting system is a precise understanding of the process
operations, including the follow-up procedures, based on sampling and measurements. A data
reconciliation algorithm allows obtaining coherent material balance from raw data, improving the
accuracy of data by a statistical distribution of measurement errors, and allowing data quality
checking.
Citing the AMIRA code, “Accounting results must be made available timeously, to meet
operational reporting needs, including the provision of information for other management
information systems, and to facilitate corrective action or investigation”. A software solution, such
as INVENTEO, automates all the repetitive tasks, ensures traceability and workflow and rapidly
provides reports.
The paper presents first the data reconciliation algorithm and highlights the need of measurement
errors estimation. Then, the main steps and actions required for the implementation of an accurate
metal accounting system are listed and illustrated with various cases. Examples such as stockpile
management, concentrator, refinery or smelter balances, demonstrate how to address specific
needs.

1
INTRODUCTION
Metal accounting estimates the quantity of saleable metals produced over a defined period of time
(Morrison, 2008a). It is widely used to quantify the performances of production plants and to
estimate as accurately as possible the metal inventory. Precise information is critical to decrease the
financial risk linked to a wrong estimate of the stocks and to quickly identify losses and process
issues. In this perspective, the AMIRA P754 project established a code of best practice for
implementing efficient and reliable metal accounting systems. Such systems often use a statistical
data reconciliation algorithm to obtain coherent material balance from raw data and improve the
accuracy of data by a statistical distribution of measurement errors. Different software solutions are
currently proposed to help in fulfilling compliance with the AMIRA code. Nevertheless, the use of a
given software alone, whatever qualities it may get or functionalities it may provide, is not
sufficient to apply completely the code of best practice. In fact, the implementation and the
maintenance of a reliable metal accounting system is first of all a project that involves the whole
company, from the top management to the laboratory team, and each actor shares the responsibility
for the project to succeed. Amongst others, the role of the process engineer is particularly important
due to his functions linked to the process management on a daily basis and his precise
understanding of the process operations. His proximity to the process and his know-how allow him
a very strong contribution in setting up the metallurgical accounting system, in maintaining it up-
to-date as the process evolves with time, and last but not least, in using the metal accounting results
to adjust the production in order to optimize it.
The paper presents first the data reconciliation algorithm and how to improve its results,
highlighting the need for measurement errors estimation. Then, the main steps and actions required
for the implementation of an accurate metal accounting system are listed and illustrated with
various cases. For each step, the main responsibilities associated to the process engineer will be
highlighted. Real cases managed with INVENTEO, a metallurgical accounting solution, such as
stockpile management, concentrator, refinery or smelter balances, demonstrate how specific needs
can be addressed.

METHODOLOGY
In metal accounting, statistical data reconciliation is often considered as necessary since data are
incoherent regarding the material conservation laws (Hodouin & Everell, 1980; Herbst, Metha &
Pate, 1988; Fourniguet et al, 1997). To appreciate such incoherence, it is necessary to perform
enough measurements, which is to say to have a redundant system. Data redundancy means
having more data than the minimum required to calculate the material balance. In fact, when we
consider a whole plant or a single operation unit, the sum of the measured masses of input streams
will never be exactly equal to the sum of the measured masses of output streams, neither for the
global mass nor for single masses of metals, even in steady state conditions. This is because
measurements of masses, moisture and metal contents are random processes, and therefore subject
to uncertainty. Such uncertainty can be evaluated for every measurement by its associated
measurement error (Morrison, 2008b). The overall measurement error includes many components
that can be divided, following the Pierre Gy’s classification (Gy, 1979), into two main ones: the total
sampling error and the analytical error. The analysis of the uncertainty budget of the overall
measurement error highlights the main components on which the efforts of improvement can be
done (Brochot, 2011). The data reconciliation algorithm hence finds a set of estimated values so that

2
they are as close as possible to the measured values according to their errors, and that they verify
the material conservation laws (Brochot & Durance, 2012). The information redundancy allows
delivering a set of coherent estimates more accurate than the initial measurements (Ragot et al,
1990; Ragot & Maquin, 2006). Therefore, data redundancy and the related measurement errors
estimation are the necessary starting point for a data reconciliation that enables to improve the
knowledge of the process (Durance, Brochot & Mugabi, 2004). Moreover, the accuracy of data
reconciliation results increases with the level of detail and the quantity of information used to
calculate the material balance. When sufficient data redundancy is available it is hence interesting
to take into account intermediate process streams (Lyman, 2005), to account for the wet mass and
related moisture content instead of the calculated dry mass alone, to include non-saleable
components in the material balance calculation and operate metallurgical accounting also over
periods of time longer than one month (Brochot & Durance, 2012). The quality of data management,
processing and checking before data reconciliation is indeed of primary importance to avoid the
“garbage in, garbage out” effect.
To perform the metal accounting calculation over a given period, raw data such as masses,
moisture and metal contents obtained by measurements have to be gathered. Once collected, the
total quantity of material can be calculated as the sum of many mass measurements such as truck
loads or production weights per shift, and mean moisture or metal contents as the weighted
average of the contents of many samples. The aggregation of these raw data gives the “basis data”.
Before being fed to the reconciliation algorithm, measurement errors are attached to the basis data.
The accurate estimation of these errors using the error propagation calculation rules is necessary to
correctly eliminate the incoherencies during the statistical reconciliation. The preparation of a basis
data set is a key operation and can be very time-consuming for plants that have numerous streams.
The use of a metallurgical accounting solution such as INVENTEO can answer the need for
automatic data aggregation, providing the traceability from starting raw data to final basis data,
while ensuring the respect of a well-defined procedure for the calculation of every variable and
timely results.
All these considerations must be taken into account during the entire life cycle of a metal
accounting system, from the first implementation of the system to its adjustments due to the
evolution of the production process, and when performing the regular metal accounting operations.
The skills and functions of the process engineer put him in a core position during the entire life
cycle of the metal accounting system. In the life time of a company, different life cycle levels can be
considered for metal accounting (Brochot & Durance, 2012) and the process engineer participates to
all of them. One life cycle concerns the metal accounting system itself, its implementation up to its
validation, the phase of metal accounting reports production, and finally the last reconciliation
model corresponding to the end of the production plant activity.
During the implementation, the process engineer is the one best suited to define the objectives of
the metal accounting and the delimitation of the system, according to the availability of raw data,
which also requires the review of process documentation and of the existing measurement system,
and in case of missing information, choosing the best locations to implement additional points of
measurement. During tests and tuning of the initial reconciliation model, thanks to his proximity to
the process he can study the reconciliation results to check if they are realistic and satisfactory.
After the validation of the metal accounting system, he manages the production of metal
accounting reports. Moreover, he has to identify any possible improvement of the reconciliation
model and evaluate possible enlargements or contractions of the delimitation of the system.

3
Changes may be due to improvements in the measurement system and data management, but also
to process modifications, new raw materials, new products or units of operation. After the
identification of the needs for updating the reconciliation model and their implementation, a new
phase of tests and tuning is necessary to validate the updated version of the reconciliation model.
On the long run, the process engineer may also identify actions to improve altogether the process
efficiency and its accountability, for example modifying the stock management using smaller
parallel stockpiles when possible.
When the choice to stop the production and close the site has been taken, the configuration of the
final reconciliation model is necessary to account for the stock exportation, tailings disposal and
management of the wastes generated by the dismantlement.
During the life cycle linked to the calculation of each metal accounting report, performed in the
frame of the production phase as defined above, the process engineer is responsible for checking
the availability and correctness of raw data, identify exceptional events that could affect the
material balance, and this, very often during the considered period. Some operation units or
measurement points may be down, or a stock may be managed with an uncommon manner due to
an unexpected problem. Again, his proximity to the process is a major advantage. Once all raw data
are available, verified and fixed he has to manage the reconciliation process and generate
intermediate reports before their validation by the management staff. A metallurgical accounting
solution such as INVENTEO can help in flagging mistakes in raw data thanks to automated
diagnostic tools. It can manage the workflow offering a complete traceability from the raw data
extraction, up to their aggregation, including their possible manual adjustment and approbation.
The next section presents a series of real cases encountered during the implementation of
metallurgical accounting systems with INVENTEO.

RESULTS AND DISCUSSION

This section presents several approaches to reach the best results with data reconciliation by
material balance. It is illustrated with real cases encountered during INVENTEO implementations.
As shown, the same principles can be applied to a variety of processes dealing with production of
different metals.

Material balance of the flotation circuit of a gold ore concentrator

In the flotation plants it is usual to lack the measurements concerning the amounts of produced
concentrates and tailings. It is generally the case when concentrate is immediately processed
downstream. Similarly, flowrate and solid percent of tailings are rarely or inaccurately measured.
Very often, such amounts are calculated using the “two products” formula or the “three products”
formula when two concentrates are produced. The formula bases its calculation on the amount of
the material feeding the flotation plant, and the contents of one of the components in feed,
concentrates and tailings. The formula is based on the assumption that there is no accumulation of
material in the plant (the tank levels are constant). When only one concentrate is produced, the
formulae to calculate the amounts of concentrate (1) and tailings (2) are:
(1)

(2)

4
Where F, C and T are the quantities of feed, concentrate and tailings; and f, c and t their
corresponding contents of one of the components. It must be reminded that such procedure is
considered an exception by the AMIRA code, that must be clearly justified in the metal accounting
protocols and used only if direct measurements are lacking. Therefore, the addition of flowmeters
or belt weighers to have better figures must be considered when possible. Nevertheless, when the
formula is used some choices allow to improve the calculation results. First of all, the metal (or
component) used in the formula has to be well defined: it could be one of the metals of economic
interest, but also another major component that can be accurately measured. In fact, a major
component, even if non-saleable, can enhance the estimates. It must be kept in mind that the larger
the measurement errors associated to the contents, the larger the errors associated to the calculated
quantities. Following is an example related to a flotation circuit treating gold ore, where solid
quantities of concentrate and tailings are calculated using the three gold contents and the dry mass
of the feed by means of the “two products” formula. As seen in Table 1, the estimates of the
standard deviations for the gold contents of feed, tailings and concentrate that take into account
sampling and analytical errors highlight a noteworthy uncertainty, a problem frequently
encountered when working with gold contents. Corresponding measurement errors, calculated as
twice the standard deviation, range between 12 and 19 percent.

Table 1 Gold contents of feed, tailings and concentrate and the estimates of their standard
deviations

Relative standard Absolute standard

Stream Au (g/t) deviation (%) deviation (g/t)

Feed 2 6 0.12

Concentrate 29 7 2.03

Tailings 0.4 9.5 0.04

Table 2 shows the standard deviations associated to the calculated quantities of concentrate and
tailings, deduced by means of the error propagation formula. Results are not accounting for the
standard deviation of the feed quantity, and yet, in Table 3 it is possible to observe that the use of
such approach leads to a non-negligible uncertainty in the amount of gold recovered in the
concentrate. The variability coming from the “two products” formula calculation method is
represented by an absolute standard deviation equal to 1.27 kg of gold per day. The amount of gold
recovered in the concentrate is then estimated as 16.22 kg 2.54 kg every day, with very relevant
financial implications. This calculation is assuming that samples are correctly selected and
managed, avoiding biases. If bias affected just a single one of the considered contents, the situation
would be even worse.
It is interesting to study how the “two products” formula can be exploited in data reconciliation, in
which, its result represents only an initialization for the values of masses of concentrates and
tailings, while their most probable values will be calculated thanks to the subsequent reconciliation
by material balance. The simultaneous material balance calculation of many components adds
redundancy to the system and improves the accuracy of the metal accounting (Morrison, Gun &

5
McCallum, 2002; Guerney, Dunglison & Cameron, 2005). In the event that only the component used
in the formula is considered in the material balance calculation, the redundancy is not sufficient,
since there is just enough information to calculate the material balance.

Table 2 Calculated concentrate and tailings quantities and their associated standard deviations
calculated with the error propagation formula

Relative standard Absolute standard

Stream Mass (t/d) deviation (%) deviation (t/d)

Feed 10000

Concentrate 559.4 10.57 59.1

Tailings 9440.6 0.63 59.1

Table 3 Quantities of gold in the concentrate and the tailings and their associated standard
deviations calculated with the error propagation formula

Relative standard Absolute standard

Stream Mass of Au (kg/d) deviation (%) deviation (kg/d)

Feed 20

Concentrate 16.22 7.83 1.27

Tailings 3.78 9.65 0.36

In the flotation circuit presented above there is no redundancy of data. Therefore, improvements of
this node balance are impossible without measuring at least another component for each stream.
These considerations were taken into account during the INVENTEO implementation. Iron, arsenic
and silica were added to the material balance calculation to increase redundancy. Unfortunately,
sulfur contents were not available for all the streams on a regular basis, not allowing a stable
redundancy with respect to this element and hence to consider it for a periodical accounting.
Moreover, since these contents can be measured more accurately than gold ones, silica was used
instead of gold in the “two products” formula for a more accurate initialization. Silica was
preferred to the others because it reaches considerably higher contents, ranging from 23 up to 66
percent of the dry masses, representing alone more than half of feed and tailings dry masses.
Moreover, as fed silica mainly reports to the tailings its contents allow to better describe the feed to
tailings mass ratio. For data reconciliation, all the elements are taken into account. Gold and arsenic
contents will contribute to a better estimate of the concentrate, while silica content will contribute to
a better estimate of the tailings. This example shows the importance of redundancy and of
measurement errors calculation necessary to exploit it, and how they can enable to get the most out
of data even under unfavorable conditions.

6
Stockpile management and batch processes in a copper smelter plant

Smelter plants receive large amounts of concentrates from various sources and the operations of
dispatching and blending require a large number of stocks. Accumulation of material concerns also
WIP (Work in progress) and intermediate stocks of reverts. Stocks are very important concerning
metal accounting as they contain large amounts of metals, especially in the last processing steps of
smelters and refineries. Stocks lock up much value but their estimate are generally less accurate
than the material movements (Wortley, 2008; Holmes, 2008; Connelly, 2009). To add redundancy
and improving results when using data reconciliation, it is interesting to account each stock
individually through a specific devoted node in the material flow diagram. One or more input
streams can feed the stock and one or more output streams can reclaim from it. The quantity and
the quality of the stock are accounted for at the beginning (opening) and at the end (closing) of the
considered period. The beginning inventory can be seen as the first input stream of the period and
the ending inventory as the last output, the ending inventory becoming the beginning inventory of
the following period. Such approach not only adds redundancy, it also gives an estimate of the
stock amount and quality at a given time.
When direct measurements of a stock are difficult or non-representative (for example, it is
impossible to correctly sample a stockpile), it can be preferable to back-calculate the quantity and/or
the contents of an inventory from the data concerning the feed and the reclaim of the stock. Such
data are measured in material flow conditions. In that case, a “stock model” considers the feeding
and the reclaiming streams with their timetable (quantity and quality of material versus time) to
draw a picture of the consecutive quantities and qualities of materials that have been fed to or
reclaimed from the stock. In order to calculate a weighted average, masses can be linked to their
corresponding contents by means of this timetable or lot numbers. The stock model can be
considered as a perfect mixer (such as an agitated tank or a blending bed), a FIFO (first input, first
output such as a silo) or a LIFO (last input, first output such as a box). The main risks of this
approach are the possible accumulation of the inaccuracies linked to the material movements into
the stock and the fact that the stock does not operate exactly as a mixer, FIFO or LIFO model. This
calculation alone can hence, after many months, create a figure far from reality. It must therefore be
considered as an initialization for the data reconciliation algorithm, to be coupled with the
appropriate calculated error. Different approaches are compared with a numerical example
considering a stockpile of copper concentrate, fed and reclaimed during the period. Table 4 reports
the known quantities and copper contents of movements and inventories. The content of Reclaim 1,
the quantity and the content of the ending inventory have to be estimated.

7
Table 4 Quantities and average copper contents of feeding movements and inventories

Movement/Inventory Mass (t) Cu (%) time

Beginning inventory 73 20 t0

Feed 1 52 28 t1

Reclaim 1 54 Not measured t2

Feed 2 56 29 t3

Ending inventory Not measured Poorly sampled t4

Table 5 reports the results of different calculation approaches. The first approach considers a global
material balance over the entire period, and that the contents of the reclaim and the ending
inventory are identical. It is wrong because it does not take into account the time when movements
occurred: the content of Reclaim 1 at t2 also considers Feed 2, happening later at t3. The second
approach considers a stock model with perfect mixing: feed streams are perfectly mixed with the
pre-existing stock. Feed 2 is not considered when calculating the content of Reclaim 1, and is
accounted in the calculation of the ending inventory content, leading to the calculation of two
different contents. This approach can be considered more accurate than the first one. The problem is
that considering a stockpile as a perfect mixer remains a bold assumption. Unfortunately, there is
no alternative since getting a representative sample is impossible. Consider the stockpile at t2
perfectly shaped as an equilateral cone, with a “true” average copper content equal to 23.33 %. Feed
2 is fed in such a way as to ensure that the final cone is still an equilateral one. At t4 a weighted
average defines a “true” average copper content equal to 25.83 %. As shown in Figure 1, if we tried
to sample such a cone with an auger from the middle of the apothem and perpendicularly to the
surface, the resulting sample would present an important increment weighing error. Due to the
geometry of the system, the increments of the two qualities would not be proportional to their real
mass proportion in the stockpile. This sample would lead to a very different estimate of the ending
inventory copper content, that based on geometrical considerations can be estimated as 24.83 %.
This simplified example does not consider the segregation phenomena depending on density and
size classes of particles and yet it shows how figures could get strongly affected by a non-
representative sample. At the end of the period, depending on the chosen approach, the copper
contained in this single stockpile can be estimated with a difference of more than one ton.

8
Table 5 Missing values obtained with different methods

Approach Movement/Stock Mass (t) Cu (%) Cu (t)

Ending inventory 127 25.08 31.86

Global mass balance
Reclaim 1 54 25.08 13.54

Ending inventory 127 25.83 32.80

Perfect mixer stock model
Reclaim 1 54 23.33 12.60

Ending inventory 127 24.83 31.53

Sampling the stockpile
Reclaim 1 54 25.68 13.87

Figure 1 Simplified geometry of a stockpile

Above all, much can be done practically in the plant in terms of stock management, such as
building small stocks with frequent depletion. This can be achieved by building small parallel
stockpiles, one being reclaimed while the other is being fed. This allows to frequently “set the clock
to zero” and avoid huge build-ups of value in the stocks, reducing financial risk. When volumes of
material are relatively small, it can also be interesting to avoid mixing different lots of production.
This practice allows the production of small stockpiles for which metal contents are well known.
Also, when the remaining volume of a stock is sufficiently small, it is the opportunity to directly
weight it, instead of calculating its dry mass from a volume measurement and assuming its density.
These practices are dictated by common sense and can really make the difference in increasing the
knowledge and accuracy of plant stocks.
The implementation of INVENTEO in this copper smelter plant was particularly successful thanks
to the coupling of good practices together with well-designed data management and data
reconciliation. Thanks to automatic filters, data concerning feeding and reclaiming streams are
automatically aggregated and used to calculate the inventory mass and contents with a stock model
chosen on case-by-case basis, depending on the stock management. The subsequent data

9
reconciliation improves the knowledge of the inventories. The role of process engineers on the field
must be underlined as it allowed the improvement of the process efficiency and its accountability
altogether.

Hydrometallurgy and liquid accounting in a zinc smelter plant

In mineral processing, the definition of a wet mass with corresponding solid and metal contents is
sufficient to track the metals contained in solids along the process. In hydrometallurgy, metals can
be present in both liquid and solid phases and they can pass from one to another. A more detailed
phase definition is hence necessary. In a zinc smelting process, leaching and electrolysis operations
must be accounted together with roasting and melting ones. Figure 2 reports a simplified flowsheet.
Some streams are wet solids (roaster feed, cements and cake from leaching…), others are
completely dry (calcine, cathodes, ingots…) or completely liquid (spent electrolyte, purified
solution…). The possibility to define several phases is a major feature to allow a detailed metal
accounting for such process.
Depending on the process conducted in each considered process step, and on the data availability,
different conservation laws should be defined for each node of the flow diagram. In the node
associated to the roasting operation the loss of sulfur and moisture and the gain of oxygen modify
the total wet mass. Liquids and solids are not conserved. Instead, metals in solids must have the
same amount in input and output streams. For leaching and cementation operations the
conservation of the total amount of liquids can be assumed. Moreover, metals can pass from solids
to liquids and vice versa. Hence their total amount must be conserved whichever the phase.
Approaching the end of the process, metals other than zinc are present in low concentration and
often poorly measured. For instance, in the electrolysis node, zinc is mainly present as other metals
have been previously removed by cementation. In this case, only the zinc conservation constraint is
considered in the material balance calculation since not enough information is available for the
other metals. For melting of cathodes and casting of ingots operations the conservation of the total
dry mass and the zinc contained in it can be assumed.

10
 Figure 2 Simplified flowsheet of a zinc smelting process

Table 6 reports an example of possible imbalances for some conservation laws considered in the
material balance. These imbalances are due to measurement errors and can be eliminated thanks to
data reconciliation. Positive values indicate that the input masses are larger than the output ones,
negative values show that output masses exceed the input ones.

11
Table 6 Example of imbalances before data reconciliation

Operation Conservation law Imbalance

(kg)

Conservation of zinc in solid phase 8291

Conservation of lead in solid phase -1217

Roasting
Conservation of copper in solid phase 96.34

Conservation of silver in solid phase 2.793

Conservation of liquid mass -146700

Global conservation of zinc in solid and liquid phases -80560

Leaching and cementation Global conservation of lead in solid and liquid phases 1056

Global conservation of copper in solid and liquid phases -93.43

Global conservation of silver in solid and liquid phases -8.586

Electrolysis Global conservation of zinc in solid and liquid phases 28390

Conservation of solid mass -12100

Melting and casting
Conservation of zinc in solid phase -12100

As shown, the definition of conservation laws has to be very flexible and can be different
depending on the process conducted in each node and on the data availability. For example, one
metal can be conserved in one node and not considered in another one if not enough information is
available to close a balance around that node.

CONCLUSIONS
During the implementation of a metallurgical accounting system, the review of the available data
can give important hints about the flow diagram to be considered. Similarly, the measurement error
estimation points the way forward the improvement of raw data, already before statistical data
reconciliation. The article reports examples of how to define the material flow diagram, the
conservation laws to be considered and the number of considered components, to improve the
accuracy of the final metal accounting by means of data reconciliation by material balance. The use
of a metallurgical accounting solution such as INVENTEO can answer the need for automatic data
aggregation, providing the traceability from starting raw data to final reconciled data, while
ensuring the respect of a well-defined procedure for the calculation of every variable and timely
results. The success of the implementation depends strongly on the implication of the company
personnel. In particular, process engineers have a core role thanks to their functions linked to the

12
process management on a daily basis and their precise understanding of the process operations. In
fact, their proximity to the process is crucial in setting up the metallurgical accounting system, in
maintaining it up-to-date as the process evolves with time, and last but not least, in using the metal
accounting results to optimize the production.

REFERENCES
AMIRA International (2007) Metal Accounting: Code of Practice and guidelines, Release 3, AMIRA
International Africa.
Brochot, S. (2011) ‘The application of sampling theory in metallurgical accounting process –
Inventeo methodology implementation’, Proceedings Fifth World Conference on Sampling and Blending,
M. Alfaro, E. Magri, F. Pitard, Santiago de Chile, pp. 185—193.
Brochot, S., and Durance, M.V. (2012) ‘A new approach to metallurgical accounting’, Proceedings
11th AusIMM Mill Operators’ Conference, The Australasian Institute of Mining and Metallurgy,
Melbourne, pp. 217—224.
Connelly, D.E.G. (2009) ‘Metallurgical accounting standards in process plants’, Proceedings Tenth
Mill Operators’ Conference, The Australasian Institute of Mining and Metallurgy, Melbourne, pp.
281—287.
Durance, M.V., Brochot, S., and Mugabi, M. (2004) ‘Material balance approach for parameter
determination in Bioleaching process’, Proceedings 9th International Symposium on Computer
Applications in Biotechnology (CAB9), Nancy.
Fourniguet, G., Villeneuve, J., Rocchia, L., Védrine, H., Brochot, S., and Guillaneau, J.C. (1997)
‘Analyse par bilans matière d’un procédé hydrométallurgique de recyclage de déchets d’aciéries
électriques’, Proceedings Sixth French Congress of Process Engineering, Société Française de Génie des
Procédés, Paris.
Guerney, P., Dunglison, M., and Cameron, P.M. (2005) ‘Leveraging accuracy and precision —
Multi-phase mass balancing and reconciliation as a tool for quality data management’, Proceedings
Second World Conference on Sampling and Blending, The Australasian Institute of Mining and
Metallurgy, Melbourne, pp. 121—125.
Gy, P.M. (1979) Sampling of particulate materials – Theory and practice, Elsevier, Amsterdam-Oxford-
New York.
Herbst, J.A., Mehta, R.K., and Pate, W.T. (1988) ‘A hierarchical procedure for mass balance closure
and parameter estimation: application to ball mill grinding’, Proceedings XVI International Mineral
Processing Congress, K.S.E. Forssberg, Elsevier, Amsterdam, pp. 1719—1732.
Hodouin, D., and Everell, M.D. (1980) ‘A hierarchical procedure for adjustment and material
balancing of mineral process data’, International Journal of Mineral Processing, vol. 7, pp. 91—116.
Holmes, R.J. (2008) An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 141—170.
Lyman, G.J. (2005) ‘Sampling, metallurgical accounting and reduction of balance estimation
variance’, Proceedings Second World Conference on Sampling and Blending, The Australasian Institute of
Mining and Metallurgy, Melbourne, pp. 127—135.

13
Morrison, R.D. (2008)a An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt
Mineral Research Centre, Indooroopilly, pp. 3—23.
Morrison, R.D. (2008)b An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt
Mineral Research Centre, Indooroopilly, pp. 27—75.
Morrison, R.D., Gun, Y., and McCallum, W. (2002) ‘Metal balancing from concentrator to multiple
ore sources’, Proceedings Value Tracking Symposium 2002, The Australasian Institute of Mining and
Metallurgy, Melbourne, pp. 141—147.
Ragot, J., Darouach, M., Maquin, D., and Bloch, G. (1990) Validation de Données et Diagnostique,
Hermès Science Publications, Paris.
Ragot, J., and Maquin, D. (2006) ‘Validation et réconciliation de données. Approche
conventionnelle, difficultés et développements’, Les Techniques de l'Industrie Minérale, vol. 29, pp.
22—30.
Wortley, M. (2008) An Introduction to Metal Balancing and Reconciliation, Julius Kruttschmitt Mineral
Research Centre, Indooroopilly, pp. 77—139.

14
A Review of the Options in Concentrator Layout
Greg Lane1, Bianca Foggiatto1, Phil Dakin1, Andrew Tew1, Niresh Deonarain2
1. Ausenco, Australia
2. Ausenco, Peru

ABSTRACT
Site and plant layouts have a significant impact on the installed cost of a concentrator. The general
arrangement of equipment, the relationship between layout, earthworks requirements, bulk
materials quantities and optimization of operational and maintenance needs all provide significant
opportunities to maximize the value of projects.
The design of leach plants and concentrators has changed incrementally over the past 50 years and
mostly to accommodate the increasing capacity of process equipment. Concentrator layouts tend to
be designed conservatively and based on previous practice. Engineering companies have tended to
rely on proven designs that decrease their risk. Owners have brought their preferences into the
design as lessons have been learnt from the operation of past projects. Through this process
designers have tended to have very little exposure to site concentrator operation and critical
analysis of previous projects from a design perspective, instead relying on feedback from
commissioning engineers.
This paper challenges some of the paradigms associated with concentrator design and summarizes
the value associated with good decisions when compared with using “standard designs”.

1
INTRODUCTION
Process plant design in the minerals industry consists of a number of interconnecting activities
commencing with ore body evaluation and culminating in critical evaluation of the design in
operation. The design process interfaces with geology, mining, metallurgy, process engineering,
discipline engineering, design, constructability assessment, maintainability assessment, safety
reviews, procurement strategy, contracting strategy and construction planning.
This paper focusses on the impact of plant design and layout on capital cost. Given that process
equipment is common to all layout options, the relationship between layout and cost is driven by
the bill of materials and the constructability of the design.
The design of a particular minerals concentrator is influenced by:
 regulatory standards and requirements for the particular location
 requirements for safe working practices
 operational and maintenance requirements
 climate in the context of the need for buildings
 paradigms in the context of operator and maintenance requirements
 contracting strategy in the context of quantity optimization
 risk management in the context of benchmarking.
Cost effective concentrator design interacts with the project infrastructure constraints, owner’s
needs, vendor’s capabilities, constructor’s logic and operator’s and maintenance team’s preferences.
However, cost effective design has some “rules of thumb”, namely:
 keep the execution strategy and plan simple, sift the “baggage” from the facts early, have a plan
(and agreed scope) and stick to it
 minimize the number of interfaces across all parties as every interface requires “management”
 invest in good equipment as it makes you money in operation
 reduce plant footprint, as capital and operating costs increase with increasing plant footprint.
As a consequence of the above, capital cost will increase if:
 scope is poorly defined and the execution strategy meanders (scope and design are not frozen)
 simplicity is replaced with opportunism (hope)
 pipe rack locations are used as the basis of plant layout or plant areas are spread apart requiring
long pipe racks
 allowance for “expandability” is a necessity.

A SUMMARY OF PREVIOUS WORK

A number of papers have been written on concentrator design and layout. Some examples
pertaining to specific aspects of mineral concentrator layout include:
 Dufour et al (2011) on SAG and ball mill circuit layout and design
 Boyd (2002) on crushing plant design and layout considerations
 Callow and Meadows (2002) on grinding plant design and layout considerations
 Erickson and Blois (2002) on the design and layout of dewatering circuits.
These papers do not specifically deal with the relationship between good layout and capital cost.
Rather they deal with process selection, flowsheet design and layout options based on technical

2
criteria and operational preferences. Whilst technical criteria and operational preferences are very
important, they need to be considered in parallel with the operating and capital cost implications.
The designers “model” should include “rules” for capital optimization, either in software logic or
empirical knowledge form.
The layout of chemical plants is well documented in the literature and typically focusses on
minimizing operational risk and improving plant safety, see for example Mecklenburgh (1973). The
improvements in CAD over the last 30 years have led to many studies into how to optimize
chemical plant design based on a number of factors; see for example Schmidt-Traub et al (1999).
Whilst predominantly focussed on the chemical processing industry, Figure 1 from Schmidt-Traub
et al summarizes the high level considerations in layout development that have been captured in
subsequent design software packages.
The key elements in Figure 1 are:
 equipment modelling – space, operation and connection
 layout modelling – alternative assessment
 pipe routing – pipe rack planning and routing
 analysis and assimilation of the above into the layout.
The Schmidt-Traub approach does not directly deal with the capital cost implications resulting
from the impact of layout on bulk material quantities. Chemical plants are designed based on:
 linking unit processes with pipe and service racks
 safety requirements based on the materials being processed
 operational and maintenance access requirements.

Figure 1 Layout optimization process for chemical plants (adapted from Schmidt-Traub et al, 1999)

THE IMPACT OF THE FINANCIAL CLIMATE

In the five year period from August 2009 to August 2014 the iron ore price was above $80/t. From
about 2 years prior to this period Chinese demand drove project development behaviour across
most mineral commodities. The high commodity prices meant that project schedule became
paramount, far outweighing the importance of development cost for the major mining houses. In

3
iron ore, establishing and maintaining market share was a driving force for expansion and new
projects (Kloppers, 2012). This style of approach to project development, coupled with the high
demand for experienced people and resultant influx of less experienced personnel, resulted in
project development conditions that adversely impacted on the quality of delivery resulting in
project capital cost overruns averaging 62% (EY, 2015).

PROJECT QUALITY, COST AND SCHEDULE

Measures of quality are often subjective except where poor quality design results in lost production.
Hence, determining the quality of the design is difficult to separate from operational factors that are
not related to design. Poor redesign and execution nearly always results in slow project ramp-up,
see for example Toromocho (Emery, 2015) and Sino Iron (Mckenzie, 2016). A well designed and
executed concentrator will ramp-up to full production well inside 6 months provided ore supply is
adequate (McNulty, 2014).
Project cost and project schedule have an interesting relationship that is established in the study
phase. Project schedule is related to project cost in design because optimization of bulk quantities
leads to a reduction construction man hours and a significant reduction in capital cost. Failure to
capture the design optimization opportunities in the study phase results in loss of project schedule
that cannot be recaptured. Rework and redesign takes time and in some cases the cost of lost
schedule (delayed construction and commissioning due to design optimization) cannot be offset by
the savings in commodities because the opportunities were not captured early enough in the design
process. Value improvement processes signifies inefficiency in the original design process that did
not capture scope and cost optimization opportunities.

STANDARD DESIGNS
There have been several attempts to generate “standard layouts” for concentrators and CIL plants.
These have developed based on commercial expedience. Minproc (and others) developed a series of
standard designs for CIL plants in the 1980’s and early 1990’s that allowed it to compete effectively
in the competitive lump sum turn-key gold plant business over a number of years and dislodge
incumbent larger engineering companies (Close, 2002). Bechtel and Xstrata reached an agreement to
design a “standard” concentrator for a number of copper concentrator projects, two of which were
constructed at Antapaccay and Las Bambas. The same design was also considered in studies for
Tampakan and Frieda River.
In the Minproc example, the driving force was cost competitiveness and reduction in project
schedule. The basic design and layout framework enabled the LSTK/EPCM for plants as large as
2 Mt/y to be completed in less than 40 weeks. This approach did lead to some performance issues
for projects such as Three Mile Hill where the ore competency was extremely high and the standard
design was not appropriate (Newell, 2015; Lane et al, 2011). However, the same situation occurred
for some bespoke plants constructed by other engineers, such as St Ives. Both St Ives and Three
Mile Hill initially installed SAG mills and had to retrofit stage crushing to achieve desired
throughput.
In the Bechtel-Xstrata example, the aim was to reduce project schedule and contractors EPCM costs
during a period of heated demand for resources but using a standard design. This approach led to
several inefficiencies in design due to the variation in ore competency across ore bodies and

4
resulting compromises in design. It is arguable that these inefficiencies more than offset all the
efficiencies of the “standard design” due to the excessive bulk materials required for generic layout.
Notwithstanding the above, standard designs are highly advantageous for the engineer and
designer. They are usually a compilation of the best aspects of design and represent the aggregation
of years of experience. This approach can be highly beneficial as long as the “standard” is
challenged for every project from technical, delivery, operations and maintenance perspectives.

FACTORS THAT INFLUENCE PROJECT LAYOUT AND COST

A well-designed plant layout suits the design criteria, flowsheet and selected equipment in the
most economical possible configuration. Thus, understanding the cost/risk/benefit relationships is
key to successfully design minerals processing plants.
The principal design parameters that drive the design minerals processing plants include: ore
characteristics and life of mine, production requirements, project location, climatic conditions and
environment, expansion plans, capital and operational costs, maintenance requirements and safety.
Prior to the definition of the design criteria, it is necessary to obtain representative ore samples,
conduct testwork and define ore properties which drive the selection of flowsheet and equipment.
Ore characteristics change over the life of mine, thus understanding the ore body, mineralogy and
process responses is necessary to allow process and market risks to be managed effectively.
Production requirements are typically defined by the client and along with the ore characteristics
drive the flowsheet selection. The flowsheet options need to be considered and evaluated prior to
detailed engineering and project execution proper. Value engineering assessments can occur during
the design process but these need to be limited to low level issues and not matters of scope or issues
material to the schedule. Value engineering exercises to contain capital cost after 30% engineering
completion mean that the project was not set up initially with the correct capital and/or design
expectations, and changes to flowsheet may cause rework and administrative churn within the
project.
Project location may have major impacts in plant design and construction costs. Survey and
geotechnical data generally hold up progress when defining the plant location and finalising
earthworks detail. Site survey and geotechnical studies need to be completed in the study and front
end engineering phases. In addition, in-country materials of construction costs need to be
understood in order to make cost effective structural decisions (e.g. concrete versus steel).
Local weather or environmental issues may define the need for a plant under roof, inside buildings
or with other protection. The cost of installing plant in buildings, particularly in the typical South
American style, is high.
Reagent delivery and on-site storage requirements need to be defined based on plant access
limitations (seasonal weather and/or other social and environmental factors). Environmental
approvals need to be finalized and permitting requirements (traffic, run-off, dust, noise, fumes, and
materials safety) need to be clearly defined before the layout design starts.
Often high grade ores are exploited in the beginning of the life of mine and as grades decline, so
does the production of final concentrate. Production is attained through expansion rather than by
increased metallurgical recoveries, thus project expansion requirements and timing need to be
clearly defined. Planned expansions should be considered prior to the commencement of layout
design in order to minimize footprint, capital and operational costs.

5
Higher operating availabilities are typically achieved when plants are designed for maintainability
and ease of access. Client maintenance preferences and local crane availability impact on the
decisions to use overhead cranes, tower cranes, monorails or davit cranes for various duties. These
decisions need to be made early in the plant layout process to avoid rework in all disciplines.
Other requirements needed early in the plant layout process include client approved flowsheets
(these determine the scope, mechanical equipment and battery limits) and preliminary piping and
instrumentation diagrams (P&ID’s) (these determine the network of piping and pipe racks
adjoining all areas), the definition of concentrate transportation methods (e.g. truck or rail) and
water storage method (ponds versus tanks).
Lane et al (2005, 2008, 2009, 2011) listed the a number of factors driving selection the layouts, which
are summarized below:
 ‘Fit for purpose’ design considering climate, environmental constraints etc. Use topography to
minimize earthworks and using the sloping topography to reduce conveyor length/lift and to
allow gravity for major slurry flows. Heavy or vibrating equipment should be mounted as low
as possible and founded in cut.
 Keep elevation of the equipment to a workable minimum. The elevation of run-of-mine bins,
mills, cyclones, flotation cells and thickeners are key drivers.
 Optimize the plant footprint with the aim of reducing concrete, structural steel, piping and
electrical/control cable and raceways. Push all areas close together, e.g. HV drives close to
substation, grinding close to CIL/flotation, desorption and reagents close to CIL/flotation,
air/water services close to plant, thus reducing roads, conveyors, pipe racks and electrical
distribution costs.
 Minimize transport distances - primary crushers and run of mine (ROM) leaching pads should
be located as close to pit exit as possible and at the same elevation.
 Avoid dominating pipe racks, building close-stacked vertical processes or large platform areas.
Have common platforms, stairs, pipe and cable ladder supports.
 Keep platework and lining lean, e.g. only put wear liners in the chute areas exposed to wear
and not all internals.
 Clearly define duty/standby equipment.
 Use designers with previous field installation, commissioning and site as-built experience, as
they are better equipped to understand the basics of layout and operability and what is ‘fit for
purpose’.

QUANTITY TARGETS FOR LARGE CONCENTRATORS

Table 1 contains typical Lang factors for various projects. The Lang factor is the installed direct cost
divided by the installed mechanical equipment cost. It has a relationship to the design philosophy
and layout of the plant, directly reflecting material quantities and local factors such as labour cost
and site location.

6
Table 1 Lang factors for copper concentrators

Project / Context Lang Factor

S.E. Asia norm 1.9
Ausenco South America 2.1
Australian 2.4
Norm for major projects 2.6 to 2.8

Approximately half of the direct capital cost of a copper concentrator is associated with the
comminution circuit, primary crushing to cyclone overflow. The layout of this area of the plant has
a large impact on overall plant layout and the overall plant capital cost.
A typical cost breakdown for a South American plant comminution circuit is illustrated in Table 2.
Approximately a quarter of the cost and more than a quarter of the man-hours are associated with
the concrete and steel quantities. Another quarter of the cost is associated with other commodities
such as electrical, platework and piping whose quantities are also related to layout. Hence,
reducing the footprint can impact on up to 50% of the grinding area capital cost with an additional
flow-on impact on indirect costs.

Table 2 Percentage cost distribution for a typical South American concentrator comminution
circuit

Steel Mechanical Grand

Area Concrete Other
Work Equipment Total
Primary crushing 1.2% 0.1% 2.1% 1.8% 5.2%
Coarse ore conveyor 0.6% 0.2% 7.0% 1.6% 9.5%
Coarse ore stockpile 0.5% 1.4% 0.1% 0.6% 2.7%
Coarse ore reclaim (incl. mill feed conveyors) 2.5% 0.2% 1.9% 1.5% 6.0%
Grinding 9.3% 5.5% 41.2% 15.2% 71.2%
Pebble crushing 0.5% 0.6% 2.5% 1.7% 5.4%
Total 14.5% 8.1% 54.8% 22.6% 100%

Figure 2 depicts various concrete and steel relationships with installed grinding mill power for
various concentrators across the world. The concrete and steel ratios vary by project based on the
layout of the plant, mill configuration and design basis. A typical South American concentrator has
0.4 m3 concrete per installed kW.

7
 0.5 0.12

0.45
0.1
0.4

0.35
0.08
0.3

0.25 0.06

0.2
0.04
0.15

0.1
0.02
0.05

0 0

Concrete m3/kW Installed Steel t/kW Installed

Figure 2 Concrete and steel benchmarked quantities

Table 3 depicts an example of reductions in concrete and steel quantities for a South American
project. These reductions in concrete and steel will directly impact the man-hours and schedule of
the project resulting in associated savings.

Table 3 An example of potential reduction in bulk material quantities based on paradigm shift in
layout and design

Current Benchmarked % Change

Area
Concrete (m )
3 Steel (t) Concrete (m ) 3 Steel (t) Concrete (m3) Steel (t)
Primary crushing 6500 125 3500 170 46% -36%
Coarse ore reclaim (incl.
8000 300 3000 340 63% -13%
mill feed conveyors)
Grinding 49500 6500 27500 3258 44% 50%
TOTAL 64000 6925 34000 3769 47% 46%

CONCLUSION
There is no panacea solution but there are some key issues to consider in the design and layout of
any plant as outlined in this paper. However, the critical considerations are to deliver the owner(s)
targets, budgets and project timing. To achieve this, it is the engineer’s role to optimize the design
within the owner’s constraints to achieve maximum value from the project.
To assist the engineer, a clear strategy in terms of scope and execution needs to be defined and
agreed upon between the owner and the engineer as early as possible in the design process. Items
of significance, which include any plant expansion requirements, need to be resolved prior to the
commencement of layout design to ensure opportunities to minimize costs, both capital and
operating.
Early engagement by the owner with the engineer will allow the engineer to challenge and
optimize the plant design during the study and front end engineering phases. It is during this work
that the greatest impact on overall project cost can be achieved without impacting the project

8
schedule. This is best achieved utilizing engineers and designers with previous field installation,
commissioning and site as-built experience.
Any ‘value engineering’ exercises to contain capital cost after 30% engineering completion is an
outcome of the project not being set up correctly during the initial engagement process between the
engineer and the owner(s). Changes to the flowsheet and layout at this stage of the design process
will likely cause rework and administrative churn within the project resulting in unnecessary
schedule delays.
By challenging ‘standard design’ convention considerable project savings can be achieved by
minimizing footprint and associated bulk quantity requirements through ‘smart’ approaches to the
layout. This mantra will also reduce construction man-hour requirements and hence project
schedule resulting in considerable savings to the project owner(s).

REFERENCES
Boyd, K. (2002) Crushing plant design and layout considerations, Mineral Processing Plant Design,
Practice and Control Proceedings, pp 669.
Callow, M.I. & Meadows, D.G. (2002) Grinding plant design and layout considerations, Mineral
Processing Plant Design, Practice and Control Proceedings. pp 801.
Close, S. E. (2002) The great gold renaissance, Surbiton and Associates.
Emery, A. (2015) Mining in Peru: Toromocho starts commercial production,
<http://www.bnamericas.com>, accessed June 2016.
Erickson, M. & Blois, M. (2002) Plant design, layout and economic considerations, Mineral
Processing Plant Design, Practice and Control Proceedings, pp 1358.
EY (2015) Mining companies continue to count the cost of megaproject blowouts,
<http://www.ey.com/GL/en/Newsroom/News-releases/News-EY-mining-companies-continue-to-
count-the-cost-of-megaproject-blowouts>, accessed May 2016.
Kloppers, M. (2012) Diversification and delivery in a cyclical world, Brisbane Mining Club Lunch.
Lane, G., Green, S., Brindley, S. & McLeod, D. (2005) Design and engineering of flotation circuits in
Australia, Centenary of Flotation Symposium.
Lane, G., Staples, P., Dickie, M. & Fleay, J. (2008) Engineering design of concentrators in Australia,
Asia and Africa – what drives the capital cost, Procemin 2008.
Lane, G. & Dickie, M. (2009) What is required for a low cost project? Project Evaluation 2009
Conference.
Lane, G. (2011) Plant design and layout for project development and optimization, IIR Mill
Optimization Conference, Perth.
Lane, G., Dakin, P. & Elwin, D. (2011) Cost effective concentrator design, Metallurgical Plant Design
and Operating Strategies (MetPlant). Perth, WA.
Mckenzie, M. (2016) Ramp up the focus at Sino Iron, < https://www.businessnews.com.au>,
accessed June 2016.
Mecklenburgh, J. C. (1973) Plant layout: a guide to the layout of process plant and sites, Wiley.
McNulty, T. (2014) Plant ramp-up profiles and an update with emphasis on process development,
COM 2014 - Conference of Metallurgists Proceedings, CIM.

9
Newell (2015) Processing in feasibility studies, AusIMM SQ Branch,
<https://www.ausimm.com.au/content/docs/branch/2015/sthqld_2015_11.2_presentation.pdf>,
accessed May 2016.
Schmidt-Traub, H., Holtkonner, T., Lederhose, M. & Leuders, P. (1999) An approach to plant layout
and optimization, Chem. Eng. Technol. 22 (2), pp 105.

10
Improving Advanced Control for Better Mineral
Throughput, Size Classification and Cu Recovery
Daniel Silva1, Bill Gough2, Tomás Carricajo3 and Andrés Chacc4
1. Minera Los Pelambres, Antofagasta Minerals, Chile
2. ANDRITZ, Canada
3. ANDRITZ, Chile
4. MINEVO Technologies, Chile

ABSTRACT
A reliable advanced control system is vital for continuous and effective operation. Therefore, AMSA-Los
Pelambres has implemented an ongoing learning process that allows for continuous improvement of the
control system and the process itself.
In order to optimize mineral processing, the main control objectives have been identified by process area.
The primary grinding control task is to minimize SAG specific energy consumption; the secondary
grinding must deliver optimal liberation size while rougher flotation control should aim to maximizing
Cu recovery.
To better meet these demands, two advanced control layers were devised, implemented and tested. The
lower layer objective is to stabilize the process based on model predictive control tools, chosen for its good
stability, convergence and handling of delays. The upper layer aims at optimizing the process targets,
while at the same time collecting information for decision making in order to increase performance and
automatically adapt its conditions to changing ore and/or process conditions.
To find such optimal operation points, years of process information was thoroughly studied using Big
Data techniques. In particular, a SAG mill phenomenological model was developed to account for
variables that cannot be measured in real-time. For secondary grinding, a virtual particle size sensor was
designed by studying classification performance. Similarly, a dynamic compensator control technique was
designed to adjust rougher flotation upstream and downstream cell levels.
After several months of operation, preliminary results show an increase in throughput up to 2%,
decreased P80 up to 30 microns and increased recovery up to 2%. This is a significant advance because -
by adapting itself to changing process conditions, anticipating disturbances and incorporating process
behavior knowledge - it can outperform known control strategies, representing thus, a further step
towards man-less operated plants.

1
INTRODUCTION
A mining advanced process control system (APC) must always be changing and improving itself because
the mining requirements and technologies are evolving every day, too. The common case - for a non-
mining industry - is that the incoming material fed to the process does not change its properties over time,
unless it is required. A highly different scenario exists in the mining industry, where the amount and
quality of the incoming ore is always changing, even hour-to-hour. While the ore’s main properties could
be statistically estimated and be of use for metallurgical purposes, it is of no practical use for the real-time
control of the plant. Furthermore, there are properties that are very hard to estimate that might cause
significant process effects.
As such, the APC of a concentrator plant must deal with any kind of ore, considering known
characteristics - changing daily - and unknown ones, which might change less frequently.
Therefore, new technologies, new strategies and new control approaches are continually implemented
and tested (Silva & Tapia, 2009). AMSA Minera Los Pelambres (MLP) has assembled an APC team aiming
for continuous process improvement. The team has implemented an ongoing learning process that allows
understanding process behavior and dynamics, predicting future scenarios, implementing new control
logic and measuring, through specially devised KPI’s, improvements in the process performance.
This paper presents the successful control strategies and insights that have been implemented in the last
years on three areas of the concentrator plant.

ADVANCED CONTROL STRATEGIES

After applying various advanced control systems by using self-developed automation schemes, vendors’
expert systems, university-developed systems, or a combination of them, the MLP APC team has
converged to one general control philosophy conceived with two APC layers, each with different
purposes.
The lower layer objective is to stabilize the process. While previously the traditional control systems relied
on conventional proportional-integral-derivative (PID) controllers to perform the underlying regulatory
control of the process wherever possible, the mineral processing operations present many control
challenges due to variations in unmeasured ore properties, material transport delays and nonlinear
response characteristics. Thus, the model-based predictive control (MPC) solution is an adaptive and
predictive controller that works better than PID under the challenges mentioned (Clarke, Mohtadi &
Tuffs, 1987).
At this moment, MLP has the BrainWave™ controller (Gough, 2011) installed in the primary grinding,
secondary grinding and rougher flotation. This MPC platform offers stable solutions (defined and
repeatable); also well understood in terms of gains, time constants, variables interactions, etc.; and they
usually have a short implementation time.
The second and upper layer is the optimizing one, which sends the desired process operation points to the
MPC layer to drive the process towards optimal performance. For this layer, several automatic control
techniques have been tested and implemented and the techniques comes from literature (Sbárbaro & del
Villar, 2010), in-house developments and from the specific plant process knowledge.
In this layer, soft-sensors are implemented to estimate the unmeasurable variables; phenomenological
models of process equipment are designed to discover optimal process operation points and; emergency
rules are written to protect and to avoid extreme conditions, as if it were an expert system. The three
mentioned process areas have an optimizing layer designed specifically for each and most of the logic has

2
been implemented on the IDEAS™ platform of advanced control (Andritz, 2016). The information
required for the process and control study was provided by the installed historian PI System (OSIsoft,
2016). For data analysis, Data-Mining and Big-Data techniques were used.

Model-based Predictive Control implemented at MLP

The advanced process controller BrainWave™ is a model-based predictive controller that uses a state-
space model derived from the Laguerre orthogonal basis functions (Zervos & Dumont, 1988). Each
Laguerre basis function is a polynomial multiplied by a decaying exponential; so these basis functions
make an excellent choice for modelling transient behaviour as they are similar to transient signals.
Particularly, BrainWave™ includes a special algorithm designed for processes with an open loop
integrating response that outperforms traditional MPC approaches, which behaves especially well both in
fast and slow changing processes, such as SAG & ball milling circuits and flotation circuits.

PRIMARY GRINDING
The MLP primary grinding mainly considers SAG mills. Filling volume %, speed and ore properties are
the more correlated variables to grinding capacity, energy consumption and throughput (Powell, van der
Westhuizen & Mainza, 2009). If ore properties and SAG speed remain constant; grind size, throughput
and power are related to filling volume % as shown in Figure 1.

Figure 1 SAG mill performance curves

By analysing historical data from SAG mills –patterns can be found and knowledge embedded in the
operational history can be revealed - “grindability” curves were constructed. Grind quality is not a
measured variable in MLP and load weight has proven to be more reliable than filling volume %. Thus,
for each type of mineral and rotation speed, power and throughput curves are built as shown in Figure 2
and therefore, the process is uniquely characterized and its performance can now be monitored.

3
 Figure 2 Throughput and Power curves by speed and load weight

The main advantages of these curves are finding the location of the load weight operation point that
optimizes the production for each mineral condition and real vs. expected performance can now be
evaluated. Also, by means of this new understanding, it is possible to effectively reduce energy
consumption when production is limited.

Control Strategy
The MPC controller was installed to control the SAG mill weight, %Solids, the ore feed tonnage and to
limit SAG mill sound emissions. These four controllers are coordinated by an optimizer supervisor (see
Figure 3) that provides the set-points and operational limits to the MPCs. The supervisor, using
information derived from the grindability curves, adjusts dynamically the relationship between ball
charge and total mineral load (JB/JT).
The MPC SAG weight controller includes measured disturbance variables such as pebbles recycle and ore
size distribution as feed-forward inputs to adapt control when changes to these variables occur.

Figure 3 SAG mill control strategy, main variables

4
Additionally, a KPI based on Residence Time + Power (RTP) was devised to measure the real-time
improvement of the process through the control actions. This RTP index delivers valuable information
about whether each control action is heading in the right/wrong direction. This KPI enables deciding
which grindability curves the control system must use.
For this, based on load weight and throughput variables, a residence time soft-sensor was devised and
implemented. Then, power v/s residence time was plotted as shown in Figure 4.

Figure 4 Residence time v/s power

With this, three zones of operation were selected and each one of them associated with a particular
grindability curve. Thus, the advanced control executes different moves depending on which area the
SAG mill is operating.

SECONDARY GRINDING
The most important objective in secondary grinding is the size classification. MLP uses the particle size
P80 as its KPI. Maintaining it with a minimum variability is a challenging task due to the high level of
fluctuation associated with the process. Thus, the utilization of APC techniques becomes mandatory to
minimize disturbances in the circuit and to achieve a particle size within the desired range. From this
point of view and to succeed in this task, P80 must be monitored continuously with a high degree of
reliability and availability.
The APC team has developed an online estimation of the P80 in the hydrocyclones overflow (Olivares,
Silva & Yacher, 2014) using a nonlinear structure combined with empirical and phenomenological
knowledge of the classification process. This soft-sensor has become an important piece of the advanced
control system, as previous sampled P80 measurements were uncertain and non-periodical.

Control Strategy
The MPC controller was installed in each classification battery to stabilize the inlet pressure and the
incoming %Solids by coordinating the hydrocyclones opening/closing sequence, pump speed and water
addition. The P80 logic is implemented in the optimizing layer to control P80 (as measured by the soft-
sensor), to improve ball mill grinding efficiency and decrease power usage whenever possible. This is
effectively achieved by an adequate control of two sub-processes; classification efficiency, which means
operating the hydrocyclones at nominal performance and; grinding efficiency, which is effectively

5
controlled by adjusting the %circulating load of the mill. With these purposes in mind, real-time decision
maps were implemented to adjust - under several conditions and process scenarios - the adequate
pressure and %Solids set-points, thus driving the process to reach the best possible and achievable P80.
The optimizing ball mill logic was designed to keep measured variables - such as mill power and sump
level - within operational range and to manage all possible emergency conditions, such as mill weight
overload, low/high fresh-feed, mill underpower, roping condition, etc.

Figure 5 Secondary grinding control strategy main variables

Additionally, a KPI based on real-time graphic operational map was devised to measure and monitor the
impact of the control actions. As before, this dynamic index delivers valuable information about whether
each control action is heading in the right/wrong direction.

ROUGHER FLOTATION
Froth control – and therefore, %Cu Recovery and Grade - can be a challenging problem for a number of
reasons. There can be large, sudden and often unmeasured changes in the incoming pulp flow which can
cause large level disturbances. The arrangement of a number of flotation cells in series means that the
outlet flow from one cell is the inlet flow to the next in the line. Therefore, as outlet valve position is
adjusted to control cell level, levels in the downstream cells will also be affected. Additionally, the dart or
pinch valves that are typically used may not always have a fast and accurate positioning response. In
short, flotation level control is difficult due to the combination of significant unmeasured process
disturbances, interacting process variables and imperfect actuators.
The objectives of the flotation APC are stabilizing the froth depth and pulp level and increasing recovery
(achieved by the increased concentrate mass-pulling) in the cells. MPC controllers have shown a
decreased level variability compared to existing PID controllers. Furthermore, a mass flow soft-sensor,
acting as a real-time compensator, estimates the flow based on the levels of two adjacent cells and its
corresponding dart valve position to improve the dynamics of the upstream/downstream cells. Also, an
algorithm is added to diagnose and help overcome the problems caused by valve/dart stiction (Silva,
Henríquez & Chacc, 2015).

6
Control Strategy
Adaptive model-based predictive controllers, as described above, were applied to eight lines of copper
rougher flotation in MLP. Each cell level is controlled by an individual controller which adjusts the cell’s
outlet dart position. To compensate for the impact of dart position changes on downstream cell levels,
each level controller takes the estimated flow coming from the real-time compensator as a feedforward.
This allows each controller to minimize disturbances by adjusting outlet valve positions at the same time
as the inlet valve positions are changing.

Figure 6 Rougher flotation control strategy, main variables

RESULTS

Primary Grinding
The MPC control is able to maintain mill weight on target with low variability and avoid overloads
without making excessive reductions to the ore feed rate. This performance improvement enables the mill
to operate closer to optimum grinding, under almost all operational conditions.
Results show:
 An increase in throughput up to 5% (see Figure 7).
 A decrease in energy consumption up to 5-8% (see Figure 8).

Figure 7 Throughput Increase Figure 8 SEC reduction

7
Secondary Grinding
The result of including a P80 soft-sensor and an optimizing control system is the reduction of the average
P80 and its variability.
Results show:
 P80 reduction up to 30 microns as shown by lab samples (see Figure 9), maintained even at high
throughput.
 A decrease in energy consumption up to 3% and variability reduction up to 30% (see Figure 10).

Figure 9 P80 Reduction Figure 10 Energy Reduction

This performance improvement allows secondary grinding not only to operate closer to optimum
classification size, but also reduce equipment effort, power consumption and thus, improve its overall
efficiency.

Rougher Flotation
There is a high correlation between froth level variability reduction and recovery increase. Thus, froth
level standard deviation must be reduced.
Results show:
 Froth level variability reduction up to 35% (see Table 1).
 An increase in %Cu recovery up to 4.7% (see Table 2).
 An increase in Concentrate mass pulling %Time up to 29% (see Table 2)

Table 1 Rougher flotation variability reduction

Line 1 2 3 4 5 6 7 8

Variability
15% 35% 24% 8% 27% 29% 28% 9%
Reduction

8
Table 2 Rougher flotation main results

Normal Throughput

Conc. %Cu Concentrate

%Cu Rec. Grade Tailings Mass-pulling
Tons Feed Grade Cu/Fe Rougher Rougher Rougher % time

W/APC 180.157 0,60% 0,42% 94,0% 3,4% 0,11% 82,4%

W/o APC 178.170 0,61% 0,42% 92,6% 2,8% 0,14% 63,2%

High Throughput

W/APC 191.329 0,63% 0,47% 92,5% 5,73% 0,090% 82,3%

W/o APC 209.567 0,63% 0,46% 87,8% 6,49% 0,096% 53,2%

CONCLUSIONS
APC systems improve process operations not only in terms of production, but also in safety, availability,
plant stability and production cost. Also, real cases of using Big Data analysis combined with APC
strategies have been presented that further prove their potential to increase the mining business value.
Based on the obtained results, the two-layer process control philosophy proved to be a replicable,
repeatable and successful solution. Also, it has been demonstrated that; choosing adequate process
variables; creating intelligent virtual variables (such as the soft-sensors) and; analyzing their behavior to
extract valuable process knowledge, helps create and deploy successful advanced autonomous APC
systems.
The MLP APC team has taken this structural framework to create more optimization opportunities and
has further realized that technology is the main engine of productivity. Therefore, efforts should be made
towards continuously incorporating higher levels of automation and process innovation.
Finally, our experience still shows that given the actual state-of-the art process control, there is plenty of
room for innovation (i.e., to focus on the business value) based on the following principles:
 The need and the solution are best known by the user.
 There is available technology and plenty of information, but the real value is in the process
understanding, and the unique and particular solutions integration.
 Consequently, technology providers should offer open and connectable solutions (not “black-
boxes”) and innovate on their way of establishing businesses by means of commitment to
results, partnership and flexibility (joint-development). Further yet, risks and expected results
should be shared with the user based on agreed KPI’s to enhance both mutual commitment and
chances of success.
 Since APC solutions provide tangible and quantifiable results they should be considered as
“business assets” and thus, subject to management, benchmarking, operational excellence, etc.
(Silva & Tapia, 2009).

9
ACKNOWLEDGEMENTS
The author gratefully acknowledges the support of ANDRITZ AUTOMATION, CONTAC INGENIEROS
and specially MINEVO TECHNOLOGIES personnel in the development of this work. Also, the
cooperation of Minera Los Pelambres Management team to share their results with the mining industry is
greatly appreciated. Thanks also to the entire Concentrator & Flotation Operations group at Minera Los
Pelambres for their patience and contributions to the success of these projects and for letting us perform
the required but unavoidable tests.

REFERENCES
Andritz (2016), http://www.andritz.com/, viewed June 29, 2016.
Clarke, D.W., Mohtadi, C., & Tuffs, P.S. (1987), Generalized Predictive Control – Part I. The Basic
Algorithm, Automatica, Vol. 23, No. 2, 1987, pp. 137-148.
Gough, W. A. (2011), BrainWave®: Model Predictive Control for the Process Industries, Advanced Model
Predictive Control, Dr. Tao ZHENG (Ed.), pp. 393-418.
Olivares, B., Silva, D. & Yacher, L. (2014), Real-time adaptive estimation framework for P80 in
hydrocyclones overflow, Automining 2014, Viña del Mar, Chile.
OSIsoft (2016), http://www.osisoft.com/, viewed June 29, 2016.
Powell, M.S., van der Westhuizen A.P. & Mainza A.N. (2009), Applying grindcurves to mill operation and
optimisation, ELSEVIER Minerals Engineering 22, pp. 625-632.
Sbárbaro, D. & del Villar, R. (2010), Advanced Control and Supervision of Mineral Processing Plants,
Advanced in Industrial Control, Springer.
Silva, D, Henríquez F. & Chacc A. (2015), Advanced Flotation Process Control and Instruments
Technology to Improve Recovery at AMSA – Los Pelambres, Procemin 2015, Santiago, Chile.
Silva, D. & Tapia, L. (2009), Experiences and Lessons with Advanced Control. Systems in Minera Los
Pelambres: A practical approach, Procemin 2009, Santiago, Chile.
Zervos, C.C. & Dumont, G.A. (1988), Deterministic Adaptive Control Based on Laguerre Series
Representation, Int. J. Control, Vol. 48, No. 6, pp. 2333-2359.

10
Measuring and Utilizing Ore Quality Variability
Using Real Time On-Belt Elemental Analysis to
Improve Mill Feed Quality
Henry Kurth and Óscar Vásquez
Scantech, Australia.

ABSTRACT
With falling ore grades and greater social pressures on mining operations to reduce tailings
generation one of the technologies has proven effective and beneficial in partly addressing these
concerns.
Real time elemental analysis of primary crushed conveyed flows allows decisions to be made on
continuously measured increments of one to two minutes. An example is presented where iron ore
conveyed to a beneficiation plant that meets product quality requirements is diverted to bypass
beneficiation to save processing cost.
Waste ore or highly deleterious material in conveyed ore stream can be diverted before it is
processed minimising not only grinding, processing and disposal cost, but also concentrate
contamination. Diversion is based on actual analysis rather than grade control decisions in the
mine. Each analysis increment typically represents a tonnage smaller than a haul truck load so this
system is objective in determining ore/waste classification and more selective. Analysis on pebble
recycle conveyors allows material which may be below cut-off grade to be rejected. This reduces the
amount of uneconomic material incurring more processing cost and generating tailings. It results in
higher average ore grade to the downstream process.
A case study is presented for a copper operation where elemental analysis is used to control ore
feed blend from multiple stockpiles into a crusher and mill prior to leaching. Copper metal content
and ore feed mineralogy are controlled to a target range. Calcium and magnesium measurements
determine carbonate content and acid addition control. Moisture content is measured and used in a
similar way. Leach capacity utilisation is improved, copper losses are prevented, and acid
consumption is optimised. Paybacks of a few weeks are achieved. Prompt Gamma Neutron
Activation Analysis through belt, full stream, continuous, multi-elemental, real time analysis is
explained. Real time analysis is essential for these operations to remain competitive.

1
INTRODUCTION
Ore quality variability in real time is not well understood in many mineral processing operations.
Mine model blocks represent hundreds or thousands of tonnes and are assigned average grades
and process parameters based on drillcore intercepts, geological modelling and metallurgical test
work. Variability within a block is not necessarily evident from this data. Planned and unplanned
dilution and ore losses occur through mining and subsequent materials handling. Ore may not be
well blended by the time it reaches the process plant and resulting short term grade variability can
upset plant performance. Pre-concentration processes help remove lower quality or waste
components, to provide a more consistent, higher quality feed to the process plants. Pre-
concentration is not used commonly and most sites rely on grade control at the mine to optimise
ore quality prior to processing. Tonnage to the mill is typically a key performance indicator in
mining rather than ore grade, mainly to keep the plant operating continuously. Ore quality and
variability are rarely measured until they are in the process plant so further opportunities for
optimisation including upgrading or blending prior to processing are missed.
Grade control may involve measurement in addition to visual assessment and personal judgement.
Most methods to analyse ore in buckets, trucks, draw points, blasted piles or even at the mine face
have limited accuracy or are not representative as they rely on surface measurement techniques or
subjective observations. Ore and waste can be incorrectly allocated to destinations further adding to
ore losses and dilution of the ore stream. Representative sampling of coarse material is rare due to
expense and effort required in managing large sample masses.
A solution has been developed to measure primary crushed conveyed ore composition that is
representative, accurate, safe, affordable and proven via Prompt Gamma Neutron Activation
Analysis (PGNAA). This paper discusses this measurement technology being adopted by the
minerals sector with paybacks of weeks in most applications. Real time measurement of multi-
elemental and moisture content provides opportunities for process improvement through good
quality analysis of conveyed increments. Increments can represent tens to hundreds of tonnes and
are typically much smaller than a haul truck load. The plant can respond to each measured
increment based on the potential impact of the composition or its variability on downstream
processes. The technology is used as a process control tool and not to totally replace sampling.
Turnaround time for results of sampled increments and laboratory analysis do not allow for
effective real time control. Utilising an effective real time sensor technology for grade management
has been particularly necessary in the “mine to mill” stages as this area has the potential for the
largest improvements to the process plant at minimal cost.

BACKGROUND
Real time sensing technologies including down-hole probes, handheld qualitative measurements
systems, and truck scanners have been utilised in mining applications. The most commonly used
technologies for frequently measuring material quality in process plants are XRF based slurry
analysers in flotation circuits. Laboratory analysis technologies such as XRF (X-Ray Fluorescence),
LIBS (Laser-Induced Breakdown Spectroscopy), NIR (Near Infrared), LIF (Laser-Induced
Fluorescence) and others have been adapted for conveyed bulk materials analysis. These surface
analysis techniques cannot provide representative analysis of the full belt load unless material is
homogenous, which is extremely rare in the resources sector. Particle sorter technologies are

2
applied to a limited size fraction and require a monolayer so each particle can be measured and its
trajectory controlled accordingly. This limits throughput to a few hundred tonnes per hour and
while effective where ores are amenable to such methods, cannot work on all particles in the plant
feed flow.
PGNAA (Prompt Gamma Neutron Activation Analysis), or Thermal Activation, is proven to be
representative irrespective of particle size, belt speed, mineralogy, layering on the belt, high dust
levels, and moisture content. It is penetrative and continuously measures through the full bed
depth and width of the conveyed load. PGNAA can be configured to measure well over 10,000 tons
per hour and has become the technology of choice for real time conveyed bulk material analysis.
Geoscan uses PGNAA in which a Californium-252 source located beneath the conveyor belt emits
neutrons so that elemental nuclei in conveyed material absorb neutrons and emit gamma rays.
Gamma ray energies measured by Bismuth Germanate detectors above the conveyor help identify
elements from which they are emitted. Spectral data is accumulated continuously over a short time
which is representative of the conveyed load. Neutron and gamma energies penetrate the material
depth ensuring the technique provides representative analysis for an appropriate source size and
detector configuration. Belt weigher and moisture analyser inputs allow tonnage weighted dry
weight per cent for each element to be reported to a control system every few minutes (typically
one to five minutes) for any ore, concentrate or intermediate conveyed flow, enabling operators to
respond accordingly to each result or trend. Enhanced calibration techniques to reduce background
are combined with optimal source size and multi-detector configuration to maximise count rates.
This has resulted in excellent measurement accuracies for most elements of interest in mineral
processing operations. Robust design, remote access capability and appropriate shielding are
incorporated to ensure the equipment configuration suits process plants conditions and remote
locations. There are no access restrictions near the unit while in operation and it requires only
periodic calibration checks.
The coal industry has utilised real time elemental analysis since the early 1990s predominantly to
assess ash oxides and sulphur content in coal for power generation. The cement industry also
employs the technology for limestone blending and raw mix control to reduce product quality
variability (typically by half or more compared with variability resulting from quality control using
sampling and XRF analyses). Since 2003 the Geoscan has been used successfully in minerals
applications in the iron ore, manganese, copper, zinc-lead and phosphate applications. There are in
excess of 60 successful Geoscan installations worldwide in the minerals sector covering iron ore,
copper, phosphates, lead and zinc. Moisture analysis is performed using a transmissive microwave
signal which is affected by free moisture. Attenuation and delay of the microwaves are measured to
indicate moisture content relative to belt load and this signal is used by the Geoscan for dry weight
correction.

METHODOLOGY

Grade Control Applications

Continuous measurements are accumulated and reported over a selected time increment (typically
one to five minutes depending on site requirements) of conveyed flow. The information can be used
in real time to make decisions about that increment but also the processes upstream and
downstream of that point. The measurements may be used passively, actively or proactively. The

3
main applications of the Geoscan discussed below are to measure the composition to determine
variability and enable control to be implemented to either benefit from the variability or minimize
it.

Monitoring
Measuring increments to determine composition is most beneficial when there is an opportunity to
respond to a change in composition. Simply monitoring composition of a conveyed flow may be
useful for feedback, such as an ore flow from a mine to a plant or product onto a ship, to confirm
that the quality meets expectations or a specification. Accumulated results can be effectively used
for metal accounting or ore reconciliation. The availability of real time ore quality data is more
beneficial when used to control the grade and its variability and provide information to
downstream processes in real time. A simple feedback and feed forward process is illustrated in
Figure 1.

Figure 1 Simple feedback and feed forward of real time analyser information for quality monitoring

Short term variability on copper and lead-zinc ores in process feeds are shown in Figure 2 and
Figure 3 based on two minute measurement increments. Tonnage weighted results over longer
periods of time, perhaps 30 minutes, are used to remove variability and smooth results to identify
trends.

4
 Figure 2 Continuously measured copper grade reported every two minutes on feed conveyor over 24 hour
period showing ore grade variability in mill feed

Figure 3 Continuously measured lead-zinc grades reported every two minutes on feed conveyor over 24 hour
period

Individual measurement increments may cover such a small tonnage (10-15 t) that an analysis is
received for each loader bucket amount being fed into a crusher upstream of the analyzer. Trends
may be used to identify when actions need to be taken to change a feed blend or ore source. Trends
can also be used to assign cut off grades for base metals operations. In Figure 2, due to 2 minute
reporting of Cu grade, a cut off grade can be assigned depending on the variability trend of the ore.
Variability within each day and hour are much higher than sites normally expect to see as it is
assumed there will be greater homogenization during mining and material handling processes.

5
Measuring grade variability is important in developing an understanding of how feed grade can be
better controlled (increased and less variable) as most processes perform optimally with a stable
feed quality and throughput rate.

Input-Output Measuring
Measuring process feed and product and/or discard quality may allow processing to be effectively
optimized as performance can be evaluated on smaller increments of throughput which may
indicate potential improvement opportunities. Efficiency and effectiveness of that process stage can
be quantified, be it a complete beneficiation circuit or one process step, such as a jig. The
information can be used to fine tune the process to maximise recovery as feed quality changes.
Measurement of feed flow quality and tonnes to, and reclaim from, stockpiles allows average
stockpile quality to be continuously calculated, enabling decisions on that stockpile to be made
affecting what is delivered to it or removed from it. This also benefits reconciliation and mass
balancing processes where stockpile tonnes and grade can have a major impact at the end of each
production and reporting period.

Bulk Diversion
This process refers to applications where different qualities of material would ideally follow
different process paths. Lag time between measurement and a diversion point is used to ensure
correct allocation of that increment to a process path, such as a stockpile or a processing stage,
depending on the composition. This involves the use of a diverter chute, flop gate or belt tripper to
ensure the increment is selected and diverted as completely as possible.
The most beneficial application of the technology is expected to be in the removal of waste
increments to prevent unnecessary treatment of feed material which is uneconomic to process.
Processing waste occupies plant capacity and consumes grinding energy, grinding media, water,
reagents, and tailings capacity. If waste can be removed at this stage then some 70-90 per cent of the
treatment cost in a plant using a flotation circuit, for example, may be saved for this material.
Figure 5 indicates where analyzers can be utilised to divert increments to waste dumps or further
processing stages. Ore can replace diverted waste in the feed flow resulting in higher average feed
grade, increased plant output for the same throughput rate, improved metal recoveries and more
efficient process plant operation.
Sorting ore onto stockpiles based on composition allows controlled reclaim downstream of
stockpiles. Having multiple stockpiles provides flexibility in managing quality to the next process
stage. Proportions of feed from such stockpiles can be controlled using analysis of the reclaimed
conveyed flow.
Bulk diversion of low grade material can be beneficial to “high grade” the ore body to generate a
faster payback on the mining project. Stockpiled low grade material can be processed once high
grade material is in short supply. Low grade or heavily diluted ore can be diverted to particle
sorting technologies if amenable to upgrading, or blended into process feed flow to improve grade
consistency. Material with high deleterious content, such as pyrite, talc, arsenic, or other elements
which detrimentally affect processing, can be diverted to avoid plant upsets.
In direct shipping ore (DSO) applications in iron ore, bauxite and other commodities, product
quality ore can be diverted to bypass beneficiation processes to ensure unnecessary processing and
its associated costs are avoided. Figure 4 illustrates the possible location for an analyzer for bulk

6
diversion purposes. Additionally, controlling downstream operations after a diverter is clearly
demonstrated. The two analyser locations in a traditional copper processing operations show the
potential energy and processing costs saved by employing real time elemental analysis.

Figure 4b Plant front end showing analyzer and diverted flows to improve waste removal

Blending and Additive Control

Geoscan can be used to control the blending process to improve quality consistency to a subsequent
process stage where multiple feed sources to a conveyed flow are available. Applications can be
prior to any process such as blending ore types by grade or other compositional variation, e.g.
pyrite content, talc content, etc., to minimize variability and optimize process performance.
Products of different quality can be blended to meet specification requirements such as arsenic
content in copper concentrates from multiple mines feeding a custom smelter. Arena and
McTiernan, (2011) outlined how the Sepon Copper-Gold operations have benefited from the
measurement of Cu, Fe, S, Ca and Mg from the Geoscan used there. The Sepon plant crushing and
grinding flow sheet with the Geoscan location is shown in Figure 5. Ore types are blended by
copper and pyrite content by using the real time elemental analysis by the Geoscan. The ore is
blended by proportioning the ROM stockpiles feed to the crusher. In other copper applications the
ratios of elements such as Cu:S are useful to indicate copper mineralogy and therefore treatment
regime.

7
 Figure 5 Geoscan installation at Sepon Cu-Au mine (after Arena and McTiernan 2011)

In processing applications where additives, such as limestone or siliceous flux materials or

bentonite, are required these can be correctly proportioned to meet a desired product composition
range by measuring the product flow and make adjustments to additive flow rates. This application
can be used at the mine, mill, pelletiser, sinter plant, smelter (or blast furnace), chemical plant or
where any blending of conveyed materials is required.

Feed Forward Control

Measuring the composition of the ore feed to a process plant using a Geoscan after crushing can
provide critical advance warning of changes which affect processing. Major changes in ore type, or
geometallurgical domain, may be detected through elemental analysis and allow operators to make
changes to the treatment process before the ore enters the slurry circuit as there is a significant time
delay between the crusher, crushed ore stockpile and grinding mill outflow. At the Sepon operation
referred to above, calcium and magnesium indicates carbonate levels, which affects acid
consumption in their leach circuit resulting in acid addition being managed through feed forward
control. Identification of major deleterious components on the crushed ore conveyor such as talc
(through Mg), pyrite (Fe, S) and others affecting flotation performance allows operators to adjust
reagent dosing to accommodate the composition change. This can reduce contamination of
concentrates and improve metal recoveries. Typically the plant responds to trends rather than
individual analysis results, however, sudden changes can result in plant upsets so identifying these
is important.
The elemental content may be used to identify changes in silica, Si:Fe ratio, or soft minerals which
can affect grinding performance. The feed rate to the mill can be adjusted where harder or more
competent material is detected in the feed to minimise mill overload, or otherwise optimise
grinding performance.

8
CONCLUSIONS
PGNAA technology on which the Geoscan is based is the most representative method of measuring
conveyed flows continuously in real time. Geoscan has proven to be safe, accurate and reliable with
over 60 successful installations in the minerals sector alone. Most units are installed on ore feed
conveyors to utilise or minimise composition variability to improve process performance.
Decisions can be made on relatively small tonnages in bulk diversion, blending, and feeding the
measured results back to upstream process operators and forward to downstream process
operators to improve performance at multiple stages simultaneously.
The technology does not totally replace site sampling and laboratory analysis, but enhances process
control by providing real time analysis for improved responsiveness to material variability. The
benefits provided by the technology increases plant performance through improved utilisation,
higher recoveries, and smarter processing.

ACKNOWLEDGEMENTS
The author wishes to acknowledge the assistance of the sites used in the examples in this paper for
permission to use their data. Thanks to Scantech for permitting publication of the information and
for the support and input from contributors to, and reviewers of, the paper.

REFERENCES
Arena, T. & McTiernan, J. (2011) ‘On-belt analysis at Sepon Copper operation’, Proceedings
Metallurgical Plant Design and Operating Strategies (MetPlant 2011) 8-9 August 2011 Perth, WA, The
Australasian Institute of Mining and Metallurgy, Melbourne, pp. 527-535.
Balzan, L., Beven, B.J., & Harris, A. (2015). Geoscan online analyser use for process control at
Fortescue Metals Group sites in Western Australia. Proceedings Iron Ore Conference, 13-15 July 2015,
Perth WA, The Australasian Institute of Mining and Metallurgy, Melbourne, pp. 99-105.
Beurton, G., Ledru, B., & Letourneur, P. (1995) ‘Perspectives for on-line analysis of bauxite by
neutron irradiation’ in Internationl Conference on Neutrons and Their Applications, June 12, 1994 - June
18, 1994. Crete, Greece: Society of Photo-Optical Instrumentation Engineers.
Cottle, I. (2007) ‘Novel online analysis applications in the iron ore industry’ in Proceedings Iron Ore
Conference, Perth WA, 20-22 August 2007, The Australasian Institute of Mining and Metallurgy,
Melbourne, pp. 241-244.
Delwig, C., Fettweis, H., Schnitzler, T.,Wienstroer, S., Ferguson, S. and Noble, G. (2011) ‘Use of an
online elemental analyser to optimize the sinter process at ThyssenKrupp Steel Europe, Duisburg,
Germany’ in Proceedings Iron Ore Conference, Perth WA, 11-13 July 2011, The Australasian Institute of
Mining and Metallurgy, Melbourne, pp. 381-388.
Kurth, H. (2007) ‘Online analyser technologies for conveyed materials - timely, representative and
reliable data for improved process control’, Proceedings 2007 Australian Mining Technology
Conference, “Smart Technologies for Overcoming the Skills Shortage” (ed: M Hood, P Knights and P Lever),
The Australasian Institute of Mining and Metallurgy, Melbourne, pp. 43-52.

9
Kurth, H. & Edwards, M. (2008) ‘Use of on-belt analysers for improved process control of conveyed
materials’, Proceedings Metallurgical Plant Design and Operating Strategies (MetPlant 2008) 18-19
August 2008 Perth WA, The Australasian Institute of Mining and Metallurgy, Melbourne, pp. 493-
507.
Leonida, C. (2015) ‘So where’s the grade control?’ in IPCC feature, Mining Magazine, April 2015:35–
36.
Matthews, D. & du Toit, T. (2011) ‘Validation of material stockpiles and roll out for overall
elemental balance as observed in the Khumani iron ore mine, South Africa’, Proceedings Iron Ore
Conference 11-13 July 2011 Perth, WA, The Australasian Institute of Mining and Metallurgy,
Melbourne, pp. 297-305.
Minnett, R.C.A. (2010) ‘The state of sampling practice in the South African minerals industry’ in
Proceedings Sampling Conference 11-12 May 2010 Perth, WA, The Australasian Institute of Mining and
Metallurgy, Melbourne, pp. 31-50.
Nelson, M.G. & Riddle, N.K. (2003) ‘Sampling and analysis for optimal ore-grade control’ in 2003
SME Annual Meeting, February 24-26, Cincinnati, Ohio. Preprint 03-128, pp. 1-8.
Patel, M. (2014) ‘On-belt elemental analysis of lead-zinc ores using Prompt Gamma Neutron
Activation Analysis’, XXVII IMPC 2014, 20-24 October 2014 Santiago, Chile, Gecamin.
Tran, K.C. and Evans, M. (2001) ‘Intelligent instruments (12) for process control of raw materials in
the basic industries’ in 2001: A Mining Odyssey – SME Annual Meeting, February 26-28, Denver,
Colorado. Preprint 01-34, pp. 1-6.

10
CHAPTER 4

Water Resources
Optimization and
Tailings Management
Effect of Seawater on Solid-Liquid Separation and
Transport Properties of Magnetite Tailings
Matías Álvarez1,2, Christian Ihle1,2, Willy Kracht1,2 and Cristian Reyes1,2
1. Department of Mining Engineering, Universidad de Chile
2. Advanced Mining Technology Center, Universidad de Chile

ABSTRACT
It is well known that seawater affects mineral flotation, especially in the presence of clays. A
comparatively less studied problem is the effect that seawater on the behavior of tailings in light of
the solid-liquid separation at the thickening stage, and the subsequent flow properties towards
tailings placement. In the present study both settling tests and rheology measurements have been
performed between iron ore magnetite tailings samples, whose P80 is close to 40 µm, and different
seawater fractions. The outcomes of experimental measurements focus on the turbidity of the
supernatant layer and the settling velocity in the case of settling tests, and, in the case of rheology
viscosity and yield stress. Settling tests have been made using a video-based system to track the
interface between the supernatant layer and the settling pulp region. Rheology measurements have
been made using a concentric cylinder rheometer, considering test times significantly smaller than
the settling time scale. Results show that yield stress, plastic viscosity and settling velocity depend
on the seawater fraction and the particles concentration in the suspension. In regard to the
rheological parameters is possible to determinate that yield stress and plastic viscosity are sensitive
to seawater fraction, especially for 75% and 100% and both parameters increase when increases the
solid fraction by weight. Settling tests allowed to determinate that seawater fraction is significant in
this parameter, generating a minimum value for a seawater fraction of 50%. By contrast, the pH
value in the slurry is less relevant to explain the changes in yield stress, viscosity and settling
velocity.

1
INTRODUCTION
Understanding the rheological properties of tailings is of vital importance in the mining processing.
Little is known about the interplay between tailings and different electrolytes including seawater.
These electrolytes have direct impact on viscosity and yield stress of suspensions. The yield stress
of particulate suspensions is typically related to particle shape and size distribution, solid
concentration and surface chemistry (Kapur et al., 1997; Zhou et al., 1999; Ancey & Jorrot 2001) and
it is expected that these factors affect the settling of particles. The viscosity and yield stress of
suspension have been routinely measured and several authors have explored the relationship of
these properties with the interaction potential between particles (Channell & Zukoski, 1997; Scales
et al., 1998; Johnson et al., 2000; Franks et al., 2000; Zhou, Scales & Boger, 2001), and it is expected
that these factors affect the settling of particles.
The viscosity and yield stress of suspensions have been measured by several authors in relation
with the interaction potential between particles (Channell & Zukoski, 1997; Scales et al., 1998;
Johnson et al., 2000; Franks et al., 2000; Zhou, Scales & Boger, 2001).
When particles are in suspension with seawater, different reactions that involve the cations and
anions in seawater with the minerals through exchange happen. This process is complex because
depends on the different bonding energies of Ca2+, Mg2+, K+ and Na+ ions, whereas the effect of the
cations originally resides in the exchange positions, the ionic activity of the seawater, the buffer
mechanism of seawater and the physical properties of particles (i.e. diameter, shape). As a result,
the inter-particles charge and the strength and range of this change is determining by the
electrolytes.
In the simplest case, the inter-particle interactions are governed by the sum of the attractive van der
Waals and the repulsive electrical double layer forces, as defined by the DLVO theory (Derjaguin,
Landau & Others, 1941; Verwey & Overbeek, 1948). The van der Waals attraction is due primarily
to the interaction of instantaneous dipoles generated within the atoms comprising each particle.
This attraction is the major force responsible for aggregation of metal oxide particles in most
processing situation. In the case of repulsive force, the most common interactions between colloidal
particles are electrical repulsive forces which are the result of like surface charges on the particles.
The electric repulsive force can easily be manipulated for controlling suspension rheology. For
metal oxide suspensions, changing the suspension pH will alter the magnitude and/or sign of the
zeta potential, while the addition of electrolyte will affect both the Debye length and the magnitude
of the zeta potential. When there are no repulsive forces between the particles, the strong van der
Waals attraction and/or other attractions will result in an interconnected network of substantial
mechanical strength. These networked flocculated suspension can exhibit a high viscosity and yield
stress, extreme shear thinning, and thixotropy in some cases (Zhou, Scales & Boger, 2001).
In non-dilute system, where the surface separation is short, when the particles approach each other,
the repulsive force increases gradually due the extension of the double layer (Krieger & Eguiluz,
1976). As a result, the electrical double layer repulsion increases the effective volume fraction of the
particles and this brings an increase in the yield stress of this suspension. The behavior of the
particles in a suspension with large electrolyte is different in terms of settling and rheological
properties, even the suspensions are typical of flocculated ones, though at the pH far away from the
isoelectric point (IEP), where the repulsive force is zero. Some studies of the rheological behavior of
metal oxide suspensions containing large electrolyte concentrations has been examined in (Leong et
al., 1993; Chang, Lange & Pearson, 1994; Channell & Zukoski, 1997; S. B. Johnson, 1998). Other
studies mentioned the maker and breaker salts, which change the structure of water. The maker

2
ions such as Li+, Na+, Mg2+, and Ca2+ are small and thus have a strong electric field causing water
molecules around these ions to become highly structured. Breaker ions such as Cs+, K+, Cl- and I-
generally have a larger size than the maker ions and weaker electric fields. It is generally observed
that the relative viscosity of a suspension increases in the presence of maker salts because of the
strengthening of the water bonds. Also It is interesting to note that the viscosity of seawater is
higher than for pure water for all temperatures, which could be postulated to be due to its structure
maker salts (Jeldres et al., 2014)
In the absence of flocculant, both the viscosity and yield stress are increased by the presence of
maker salts for suspensions of both silica (Colic, Fisher & Franks, 1998) and alumina (Colic &
Fisher, 1998). These experimental results suggest that maker counter ions give rise to inter-particle
pair potentials leading to stronger particle networks.
In the present paper, the effect of seawater in the rheological properties of dense magnetite
suspensions is analyzed.

Rheological models
A number of models have been proposed in the past to estimate the yield stress. The linear
Bingham plastic model is very often used,
τ = τb + ηγ̇
Where τ is the measured shear stress, γ̇ is the imposed shear rate and η is the plastic viscosity. To
take into account the shear – thinning characteristic of these suspensions at sufficiently low shear
rates, nonlinear models such as Herschel-Bulkley model have been used.
τ = τHB + Kγ̇ n
This model uses three rheological parameters, the Herschel-Bulkley yield stress (τHB ), the flow
consistency index (K) and the flow behavior index (n)
There are investigators that uses the Bingham model in the full rheological data, meanwhile
another uses only in the linear portion of the curve. There are several reports showing the disparity
between theoretical and experimental results using the Bingham plastic model for the flow of many
non-Newtonian fluids (Hansen et al., 1999; Coussot, 1994; Fordham, Bittleston & Tehrani, 1991) and
a reason for the discrepancies could be the different Bingham yield stresses obtained by the
extrapolation of flow data. On the other hand, the use of Herschel-Bulkley model has proven to be
good choice for many cases (Kelessidis et al., 2007). In this paper the Bingham model was used in
the linear portion and the Herschel-Bulkley in the non-linear portion to get a better fit.

Magnetic separation
The magnetic separation process features the three-way competition between tractive magnetic
forces, gravitational, frictional or inertial forces, and attractive or repulsive interparticle forces. The
feed is split in the magnetic separator into two or more components. If the separator is to produce a
magnetic concentrate, then the tails are the nonmagnetic components (Figure 1). The magnetic
concentrate is called the mags and other less magnetic components are called madlings. The
magnetic and competing gravitational, friction, hydrodynamic, or inertial forces tend to separate
the particles, while attractive interparticle forces tend to reduce the degree of separation
(Oberteuffer et al., 1974).

3
 Figure 1 Diagrammatic illustration of dry drum separator

In a magnetic separation system, the force on a particle toward increasing field intensity (F) is
expressed by:
dH
F ∝ V (Sp − So )H
dt (1)

Where V is the volume of the particle, Sp and So are the magnetic susceptibility of the particle and
surrounding media, respectively, H is the magnetic field intensity and dH/dt is the magnetic field
gradient (Song et al., 2002). This equation shows that the difference of magnetic force between
magnetic minerals and non-magnetic minerals considerably decreases with the reduction of particle
size, thus the separation efficiency of magnetic separation collapses in the fine size range. The poor
efficiency of magnetic separation of weakly magnetic mineral fines can be improved through
increasing magnetic field gradient, field intensity and particle size.
The iron magnetite separation process involves fine grinding. Downstream the thickening process,
the iron ore tailing can be either transported to an aerial disposal place or below at the sea bottom
(Ramírez-Llodra et al., 2015; Dold, 2014 & Ellis et al., 1995). In the present paper, we discuss the
impact of electrolyte in the rheology and settling in this ultra-fine grinding system.

METHODOLOGY
The magnetite tailings particles used in this study had a P80 close to 48 μm and had a density of 2,9
ton/m3 . These particles were dried to 80°C for 24 hours. to eliminate the humidity content in the
sample. The water used for all experiments was a combination of two different sources: Process
water (PW) and Seawater (SW), in different combinations. The water was protected from light, to
avoid the generation of organic matter.

4
Settling tests
The experiments were made using a glass test tube of 1 L and a video-based system to track the
interface between the supernatant layer and the settling pulp region and being processed with the
tool Matlab. The experiments were done using at tailing solid fractions by weight of 45%, 50% and
55%, respectively. The water used for these experiments was a combination of process water and
seawater in different compositions: 0, 25, 50, 75 and 100% of seawater, measured by mass balance.
The height of the interface was registered for 24 hrs. Before to start the tests, the glass test tube
(containing the solid and the water), was stirred during 5 min to get a homogeneous concentration
of the slurry along the test tube.

Rheology test
The experiments were conducted using an Anton Paar RheolabOC Couette viscometer, fitted with a
cup-and-bob CC39/S-SN32517 accessory (internal/external cylinders with internal/external
diameters of 26,66 mm and 28,92 mm, respectively). While the cup is kept fixed, the internal
cylinder angular velocity has been imposed between 0 and 24.3 rad/s-with the corresponding
conversions following ISO 3219 standard. The experiments were done using three different solid
fractions by weight: 45, 50 and 55%, measured by mass balance. The water used for these
experiments was a combination of process water and seawater in different compositions: 0, 25, 50,
75 and 100% of seawater, measured by mass balance. Different shear rate, linear sequences between
0 and a maximum value of 300 s were tested. The particles and the water were mixed during 5 min
-1

to get a homogeneous concentration in the cup before to start the tests. Each test was performed by
triplicate and before to start each test, the slurry was mixed during 45 s to prevent the
sedimentation of particles in the slurry.
A preliminary analysis showed that the viscosity was stable in test times between 40 and 60 s, hence
it was decided to perform shear rate ramps at 50 s. For the analysis of flow curves, shear rates
between 4 and 130 s-1 were chosen. In a first range, shear rates between 4 and 40 s -1, the Herschel -
Bulkley behavior was observed, while in a second range, shear rates between 40 and 130 s -1, the
Bingham behavior was observed. In any case, the parameters were obtained, yield stress and plastic
viscosity in the Bingham behavior, and the yield stress and consistency index in the Herschel -
Bulkley behavior.

RESULTS AND DISCUSSION

Figure 2 shows that the shapes of the flow curves depend on the seawater fraction and the particles
concentration in the suspension. The yield stress and the plastic viscosity for all cases was higher, in
the range studied, at 100% seawater for the Bingham and Hershel Bulkley model, probably because
the presence of maker salts that change the water structure and for the decrease of the repulsion
between the particles by the cations adsorption that causes. It is observed than the case of 50%
particle concentration and 75% seawater presents a shear stress equal to the case of 100% seawater
in the beginning, but slightly higher for shear rates greater than 50 s-1.

5
 Figure 2 Experimental flow curves vs fraction of seawater (SW) for 50% solids fraction (by weight)

Tables 1 and 2 summarize the fitted rheology results obtained from the flow curves. In particular,
the yield stress and the plastic viscosity increase with the particle concentration. For the cases
studied we could see that the effect of the particle concentration is higher than the effect of the
seawater fraction in the yield stress and the viscosity but it is necessary keep studying this topic to
use this kind of water in the industry.

Table 1 Parameters according to Bingham model

Solid fraction by weight, %

SW fraction
45 50 55
%
η [Pa · s] τb [Pa] η [Pa · s] τb [Pa] η [Pa · s] τb [Pa]
0 0.0134 3.477 0.0221 8.566 0.0578 30.960
25 0.0140 3.652 0.0229 8.700 0.0585 30.836
50 0.0136 3.669 0.0231 8.873 0.0580 31.110
75 0.0141 4.146 0.0245 9.863 0.0578 31.159
100 0.0152 4.924 0.0249 10.324 0.0628 35.117

6
Table 2 Parameters according to Herschel - Bulkley model

Solid fraction by weight, %

SW fraction 45 50 55
%
K [Pa · s-0,55] τHB [Pa] K [Pa · s-0,55] τHB [Pa] K [Pa · s-0,55] τHB [Pa]
0 0.177 2.733 0.384 6.414 1.164 24.363
25 0.182 2.883 0.403 6.500 1.196 23.989
50 0.169 2.938 0.425 6.854 1.216 24.120
75 0.196 3.251 0.460 7.597 1.200 24.280
100 0.231 3.799 0.470 7.533 1.455 26.506

Regarding to the pH in the slurry, according Figure 3, this parameter is less relevant to explain
changes in viscosity, yield stress and settling velocity. This includes slurries with different solid
fraction by weight. The adsorption or desorption (especially of OH- ions), is negligible. The slight
decrease of pH could be due the particles adsorb cations causing the release of H+, this adsorption
process is due that the ions seek to compensate the charge of the surface and brings a decrease of
the repulsion inter particles that affect the shear stress of the solution and also the settling of
particles.

Figure 3 pH value of slurry to different seawater fraction and solid fractions (by weight)

In relation to ions concentration, according Table 3, there is greater increase of Cl - respect to other
ions. The change in the concentration should generate a decrease in the viscosity (as a breaker ion).
Nevertheless, the viscosity increases with increasing of seawater (SW), especially in fractions of 75
and 100%. It is possible to considerate, in this case, that breaker ions are less important than maker
ions, even when the variation of maker ions is much lower. The maker ions are more important for
to explain the change in the viscosity than breaker ions. It is possible to explain this effect if it

7
considered than the potential around particles is equal sign than Cl- and opposite to ions as Mg2+ or
Ca2+, that have greater charge, are more important in viscosity and yield stress changes.

Table 3 Ions concentration and water conductivity

SW fraction Cl- Ca2+ K+, Mg2+ Na+ Conductivity

% mg/L mg/L mg/L mg/L mg/L μs/cm
0 2254 763 151 208 835 9560
25 6692 1911 213 407 5309 27710
50 11420 1866 260 517 6816 32700
75 15568 1925 299 614 8953 43200
100 20227 1904 354 723 9714 52800

In the figure 4 it observes that the effect of the seawater fraction in the settling process is significant.
It was expected that the settling velocity increase with the increasing of seawater fraction because
the decreasing of repulsion drives the formation of aggregates and probably this happen from 50%
seawater, for the case of 25% and 0% the behavior could be explaining for the use of process water
that have a different concentration of fresh water and could cause the formation of flocs, with
different shapes and diameter, and for this reason is faster than the another seawater fraction. To
explain this in a complete way is necessary make another studies to check if exist other elements
that affect the particles behavior. A studies of formation of aggregates and zeta potential
measurement could be necessary.

Figure 4 Settling velocity vs seawater fractions for 50% solids fraction (by weight)

8
CONLUSIONS
The sea water has an effect in the rheological properties and settling velocity of magnetite tailings.
For the case of yield stress and viscosity, when the sea water fraction increases both increases, but
this effect is lesser than the effect of the particles concentration. Meanwhile the settling velocity of
particles first decreases and then increases when the sea water fraction increases, probably due to a
coagulation-flocculation process induced by the superficial chemistry of the particles and the ions
in the water. For the full understanding of this, is necessary to investigate deeply the changes in the
superficial chemistry due the ions concentration in the liquid by zeta potential measurement,
agglomerate sizes, distribution size of particles or another similar studies.

REFERENCES
Ancey, Christophe, and Hélene Jorrot. "Yield stress for particle suspensions within a clay
dispersion." Journal of Rheology (1978-present) (The Society of Rheology) 45, no. 2 (2001): 297-319.
Carroll, Dorothy, and C. Starkeu Harry. "EFFECT OF SEA-WATER ON CLAY MINERALS 1 by."
Clays and Clay Minerals: Proceedings of the Seventh National Conference on Clays and Clay
Minerals. 2015. 80.
Chang, Jeanne C., Fred F. Lange, and Dale S. Pearson. "Viscosity and yield stress of alumina slurries
containing large concentrations of electrolyte." Journal of the American Ceramic Society (Wiley
Online Library) 77, no. 1 (1994): 19-26.
Channell, Glenn M., and Charles F. Zukoski. "Shear and compressive rheology of aggregated
alumina suspensions." AIChE Journal (Wiley Online Library) 43, no. 7 (1997): 1700-1708.
Colic, Miroslav, and Matthew L. Fisher. "Effect of the size and affinity for water of excluded ions on
interparticle forces." Chemical physics letters (Elsevier) 291, no. 1 (1998): 24-30.
Colic, Miroslav, Matthew L. Fisher, and George V. Franks. "Influence of ion size on short-range
repulsive forces between silica surfaces." Langmuir (ACS Publications) 14, no. 21 (1998): 6107-6112.
Coussot, P. (1994). Steady, laminar, flow of concentrated mud suspensions in open channel. Journal
of Hydraulic Research, 32(4), 535-559.
Derjaguin, B. V., L. Landau, and Others. "Theory of the stability of strongly charged lyophobic sols
and of the adhesion of strongly charged particles in solutions of electrolytes." Acta physicochim.
URSS (Elsevier) 14, no. 6 (1941): 633-662.
Dold, B. (2014). Submarine tailings disposal (STD)—A review. Minerals, 4(3), 642-666.
Ellis, D. V., Pedersen, T. F., Poling, G. W., Pelletier, C., & Horne, I. (1995). Review of 23 years of
STD: Island copper mine, Canada. Marine georesources & geotechnology, 13(1-2), 59-99.
Fordham, Edmund J., Simon H. Bittleston, and M. Ahmadi Tehrani. "Viscoplastic flow in centered
annuli, pipes, and slots." Industrial {\&} Engineering Chemistry Research (ACS Publications) 30,
no. 3 (1991): 517-524.
Founargiotakis, K., V. C. Kelessidis, and R. Maglione. "Laminar, transitional and turbulent flow of
Herschel--Bulkley fluids in concentric annulus." The canadian journal of chemical engineering
(Wiley Online Library) 86, no. 4 (2008): 676-683.

9
Franks, George V., Zhongwu Zhou, Nanda J. Duin, and David V. Boger. "Effect of interparticle
forces on shear thickening of oxide suspensions." Journal of Rheology (1978-present) (The Society of
Rheology) 44, no. 4 (2000): 759-779.
Hansen, Svein A., Rolv Rommetveit, Nj{\aa}l Sterri, Bjarne Aas, Antonino Merlo, and Others. "A
new hydraulics model for slim hole drilling applications." SPE/IADC Middle East Drilling
Technology Conference. 1999.
Hemphill, Terry, Wellington Campos, and A. Pilehvari. "Yield-power law model more accurately
predicts mud rheology." Oil and Gas Journal;(United States) 91, no. 34 (1993).
Jeldres, Ricardo I., Pedro G. Toledo, Fernando Concha, Anthony D. Stickland, Shane P. Usher, and
Peter J. Scales. "Impact of seawater salts on the viscoelastic behavior of flocculated mineral
suspensions." Colloids and Surfaces A: Physicochemical and Engineering Aspects (Elsevier) 461
(2014): 295-302.
Johnson, Stephen B., George V. Franks, Peter J. Scales, David V. Boger, and Thomas W. Healy.
"Surface chemistry--rheology relationships in concentrated mineral suspensions." International
Journal of Mineral Processing (Elsevier) 58, no. 1 (2000): 267-304.
Johnson, Stephen Bruce. The relationship between the surface chemistry and the shear yield stress
of mineral suspensions. University of Melbourne, Department of Chemistry, 1998.
Kapur, Prakash C., Peter J. Scales, David V. Boger, and Thomas W. Healy. "Yield stress of
suspensions loaded with size distributed particles." AIChE Journal (Wiley Online Library) 43, no. 5
(1997): 1171-1179.
Kelessidis, V. C., and R. Maglione. "Yield stress of water--bentonite dispersions." Colloids and
Surfaces A: Physicochemical and Engineering Aspects (Elsevier) 318, no. 1 (2008): 217-226.
Kelessidis, Vassilios C., et al. "Gelation of water--bentonite suspensions at high temperatures and
rheological control with lignite addition." Applied clay science (Elsevier) 36, no. 4 (2007): 221-231.
Krieger, Irvin M., and Miguel Eguiluz. "The second electroviscous effect in polymer latices."
Transactions of The Society of Rheology (1957-1977) (The Society of Rheology) 20, no. 1 (1976): 29-
45.
Leong, Y. K., N. Katiforis, D. B. O. Harding, T. W. Healy, and D. V. Boger. "Role of rheology in
colloidal processing of ZrO sub 2." Journal of Materials Processing {\&} Manufacturing
Science(USA) 1, no. 4 (1993): 445-453.
Oberteuffer, J. A, “Magnetic separation A review of principles, devices, and applications”,
Magnetics, IEEE Transactions on, 10(2) (1974): 223 238.
Ramirez-Llodra, E., Trannum, H. C., Evenset, A., Levin, L. A., Andersson, M., Finne, T. E., ... &
Vanreusel, A. (2015). Submarine and deep-sea mine tailing placements: A review of current
practices, environmental issues, natural analogs and knowledge gaps in Norway and
internationally. Marine pollution bulletin, 97(1), 13-35.
Scales, Peter J., Stephen B. Johnson, Thomas W. Healy, and Prakash C. Kapur. "Shear yield stress of
partially flocculated colloidal suspensions." AIChE Journal (Wiley Online Library) 44, no. 3 (1998):
538-544.

10
Song, S., Lu, S., & Lopez-Valdivieso, A., “Magnetic separation of hematite and limonite fines as
hydrophobic flocs from iron ores”, Minerals engineering, 15(6) (2002): 415-422.
Verwey, E. J. W., and J. Th G. Overbeek. "Theory of Stability of Lyophobic Solids." {Theory of
Stability of Lyophobic Solids}. Elsevier, Amsterdam, 1948.
Zhou, Zhongwu, Michael J. Solomon, Peter J. Scales, and David V. Boger. "The yield stress of
concentrated flocculated suspensions of size distributed particles." Journal of Rheology (1978-
present) (The Society of Rheology) 43, no. 3 (1999): 651-671.
Zhou, Zhongwu, Peter J. Scales, and David V. Boger. "Chemical and physical control of the
rheology of concentrated metal oxide suspensions." Chemical Engineering Science (Elsevier) 56, no.
9 (2001): 2901-2920.

11
Use of Seawater in Copper Mines in Chile:
Challenges and Opportunities
Rubén Muñoz1, Macarena Terroba2, and José Adriasola3
1. Technical Solutions, IDE Technologies, Chile
2. Copper Center of Excellence, Bechtel, Chile
3. Hydraulics and Hydrology, Bechtel, Chile

ABSTRACT
Fresh water is an essential input in copper ore processing plants. It is typically obtained from natural
sources, such as rivers, reservoirs and groundwater aquifers (generally denominated “continental
waters”), located in the areas surrounding the mine site. However, conflicts with other users have
been increasing during the last several years and, in parallel, environmental regulations have been
changing in order to prioritize human consumption and agricultural irrigation. Furthermore, at the
present time special emphasis has been placed on protecting surface and underground water bodies,
and in promoting sustainable use.
At the same time that access to continental water for use in copper ore processing plants is being
significantly reduced, Chilean copper production is expected to increase in the medium and long
term. As a result, fresh water has become a critical and strategic input for numerous copper mines in
Chile over the last years, especially for those located in areas with severe continental waters scarcity.
Currently, there is a clear and logical trend in the industry to use seawater instead of continental
waters, since the former represents a very stable source in terms of quantity and quality, with fewer
environmental constraints and conflicting consumers.
This paper discusses three aspects related with seawater consumption in the copper mining industry:
quantity & quality, physical and chemical seawater treatment, and intake and conveyance
infrastructure. These topics are crucial in order to confront significant challenges and understand
how to take advantage of some opportunities in the near future. Both technical and business aspects
are offered to the reader.

1
INTRODUCTION

Mining is an important economic activity in Latin America. The exploitation of metallic minerals such
as iron, silver, gold and copper, among others, position the region as one of the major global suppliers
of these raw materials. Sufficient to note that in 2012 the exploitation of mineral represented 7% of
gross domestic product (GDP) of Latin America (ECLAC, 2013), in 2013, a 6% (ECLAC, 2014) and in
2014 again by 6% (ECLAC , 2015).
While use of surface and underground water tends to decline and increasingly restricted, projections
of production in mining - and therefore consumption of fresh water - are increasing. Thus, currently
the fresh water supply for mining industry represents a particular risk for secure water resources for
the operations and new developments, especially for many mining facilities located in areas with
water shortages. Definitely, the current strategy in such cases is to remove the sources of fresh water
and replace them with a reliable, unlimited and competitor’s source: the seawater.
Since mining in general and its processes constitute a broad topic, the focus of this article will focus
on copper mining. This is because in Latin America, six countries currently produce, overall, 48% of
copper globally, and over the next 10 years, this percentage could exceed 60% (Cochilco, 2014).
Following the order of magnitude of their production, these countries are: Chile, Peru, Mexico,
Argentina, Brazil and Ecuador (Cochilco, 2014).
This article describes three specific water issues are addressed in copper mining, which are
considered key to meet the challenges and opportunities in the near future:
 Quantity and quality of fresh water
 Fitness seawater for use as fresh water
 Infrastructure for collection and transport of fresh water

DISCUSSION

Quantity and quality of fresh water

With regard to fresh water consumption per ton of processed copper ore, there has been a downward
trend in copper concentrators and plants leaching (hydrometallurgy) in Chile (Cochilco, 2015). This
reduction in the unit consumption should mainly to operational improvements implemented in
different processes, especially in the thickening and tailings disposal, who have managed to
significantly increase the efficiency of recovery and reuse of water.
As an example, Figure 2 shows the evolution in recent years and the future projection of the total
amount of fresh water demand by plants producing copper in Chile, along with the proportional
contribution of inland water and seawater (Cochilco, 2015). The projection includes both new projects
and existing operations
Figure 1 shows the evolution of this variable in recent years.

2
 Figure 1 Chilean Mining Unitary Consumption Fresh Water Consumption (COCHILCO, 2015)

Figure 2 Make up Fresh Water Consumption in Chilean Mining (COCHILCO, 2015)

In relation to the amount of water, some of the main challenges and opportunities presented to the
community around copper mining are:
 Implementation of new technologies for final tailings disposal (eg filtering paste tailings
and tailings)
 Use of surface water reservoirs coverage to minimize evaporation losses
 Treatment and reuse of water disposal (eg, gray water generated in the camps and
permanent installations) (Gross et al., 2015)
 Use of rainwater
 Use of treated effluent from other industries (eg treated domestic sewage)
 Continuous measurement of water consumption in different production areas, so as to
establish baselines and targets for reducing water consumption
In general terms and as an example, Figure 3 (Bechtel, own development) shows the overall balance
of water from a copper concentrator plant 100,000 ton/day processing capacity, with a net water
consumption of 0.53 m³ / ton. As can be seen, between 30 and 40% of total consumption required

3
quality fresh water and mainly used as seal water, cooling water, preparation of reagents and water.
The surplus is used as makeup water and added to the recovered water to return to the process.

EJEMPLO BALANCE GLOBAL DE AGUA - PLANTA CONCENTRADORA

Agua Repos i ci ón
49 529 m3/día PISCINA AGUA 34 858 m3/día
AGUA FRESCA

211 046 m3/día PISCINA AGUA 28 006 m3/día Agua Recuperada

94%
14 672 m3/día DE PROCESOS 35% recupera ci ón de Tranque Relaves

MINERAL 100 000 ton/día 148 182 m3/día

3 000 m3/día
3% humeda d

97 800 ton/día 97 800 ton/día DEPOSITO DE

228 200 m3/día 80 018 m3/día RELAVES
30% s ól i dos ESPESAMIENTO 55% s ól i dos
PLANTA CONCENTRADORA DE RELAVES

3
Pérdidas 300 m /día
2 200 ton/día
3 CONCENTRADO CONSUMO NETO DE AGUA = 0.53 m³/ton
218 m /día
9% humeda d

Figure 3 Global Water Balance Cupper Concentration plant (Bechtel, own elaboration)

Regarding the physicochemical quality of the fresh water, the fundamental concept is that the
makeup water does not produce negative impact on the recovery of copper and molybdenum (if
applicable), which is the main objective of the process. This becomes particularly relevant when
deciding whether use seawater or desalinated water, and the scenario of using desalinated water, the
key factor is to optimize plant configuration that allow to minimize investment costs (CAPEX) and
operating (OPEX).
Table 1 (Bechtel, own development) shows a benchmarking of physicochemical quality of process
water (PW) and desalinated water (Desai-W) which is currently used in several production plants
copper in Chile, called " Ref.1 "a" Ref.8 ". In addition, the average typical continental water quality
(called "Agua Fresca") and the typical average quality natural seawater (called "Seawater") is shown.
It is noted that the desalinated water has similar characteristics to fresh water continental in most of
the contents of ions, except in relation to dissolved solids and sulfate, while the process water sample,
as expected, a higher concentration in ion content and dissolved solids. Finally, the untreated
seawater shows high contents of chloride, calcium, magnesium, sodium and dissolved solids.
The analysis of use of seawater in the flotation of copper and molybdenum, have shown the following
effects (Castro, 2013):
 Seawater is able to decrease the size of the bubble, replacing in part the role of frother
 in the use of sea water creates a buffer effect, therefore, for pyrite depression using
additional specific reagents required
 The recovery of copper in the flotation circuit may experience a decrease of 1 to 2
percentage points comparing to the use of standard water for processes (similar to
desalinated quality) (Figure 4)

4
  The recovery of molybdenum in the flotation circuit may experience reductions ranging
from 2 to 30 percentage points comparing to the standard water processes depending on
the pH used in the flotation (Figure 4)

Table 1 Benchmarking Water Quality used in mine cupper process (Bechtel, own development)

Figure 4 Variation of Cu y Mo recovery versus pH – seawater usage vs desalinated/fresh water usage

(Adapted from Castro, 2013, with permission)

Observations from recent projects (Veki, 2013) confirm the decreased molybdenum recovery as
shown in the figure above, when seawater quality is used. As a result, desalinated water use is an
option that guarantees there will be no negative impacts on mineral production.
Regarding water quality, primary challenges and opportunities include:
• Optimizing seawater treatment to obtain a product similar to standard industrial water, without
sacrificing efficiency in copper and molybdenum recoveries
• Managing corrosion in both transportation systems and plant infrastructure that come into contact
with seawater

5
Seawater usage
In comparison to other industries, seawater usage in copper mining is relatively new. Its corrosive
properties negatively affect elements in carbon steel. When using seawater, multiple measures must
be taken to control corrosion in order to achieve the operating and maintenance costs estimated in
the design stage.
In many cases, the use of chemical products to prevent corrosion is the best way to counteract this
problem.
However, these products present limitations when seawater is transported over long distances.
During operation of long-distance pipelines, a deterioration in the quality of the water received in
the mineral process plant (at mine site) has been observed, with a greater concentration of corrosion
byproducts, such as iron. In addition, there are phenomena such as localized corrosion and microbial
corrosion. As a result, the pipelines’ interior walls become rougher and this leads to an increase in
power consumption in the pumping systems
Principally, seawater treatment must be capable of removing suspended solids, inactivating bacteria
y reducing potential corrosion. Furthermore, it must be considered that the flow velocity is directly
proportional to the wear rate due to the erosion-corrosion phenomena (Lienenweber et al., 2016a
and 2016b). The treatment recommended by NACE (2012) to reduce the corrosive potential of
seawater consists of removing dissolved oxygen and inactivating the bacteria responsible for
corrosion. As a result, the corrosion rate could be limited to a maximum of 20 mil per year (1 mil =
0.0254 mm), but this process is extremely expensive.
Among the numerous choices that must be made, the decision to use or not use internal coating in
the carbon steel pipes is crucial. This is linked to research of the erosion / corrosion process in
pressurized piping for seawater. Current state of the art is insufficient to provide conclusive answers
regarding the erosion/ corrosion process in different carbon steel qualities, transporting different
seawater properties and under distinct hydraulic conditions. Therefore, this presents a clear
challenge for applied research. Sharing information and experience, and creating opportunities for
collaboration between mining companies and the academic world will help to advance this research.

Desalination
The seawater reverse osmosis (SWRO) process is a proven technology that recently has managed to
reduce desalinated water production costs to less than USD 0.57/m³, see table below from
(Voutchkov, 2013). The process primarily consists of pre-treatment, reverse osmosis and post-
treatment. Depending on the quality of the seawater at source, pre-treatment consists of some or all
of the following processes: coarse screening, chlorination, clarification, coagulation, flocculation and
filtration. In the presence of harmful algal bloom (red tide) and severe proliferation of algae,
treatment could include a dissolved air flotation (DAF) process. However, the requirement of this
process must be analyzed in detail in accordance with the site’s conditions and seawater quality
records.
When seawater quality is below normal conditions, with low suspended solids, proper treatment
only requires flocculation, coagulation and filtration with minimal or no chemical products.
The seawater reverse osmosis (SWRO) process receives pre-treated water and delivers it to a group
of modules working in parallel. Each module has a high-pressure pump, a circulation pump, an
energy recovery devices and a reverse osmosis train that has pressurized vessels with membranes
elements. The saline concentrate resulting from the SWRO process has high pressure and higher

6
salinity. By using an isobaric energy recovery system and a circulation pump, the residual pressure
can be passed to the osmosis train’s feed manifold. Following energy recovery, the saline concentrate
is returned to the sea, where its high salinity is diluted by means of diffusors.
Finally, the desalinated water is transported to the re-mineralization process, where its quality must
be adapted to what is required for the infrastructure coming into contact with the water and the end
users. Figure 5 depicts the Sorek facility, one of the world’s largest desalination plants.

Figure 5 Sorek Desalination Plant, Israel, Courtesy of IDE Technologies

The key economic parameters of an SWRO plant are capital and Operational & Maintenance (O&M)
cost, which together determine the cost of water US$/m3. Table 2 illustrates the lowest desalinated
water cost worldwide (Voutchkov, 2013). All the projects in the table were implemented under Built
Own Operate and transfer (BOO/BOOT) contract model, which reach the lowest costing resulting
from customized and optimized process configuration.
With respect to seawater treatment systems, opportunities for cost reduction include:
• Minimizing power consumption by utilizing new technologies and production strategies
• Creating synergies with other industrial facilities near the coast such as power plants
• Developing clean technology solutions that permit eliminating chemicals in desalination and
enable an operation with increased availability and minimize marine impact on the environment

Intake infrastructure and fresh water transportation

Seawater intake infrastructures are expensive. For various reasons, owners prefer to build maritime
works that cover long-term requirements. Depending on each site’s characteristics, the challenge is
to find the lowest total installed cost configuration for intake, transportation to SWRO and discharge
of the saline concentrate. In addition to requiring technical experts, this entails the involvement of
constructability and cost estimating experts during study stages to ensure that the best alternative is
executed.

7
Table 2 Recent Lowest Cost SWRO Projects. Source: adapted from (Voutchkov, 2013)

Cost of
Desalinated
SWRO Plant Water Energy consumption

(US$/m³) (kWh/m³)

Sorek, Israel – 624,000 m3/d 0,53 3,2-3,7

Mactaa, Algeria – 550,000 m3/d 0,56 3,7

Hadera, Israel –470,000 m3/d 0,57 3,2-3,7

Cap Djinet, Algeria – 100,000 m3/d 0,72 4

Carlsbad, USA – 200,000 m3/d 0,74 2,9

CONCLUSION
The main challenge of seawater supply to the mining industry is to reduce and minimized the impact
on mineral recovery as well as identify clearly the impacts on the integral projects, such as seawater
transport, process plant equipment materials, tailings, thickening, etc. This article presented the main
implications due to seawater quality is in the molybdenum recovery, which might be negatively
affected; the main opportunities in this aspect is into adapt the seawater quality in order to minimize
this effect. On the other hand, the main challenge of the desalination industry is to continuous
innovating to decrease capital and operational cost; where the options of water purchase under
contract models of BOO/BOT might offers the right opportunities to minimize the risk of operation
and avoid investment in non-mineral production infrastructure.

REFERENCES
Castro, S. (2013) Avances tecnológicos en flotación de minerales, FLSmidth Flotation Seminar, May 23-24,
2013, Viña del Mar, Chile.
Brusilovsky, M. (2009) Challenging the Technologies for Sustainable Desalination Solutions, III Conferencia
Asocisión latinamericana de Desalación y Reuse (Aladyr).
Cochilco (2015) Proyección del consumo de agua en la minería del cobre al 2026, Chilean Copper
Commission (Cochilco), Ministry of Mining, Chilean Government.
Lienenweber, G., Cáceres, L. and Bustamante, P. (2016a), Corrosion Behavior of API5LX65 Steel in
Seawater, Piping for Northern Chile’s Mining Industry.” Hydroprocess 2016 | 8th International
Seminar on Process Hydrometallurgy, June 15-17, 2016, Santiago, Chile.
Lienenweber, G., Cáceres, L. and Bustamante, P. (2016b), Corrosion Studies for Carbon Steel Pipe API 5L
X70 Exposed to Sea Water at High Fluid Velocities, XXVIII International Mineral Processing Congress,
September 11-15, 2016, Québec, Canada.

8
NACE (2012) Standard Practice: Corrosion Control and Monitoring in Seawater Injection Systems. National
Association of Corrosion Engineers (NACE), SP0499-2012.
Veki, L. (2013) The use of seawater as process water in concentration plant and the effects on the flotation
performance of Cu-Mo, Thesis for Master’s degree in Process Engineering, University of Oulu, Faculty
of Technology, Finland.
Voutchkov, N. (2013) Cost Estimation of SWRO desalination plants, The Middle East Desalination
Research Center.

9
CHAPTER 5

Lab-Scale Metallurgical
Tests and Pilot Plant for
Process Control
On the Capacity to Predict Metallurgical Indexes
from Mineralogy, Particle Size and Operating
Conditions at Batch Scale
Felipe Contreras1, Juan Yianatos2, Luis Bergh2, Luis Vinnett2 and Isabel Panire2
1. Minera Los Pelambres, Antofagasta Minerals, Chile
2. CASIM, Universidad Técnica Federico Santa María, Chile

ABSTRACT
At present, several concentrators use geo-statistical models to predict feed particle size, final
concentrate grades and potential recoveries from the combination of different geo-metallurgical
units. These tools have shown predictability, however, the actual operating conditions (e.g. reagent
dosage, gas flowrate) and feed characteristics (e.g. feed grade, liberation) are required to adequately
fit the metallurgical indexes of the industrial operation.
In this paper, the potential to predict the copper and molybdenum recoveries and concentrate grade
from laboratory scale was studied. The tests were conducted in a standard batch flotation cell from
Minera Los Pelambres (MLP). Linear models were proposed to describe each metallurgical
parameter. Input data included three different particle sizes, mineralogy and liberation
(QEMSCAN) and feed characteristics, such as copper, molybdenum and iron grades, among others.
Datasets involved information per shift since 2012 to 2015. The variability was decreased by
applying the Hotelling T2 and the Q residuals filters in order to remove outliers.
Partial Least Square regression was performed to explain the recovery and concentrate grade
variability. Cu recovery was mainly explained by P80, content of Cu soluble (or
‘soluble Cu/total Cu’ ratio) and presence of clays and locked Cu minerals in the raw feed, whereas
Cu concentrate grade was mainly explained by Cu grade, pyrite grade (or Cu/Fe ratio) and clays
content in the raw feed. Mo recovery was not clearly explained by the input data, whereas Mo
concentrate grade was mainly explained by the Mo/Cu and Mo/Fe ratios. The regressions showed a
fairly good predictive capacity at laboratory scale, which is useful for scale-up purposes. Further
developments are being carrying out to apply the proposed methodology at industrial scale.

1
INTRODUCTION
Flotation processes are influenced by a large number of variables that change over time, e.g.
mineralogy, morphology, particle size, surface properties, liberation and operating conditions
(reagent concentration, pH, and gas flowrate). The first-order kinetic model introduced by Garcia-
Zuñiga (1935) has been extensively used to characterize the flotation process (Wills & Finch, 2015)
and has shown appropriate descriptions of the flotation performance at laboratory scale (Dowling,
Klimpel & Aplan, 1985) as well as at industrial scale (Yianatos, Henríquez & Oroz, 2006). Empirical
models have also been used to obtain the performance indexes based on the available information
from the plant operation. Variables such as mineralogy, liberation, particle size and operating
conditions are incorporated to statistical models to predict the metallurgical response. Multi-linear
regressions (MLR) were proposed by Hunt et al. (2014) to forecast the copper recovery in a batch
flotation machine. Mineral properties such as grades, lithology, and sulphide and gangue contents
were used as parameters for the principal component analysis (PCA) (Jolliffe, 1986). In this case, the
strongest predictor was the gangue mineralogy, where the regressions presented R2 between 0.69
and 0.85.
Different approaches to obtain empirical models have been reported in the literature (Sepulveda,
Dowd & Xu, 2015). A model that describes the coal recovery in a Denver cell based on the floating
time and the particle size as predictor variables was developed by Kor et al. (2013), where
polynomials of different degrees were tested. An adequate fit (R2=0.96) was obtained by using a
cubic polynomial. Rath, Sahoo & Das (2013) proposed grade and recovery models for an iron ore
from India. The flotation tests were carried out in a 2L Denver cell. The input variables of the model
were pH, collector and dispersant concentrations and conditioning time. Dispersant concentration
was the most important variable for the Fe grade, whereas collector concentration was the most
important variable for the Fe recovery. Sepulveda, Dowd & Xu (2015) modelled Au and Cu rougher
recovery in a flotation circuit by means of a projection Pursuit Regression (Friedman & Tukey,
1974). This regression finds the direction in which the explained variance of the dependent variable
is maximized. The total regression consisted of a combination of models in different directions. The
input variables of the model included grades, lithology, alteration and mineralization. The final
model only included Au, Cu, S, Fe and Mo feed grades. Bulled & Connolly (2009) used non-linear
and linear models to describe the Rmax for molybdenite, chalcopyrite and non–sulphide gangue
from standard rougher flotation tests. The feed material was grouped in three classes according to
their lithology. For the molybdenite, the Rmax was described by a potential model as a function of
the Mo feed grade, obtaining a standard error of 4%. An exponential model as a function of the Cu
feed grade was obtained for the chalcopyrite Rmax. The model fitting had a standard error of 19%.
For the non–sulphide gangue, a linear regression as a function of the Mo feed grade was proposed
(standard error of 0.35%). No effect of the lithology was observed for the three minerals.
In this paper, multi-linear regressions to predict concentrate grade and Cu recovery in batch
flotation tests are presented, using minerals from Minera Los Pelambres (MLP). Several variables
such as feed grades, mineralogy, liberation classes and particle size were used as preliminary
predictor variables. The importance of each variable in the model was evaluated using a Partial
Least Square (PLS) regression to decrease the data dimension. The models might be used in mine
planning, optimization and scaling–up of flotation indexes.

2
METHODOLOGY

Laboratory tests
A total of 1056 flotation tests were performed at laboratory scale, using drill cores that represented
the ore processed by the MLP concentrator over the 4-year period of 2012-2015. The mineral was
crushed and ground to obtain P80 values of 210, 250 and 300 µm per drill core. In each test, the feed
mineral at P80=250 µm was analysed using QEMSCAN system, which allowed identifying the
mineralogy and the distribution of surface liberation. The total Cu (Cut), soluble Cu (Cus), Mo and
Fe grades were also measured for the raw mineral of the flotation tests.
Each batch flotation test was conducted during 10 minutes. The average pH of the slurry in all tests
was 9 with a sample standard deviation of 0.4. The average cumulative concentrate grade was
12.5%Cu with a sample standard deviation of 4.7%, whereas the average Cu recovery (10 minutes)
was 84.5% with a sample standard deviation of 7.0%. For Mo, the average cumulative concentrate
grade was 0.31%Mo with a sample standard deviation of 0.35%, whereas the average Mo recovery
was 74.4% with a sample standard deviation of 10.7%.

Modelling
The multiple-linear regression (MLR) was proposed to describe the recovery and the concentrate
grade of Cu and Mo at laboratory scale. The feed characteristics (e.g. feed grades, particle size,
mineralogy, liberation) and the operating conditions (e.g. pH) were inputs to the empirical model.
The datasets were normalized according to Eq. (1) to obtain the relative relevance of the model
coefficients:
Y  mY
Yˆ  (1)
sY
where Y is the original variable, Ŷ is the normalized variable, mY is the estimated average of Y and
sY is the sample standard deviation of Y.
The general form for the model was y = β1x1 + β2x1 + … + βnxn, where xn is the n-th independent
variable (normalized), βn is the n-th regression coefficient and y corresponds to the dependent
variable. The coefficients were obtained by a least-square estimation, in which 70% of the data was
used for training (model fitting) and 30% was used for testing. It should be noted that the intercept
of the proposed model is zero because of the normalization, Eq. (1).
A total of 41 independent variables were available for each flotation test. These variables included
concentration of 31 mineralogical species, 5 liberation classes, Cu, Fe and Mo feed grades, Cu/Fe,
Mo/Cu, Mo/Fe and Soluble Cu/Total Cu feed ratios, pH and P80 of the raw material.

Data Filtering
A fraction of tests with recoveries or concentrate grades outside the normal operating range was
removed. For instance, tests with Cu recoveries lower than 70% and concentrate grades lower than
5% were removed. As a result, a total of 70 tests were eliminated in this first filter.
Principal component analysis (PCA) was applied to the data that overcame the first filter to detect
unusual values in the X-data. PCA was performed using the PLS_Toolbox 4.0 from Eigenvector
Research (Wise et al., 2005). The Hotelling’s T2 and Q statistics were used to identify anomalous

3
samples in the PCA model (Mujica et al., 2011). Q statistic indicates the lack-of-fit of a sample, i.e. Q
is the difference between a sample and its projection in the PCA model (Wise et al., 2005). The
Hotelling’s T2 is a measure of the variation of each sample inside the model (Wise et al., 2005). A
total of 189 samples were removed from this second filter.

RESULTS AND DISCUSSION

Linear multivariate models were fitted to the data per year (2012 – 2015). The relative relevance of
each independent variable on the concentrate grade and on the mineral recovery was determined
with the PLS_Toolbox, using PLS regressions. The variable importance in projection (VIP) was used
to select predictors according to the significance of the VIP values for each variable (Wold, Sjöström
& Eriksson, 2001; Chong & Jun, 2005). As an example for Cu recovery and concentrate grade, Table
1 shows variables with significant VIP scores based on the available data for year 2015 (ordered by
significance). This dataset consisted of 300 flotation tests. Chalcopyrite feed grade, P80, Cus/Cut
ratio and Cu sulphides locked fraction were the variables with more influence on the PLS recovery
model. For the Cu concentrate grade, the Cu, bornite and the chalcopyrite feed grades along with
the Cu/Fe ratio in the feed were the most important variables.

Table 1 Variables with significant VIP scores on concentrate grade and recovery models

Cu Recovery Cu Concentrate Grade

P80 Cu feed grade
Chalcopyrite feed grade Bornite feed grade
CuS/Cut ratio Chalcopyrite feed grade
% Cu Sulfides Locked Cu/Fe ratio
% Cu Sulfides Liberated Pyrite feed grade
Covellite feed grade % Cu Sulfides Liberated
Covellite feed grade
K-Feldspar feed grade
Anhydrite feed grade
Chalcocite-Digenite feed grade
Muscovite-Sericite feed grade

Preliminary linear regression models were implemented using the predictor variables obtained
from the VIP analysis. Secondly, the amount of input variables was again reduced according to the
significance (confidence interval) of the model coefficients. Figure 1 shows the modelled versus
measured Cu recoveries (Fig. 1a) and Cu concentrate grades (Fig. 1b), based on the final linear
regression for the year 2015. In the regression, 70% of the data was used for training and 30% was
used for testing. For the Cu recovery, an adequate model fitting was obtained, in which more than
98% of the data were in the range ±10%. For recoveries lower than 75%, a slight bias is observed,
where the model overestimates the experimental recovery. In addition, the modelled Cu
concentrate grades also shows a good agreement with the experimental data. From Fig. 1b, more
than 75% of the data was in a ±20% error band.

4
 (a) (b)

Figure 1 Comparison of modelled versus measured (a) Cu recovery and (b) Cu concentrate grade, year 2015

Table 2 shows the regression statistics for the Cu recovery and Cu concentrate grade. The adjusted
coefficient of determination, R2, was 0.59 for the Cu recovery and 0.67 for the Cu concentrate grade.
For both models, the p-value of the overall F-test were lower than significance level (0.05),
therefore, there was no evidence of lack of fit. In addition, normal distribution of residuals was
analysed using a Lilliefors test (Gosling, 1995). This test did not reject the hypothesis of normality of
residuals. Figure 2 shows the cumulative distribution function for the residuals of the recovery and
concentrate grade regressions, together with the fitted normal distribution for the dataset
(maximum likelihood estimator). The modelled normal distribution was in good agreement with
the residual histograms.

Table 2 Regression statistics for the Cu recovery and concentrate grade, year 2015

Recovery Concentrate grade

Multiple R 0.78 0.84
Adjusted R2 0.59 0.68
Standard error 0.64 0.57
Significance F 1.39E-45 1.16E-55

5
 (a) (b)
1 1

0.8 0.8
Cumulative probability

Cumulative probability
0.6 0.6

0.4 0.4

0.2 0.2
Data Data
Normal Dist. Normal Dist.
0 0
-10 -5 0 5 10 -10 -5 0 5 10
Residuals Residuals

Figure 2 Cumulative distribution function for the residuals:

(a) recovery model and (b) concentrate grade, year 2015

Figure 3 shows the model coefficients along with their 95% confidence intervals for the recovery
(Fig. 3a) and concentrate grade (Fig. 3b). The 95% confidence intervals did not include zero,
therefore, these parameters were significant. The variables associated to the recovery were related
to the quality of the feed mineral and the maximum achievable recovery. The particle size, P80, was
the most significant parameter for the Cu recovery, according to its absolute coefficient value. For
the Cu concentrate grade, the effects of Cu and bornite feed grades and the Cu/Fe feed ratio were
most significant. The negative sign of the Cu/Fe effect was caused by the correlation with the Cu
and pyrite grades, which were both significant.

6
 (a) (b)

Figure 3 Model coefficient for (a) Cu recovery and (b) Cu concentrate grade, year 2015

Additionally, all the available information (period 2012 to 2015) was used to obtain an overall
multivariate linear regression for Cu recovery (Fig. 4a) and Cu concentrate grade (Fig. 4b). A
suitable model fitting for the Cu recovery was obtained, in which more than 95% of the data
(training and testing) were again in the range of ±10%. For recoveries lower than 80%, a bias was
also observed. From Fig. 4b, high variability was observed for the Cu concentrate grades. In this
case, 65% of the data was in a ±20% error band. This result was attributed to the wide range of feed
minerals processed during 2012-2015, where the available input data did not justify the concentrate
grade variability. The adjusted R2 was 0.53 for both regressions. The p-value of the overall F-test
was lower than the significance level for both regressions. Again, the normality of the residuals was
corroborated. From these overall regressions, the particle size and the ‘Soluble Cu/Total Cu’ ratio had
the strongest impact on the recovery whereas the Cu feed grade was the most important variable
for the Cu concentrate grade (Figs. 4c and d).

7
 (a) (b)

(c) (d)

Figure 4 Model coefficient for (c) Cu recovery and (d) Cu concentrate grade

8
An overall multivariate linear regression model was conducted for Mo recovery and Mo
concentrate grade. The model for the Mo recovery did not show an adequate predictive capacity
(R2=0.27). Figure 5 presents a comparison between the measured and the modelled Mo concentrate
grades. A good agreement between the experimental and the modelled grade was obtained, for a
range of 0.0 – 3.0% Mo. The adjusted R2 was 0.9 and more than 50% of the data (training and
testing) were in the range of ±20%. In this case, the significant variables were: the Mo/Cu and Mo/Fe
feed ratios and the Mo and Fe feed grades. The p-values of this model (F-test for residuals) was less
than the significance level (0.05), so that the hypothesis of normality of residuals was not rejected.

Figure 5 Model coefficients for (a) Mo recovery and (b) Mo concentrate grade

CONCLUSIONS
Empirical multivariate linear models for batch flotation tests with ores from Minera Los Pelambres
(MLP) were implemented. Cu and Mo concentrate grades and recoveries were evaluated. The feed
ore included processed mineral in the MLP concentrator from 2012 to 2015. Independent variables
involved 41 measured input data, including feed properties (e.g. feed grades, particle size,
mineralogy, liberation) and operating conditions (e.g. pH). Principal component analysis (PCA) and
Partial Least Squares (PLS) were used to remove outliers and to decrease the input data dimension.
A preliminary Cu model included only data from 2015. The significant variables for the Cu
recovery were related to the maximum achievable recovery (P80, Cus/Cut ratio, Cu locked and Cu
liberated fractions). Chalcopyrite feed content was also important for the Cu recovery. For the Cu
concentrate grade, the significant variables were related to the quality of the raw feed in terms of
floatable minerals (Cu and Fe mineral content). Both models showed fairly good predictive
capacity. From an analysis of variance (ANOVA), there was no evidence of lack of fit.
Overall regressions for Cu recovery and Cu concentrate grade using all data (years 2012 to 2015)
were performed. The modelling results were in good agreement with the experimental data. The
most significant variables observed in the previous model were kept in the extended model.
Particle size, P80, in the recovery model and the Cu feed grade in the concentrate grade model were

9
the most significant variables. Multivariate linear model for Mo concentrate grade showed an
adequate predictive capacity (R2 = 0.9). The most significant predictor variables were associated to
the relative content of Mo minerals with regard to Cu and Fe minerals in the raw feed. The model
for the Mo recovery did not show an adequate predictive capacity (R 2=0.27).

ACKNOWLEGEDMENTS
The authors are grateful to Cía. Minera Los Pelambres for providing access to their plant and for
valuable assistance in the experimental work. Funding for process modelling and control research
is provided by CONICYT, Project Fondecyt 1160547, and Federico Santa María Technical
University, Project 116.27.2.

REFERENCES
Bulled, D. and Connolly, J., (2009) A case study in the flotation characterization of an ore body and
its use in circuit design and risk assessment, in Malhotra, D., Taylor, P., Spiller, E., LeVier, M. (Eds.),
Recent Advances in Mineral Processing Plant Design, SME, pp. 3647.
Chong, I. and Jun, C., (2005) Performance of some variable selection methods when
multicollinearity is present, Chemometrics and Intelligent Laboratory Systems, 78 (1–2), pp. 103112.
Dowling, E., Klimpel, R. and Aplan, F., (1985) Model discrimination in the flotation of a porphyry
copper ore, Minerals and Metallurgical Processing, 2, pp. 87101.
Friedman, J.H. and Tukey, J.W., (1974) A projection pursuit algorithm for exploratory data analysis.
IEEE Transactions on Computers, C-23(September), pp.881–890.
García-Zúñiga, H., (1935) La eficiencia de la flotación es una función exponencial del tiempo,
Boletín Minero, Sociedad Nacional de Minería, 47, pp. 8386.
Gosling, J., (1995) Introductory Statistics, Pascal Press, Sydney, Australia, pp. 148-154.
Hunt, J., Berry, R., Bradshaw, D., Triffett, B. and Walters, S., (2014) Development of recovery
domains: examples from the Prominent Hill IOCG deposit, Australia. Minerals Engineering, 64, pp.
7–14.
Jolliffe, I., (2013) Principal Component Analysis, Springer Science & Business Media, Illustrated Ed.,
Kent, England.
Mujica, L., Rodellar, J., Fernández, A. and Güemes, A., (2011) Q-statistic and T2-statistic PCA-based
measures for damage assessment in structures. Structural Health Monitoring, 10 (5), pp. 539–553.
Kor, M., Abkhoshk, E., Gharibie, K. and Shafaei, S., (2010) An investigation of the particle size effect
on coal flotation kinetics using multivariable regression, International Journal of Mining &
Environmental Issues, 1(1), pp. 41–47.
Rath, S., Sahoo, H. and Das, B., (2013) Optimization of flotation variables for the recovery of
hematite particles from BHQ ore, International Journal of Minerals, Metallurgy and Materials, 20(7),
pp. 605–611.

10
Sepulveda, E., Dowd, P. and Xu, C., (2015) Modelling geometallurgical response variables using
projection Pursuit Regression, XI International Mineral Processing Seminar, Procemin 2015,
Santiago, Chile
Wills B. and Finch J., (2015) Chapter 12 - Froth Flotation, in Wills B. and Finch J., Wills' Mineral
Processing Technology, 8th Ed., Butterworth-Heinemann, Boston, pp. 265380.
Wise, B., Gallagher, N., Bro, R., Shaver, J., Windig, W. and Koch, R., (2005) PLS Toolbox 4.0 for use
with MATLABTM, Eigenvector Research, Inc.
Wold, S., Sjöström, M. and Eriksson, L., (2001) PLS-regression: a basic tool of chemometrics,
Chemometrics and Intelligent Laboratory Systems, 58 (2), pp. 109130.
Yianatos, J., Henríquez, F. and Oroz, A., (2006) Characterization of large size flotation cells,
Minerals Engineering, 19 (6-8), pp. 531–538.

11
A Comparison from Pilot Test to Industrial Results:
Implementation of Pneumatic Technology in
Molybdenum Selective Flotation Plant
Patricia Véliz 1, Claudio Muñoz1, Egidio Catalán1 and Heinrich Sprenger2
1. Cía. Minera Doña Inés de Collahuasi, Chile
2. Malegwyn Mineral Services, Germany

ABSTRACT
A Comparison between results of a previous pilot test with pneumatic cells against the preliminary
results of the industrial implementation of pneumatic cells as cleaner stage, was developed in
selective molybdenum flotation plant, at Compañía Minera Doña Inés de Collahuasi.
The molybdenum plant consist on a rougher stage, followed by a first and second cleaner stages with
conventional cells technology. Original design considered a third and fourth cleaner stage with
column cells. Due to a low quality of the molybdenum concentrate achieved with the columns,
especially when hydrophobic clays fed the plant, it was developed a pilot campaign with pneumatic
technologies. Pilot circuit ran with one Primetals® pneumatic cell as a rougher and two G-Cell®-08
pneumatic cells in series as cleaner stages, parallel to the industrial circuit. Results were an increase
of molybdenum grade and a decrease in insoluble recovery. Supported by this results, columns cells
were replaced by Maelgwyn technology, specifically H-Cell® and G-Cell® as third and fourth
cleaner, respectively.
Preliminary results of the industrial circuit were different than expected, because a high insoluble
content was recirculated from pneumatic tail to conventional stages. Based on that, pneumatic tail
was redirected to final tail. Metallurgical results improved with this modification, approaching to
pilot results.
However, there was a high content of molybdenum losses in final tail. Due to that, a treatment was
proposed for pneumatic tail with settle separation equipment that shows good results in a pilot test
recently developed.

1
INTRODUCTION
The selective molybdenum plant, located in Collahuasi’s port, has about 5,000 tpd feed capacity to
receive and process collective concentrate Cu-Mo, which is coming from the concentrator plant. The
circuit consists of a selective flotation in counter flow, it means that tails returns to the previous stage.
The final tail from selective plant, is the copper concentrate, which is filtered and shipped to national
and international market. The Cu-Mo separation process begins with a first rougher stage flotation
with nine 28 m3 cells, followed by a first cleaner with four 14 m3 cells and a second cleaner stage with
two 14 m3 cells. All these cells are self-aerated and encapsulated conventional technology. The
original circuit consisted of a third and fourth cleaner with column cells, obsolete technology replaced
for the actual pneumatic cells provided from MMS Limited.
Molybdenum final concentrate grade is mainly affected by hydrophobic clays, presents as part of the
mineralogy of Collahuasi. Looking for new technologies to separate molybdenite from this type of
clays, it was installed a pneumatic pilot circuit, working in parallel to industrial circuit.
Pilot circuit consisted in a single cell as rougher stage (Primetals® FT02 pneumatic cell) and two cells
in series as cleaner stage (G-Cell®-08 pneumatic cells) (Fig.1). This circuit obtained positive results at
final concentrate molybdenum grade, through an effective separation from high content of clays.
Results shown in Table 1 and 2.

Figure 1 Pilot circuit layout tested

2
Table 1 Metallurgical results from pilot circuit tests
Stage Recovery Enrichment
Mass (%) Mo (%) Cu (%) Ins (%) Mo Cu Ins
Rougher 2,9 51,3 2,4 11,4 17,4 0,5 2,5
1st Cleaner 49,2 79,1 4,7 10,5 1,6 0,4 0,8
2nd Cleaner 56,1 70,8 44,1 54,0 1,3 0,2 0,7
Cleaner 35,2 72,9 2,1 6,0 2,1 0,2 0,5
Pilot Plant 1,1 37,4 0,1 0,7 36,0 0,1 1,3

Table 2 Concentrate grade in samplings

Feed Tail Con Con Tail Feed Con Tail Feed
Grade Ro Ro Ro Cl1 Cl1 Cl2 Cl2 Cl2 Cl1
Mo (%) 1,2 0,7 20,8 35,8 7,7 33,6 45,3 22,0 20,1
Cu (%) 25,5 26,0 13,2 6,4 20,0 5,1 2,2 12,9 15,5
Ins (%) 8,2 8,0 18,4 17,5 20,6 20,3 9,4 30,3 24,3

The purpose of this study is a comparison between pilot result tests at cleaner stages from industrial
results of currently installed pneumatic cells.

Dimensioning and engineering development to third and fourth cleaner industrial

circuit implementation
For the implementation of pneumatic cells as third and fourth cleaner at molybdenum plant, a mass
balance was performed to identify and quantify the different flowrate of the pneumatic Maelgwyn
circuit. At this study established a line of three cells H-Cell® (hybrid feed) for the third cleaner and a
line of three cells G-Cell® for fourth cleaner.
Engineering considered two main circuits, depending on the quality of the concentrate feed to the
molybdenum plant. The first one is to use hydrocyclone cluster before feeding the pneumatic cells to
separate a significant quantity of insoluble present in the ultrafine particle size fraction. A lot of this
corresponds to hydrophobic clays, however this circuit involves to decrease final molybdenum
recovery. A second circuit did not consider to use hydrocyclones, and pneumatic cells are feeding
with second cleaner concentrate, directly (Fig.2). See results expected on both circuits in Table 3.
It is important to mention that currently there is developing an over haul at hydrocyclone cluster,
this is the reason why third and fourth cleaner Maelgwyn circuits presented below, do not include
the use of hydrocyclones in any of its configurations.

3
 Figure 2 Proposed circuits (a) No use of hydrocyclone circuit (b) Use of hydrocyclone circuit

Table 3 Expected metallurgical performance of industrial pneumatic stages

Stage pyrophillite presence Cy On No pyrophillite & Cy off
Mo Recovery Mo grade Mo Recovery Mo grade
Rougher 86,3% 13,2% 88,7% 13,8%
1° Cleaner 55,0% 12,1% 78,4% 21,2%
2° Cleaner 93,9% 22,5% 72,5% 30,8%
3° Cleaner 33,2% 21,7% 74,9% 32,7%
4° Cleaner 50,9% 26,1% 63,1% 40,6%
Overall 74,8% 26,1% 84,6% 40,6%

The implemantation of Maelgwyn cells was located at molybdenum plant near conventional selective
flotation circuit and commissioning started in February 2016 (Fig. 3 and 4).

4
 Figure 3 Pneumatic Maelgwyn general layout (a).Conventional plant layout (b)

Figure 4 Pneumatic Maelgwyn third and fourth cleaner installed in Collahuasi molybdenum plant

5
METODOLOGY

Sampling procedure
In order to characterize and evaluate the entire Maelgwyn cells system, each of its components, a
sampling program was performed to determinate Mo, Cu and Insoluble grade in all stream.
Additionally, pH and potential (ORP) were measured, key parameters to reagents control and doses.
This program consisted about four months to evaluate each circuit used, with a frequency of three
hours to sampling only in the main stream as operational control, and twice daily frequency sampling
to each point of the circuit to evaluate each cell performance (Fig. 5)
Sampling point were defined according to the circuits used. Third cleaner as only one line and both
lines as third and fourth cleaner as Table 4 shows
Additional sampling was taken from feeding at third cleaner to make a characterization of particle
size distribution and Mo, Cu and insoluble grade and grade distribution by particle size.

Figure 5 General diagram sampling points third and fourth cleaner Maelgwyn

6
Table 4 Sampling points

Circuit Sampling Stream

1 3°Cl Feed
2 3CC1 Concentrate
3 3CC1 Tail
4 3CC2 Concentrate
3rd Cleaner
5 3CC2 Tail
6 3CC3 Concentrate
7 3CC3 Tail
8 3° Cl Concentrate
9 4° Cl Feed
10 4CC1 Concentrate
11 4CC1 Tail
12 4CC2 Concentrate
4th Cleaner
13 4CC2 Tail
14 4CC3 Concentrate
15 4CC3 Tail
16 Final Concentrate (4°Cl)

Defined circuits
The circuit configuration for cleaner stage operation with a single line as third cleaner and two lines
as third and fourth cleaner is shown below.

Third cleaner
1. Design circuit. Third cleaner feeds with second cleaner concentrate .Third cleaner tail returns to
second cleaner (Fig. 6)

Figure 6 Third cleaner design circuit

7
2. Modified circuit. Third cleaner is feeds with second cleaner concentrate .Third cleaner tail do not
returns to second cleaner. (Fig. 7)

Figure 7 Third cleaner modified circuit

Third and fourth cleaner

3. Design circuit. Third cleaner feeds with second cleaner concentrate .Third cleaner tail returns to
Second cleaner. Fourth cleaner tail returns to third cleaner. (Fig. 8)

Figure 8 Design circuit third and fourth cleaner

8
4. Modified circuit. Third cleaner is feeds with second cleaner concentrate .Third cleaner tail do not
returns to second cleaner. Fourth cleaner tail returns to third cleaner. (Fig. 9)

Figure 9 Modified circuit third and fourth cleaner

RESULTS AND DISCUSSION

The circuits described above showed different results in molybdenum, copper and insoluble grade.
Similarly, metallurgical results show differences in recovery and enrichment ratio, depending on the
configuration used.
It is important to consider that during entire period of start-up, the quality of concentrate was low,
because of high content of hydrophobic clays. In some periods, molybdenum grade was higher at
selective plant feeding, but the presence of hydrophobic clays was constant. This does not allow to
evaluate the circuit with a high quality of concentrate fed to selective plant.
Another important point is that the hydrocyclone stage did not run at any configuration, so the real
results differ from the expected when the hydrophobic clays was present.

Third cleaner metallurgical results

The results obtained at industrial level, during a single cell line operation, as third Maelgwyn cleaner,
are shown below.

Design circuit third cleaner

This circuit was fed from second cleaner concentrate, and third cleaner tail returns to conventional
circuit (rougher to second stages), therefore recirculation of insoluble was very high as shown in
Table 5.

9
Table 5 Third cleaner design circuit molybdenum, copper and insoluble grade

Stream %Mo %Ins %Cu

Con 2nd Cleaner (Feed 3rd cleaner) 24,4% 43,8% 1,2%
Tail 3rd Cleaner (Feed 2nd cleaner) 21,4% 45,6% 1,2%
Con 3rd Cleaner (Final concentrate) 25,3% 44,8% 1,1%

Table 6 Third cleaner design circuit metallurgical results

Recovery % Enrichment Ratio

Metallurgical %Rec Mass % Rec Mo % Rec Ins % Rec Cu E.R Mo E.R Ins E.R Cu
behavior
3rd Cleaner 74,5% 77,2% 75,6% 67,2% 1,0 1,0 0,9

100% (a) (b)

80%
60%
40%
20%
0%
Mass Mo Ins Cu

Figure 10 Design third cleaner circuit metallurgical results. Recovery (a) Enrichment ratio (b)

Molybdenum recovery at design third cleaner circuit is 77,2%. Enrichment ratio is 1,0 as shown in
Table 6 and Fig. 10. Low molybdenum enrichment ratio results were associated to effect of high
circulating load of hydrophobic clays from third cleaner tail to the conventional circuit, not allowing
to come out within the final tail.

Modified circuit third cleaner

This circuit was fed from second cleaner concentrate. Third cleaner tail does not returns to
conventional circuit. Grade results are shown in Table 7.

Table 7 Third cleaner modified circuit molybdenum, copper and insoluble grade

Stream %Mo %Ins %Cu

Con 2nd Cleaner (Feed 3rd cleaner) 23,0% 49,6% 2,4%
Tail 3rd cleaner (Feed 2nd cleaner) 18,4% 54,2% 2,8%
Con 3rd Cleaner (Final concentrate) 29,1% 43,0% 1,5%

10
Table 8 Third cleaner modified circuit metallurgical results

Recovery % Enrichment Ratio

Metallurgical %Rec Mass % Rec Mo % Rec Ins % Rec Cu E.R Mo E.R Ins E.R Cu
behavior
3rd Cleaner 60,4% 71,5% 55,7% 50,7% 1,3 0,9 0,8

(a)
100%
80%
60%
40%
20%
0%
Mass Mo Ins Cu

Figure 11 Third cleaner modified circuit metallurgical results. Recovery (a) Enrichment ratio (b)

Molybdenum recovery at modified third cleaner circuit is 71,5%, lower than design circuit.
Enrichment ratio is 1,3, higher than design circuit as shown in Table 8 and Fig. 11. The above results
about recovery were associated to molybdenum losses from Maelgwyn tail to final tail in selective
flotation plant, therefore enrichment results were associated to a discard of hydrophobic clays
through third cleaner tail going to final tail, and it does not increase insoluble grade at conventional
stages as design circuit does.

Third and fourth cleaner Metallurgical results

The results obtained at industrial level during both cells lines are in operation, as third and fourth
Maelgwyn cleaner, are shown below.

Design circuit third and fourth cleaner

This circuit was fed from second cleaner concentrate, and third cleaner tail returns to conventional
circuit (rougher to second stages). Finally fourth cleaner tail returns to third cleaner. Grade results
are shown in Table 9.

Table 9 Third and fourth cleaner design circuit molybdenum, copper and insoluble grade

Stream %Mo %Ins %Cu

Con 2nd Cleaner 27,7% 31,9% 1,8%
Tail 3rd Cleaner (Feed 2nd Cleaner) 23,0% 47,9% 1,8%
Con 3rd Cleaner (Feed 4th Cleaner) 33,5% 37,3% 1,7%
Tail 4th Cleaner (Feed 3rd Cleaner) 25,5% 45,2% 1,5%
Con 4th Cleaner (Final concentrate) 39,0% 34,0% 1,5%

11
Table 10 Third and fourth cleaner design circuit metallurgical results

Recovery % Enrichment Ratio

Metallurgical %Rec Mass % Rec Mo % Rec Ins % Rec Cu E.R Mo E.R Ins E.R Cu
behavior
3rd Cleaner 39,4% 56% 65% 54% 1,2 1,1 0,9
4th Cleaner 64,7% 76% 54% 39% 1,2 1,0 0,9
Global 30,1% 42% 44% 31% 1,5 1,0 0,8

100% (a)
80%
60%
40%
20%
0%
Mass Mo Ins Cu

Figure 12 Third and fourth cleaner design circuit metallurgical results. Recovery (a) Enrichment ratio (b)

Molybdenum recovery at design third and fourth cleaner circuit is 42%, which is very low.
Enrichment ratio is 1,5 as shown in Table 10 and Fig. 12. A high insoluble recovery was obtained in
this circuit.
Molybdenum recovery decreases because fourth cleaner feed has a low flowrate from third cleaner
and a low mass of concentrate is generated in fourth cleaner. Molybdenum enrichment ratio is higher
than third cleaner, but lower than expected because of the same recirculation effect of hydrophobic
clays that was mentioned before in third cleaner design circuit.

Modified circuit third and fourth cleaner

This circuit was fed from second cleaner concentrate, and third cleaner tail does not returns to
conventional circuit (rougher to second stages). Fourth cleaner tail returns to third cleaner. Grade
results are shown in Table 11.

Table 11 Third and fourth cleaner modified circuit molybdenum, copper and insoluble grade

Stream %Mo %Ins %Cu

Con 2nd Cleaner (Feed 3rd cleaner) 18,3% 50,5% 3,9%
Tail 3rd Cleaner 9,1% 58,5% 5,3%
Con 3rd Cleaner (Feed 4th cleaner) 26,9% 48,9% 1,9%
Tail 4th Cleaner (Feed 3rd cleaner) 18,3% 62,6% 2,6%
Con 4th Cleaner (Final concentrate) 35,2% 35,7% 1,4%

12
Table 12 Third and fourth cleaner modified circuit metallurgical results

Recovery % Enrichment Ratio

Metallurgical %Rec Mass % Rec Mo % Rec Ins % Rec Cu E.R Mo E.R Ins E.R Cu
behavior
3rd Cleaner 51,5% 76% 47% 28% 1,5 0,9 0,5
4th Cleaner 51,1% 67% 37% 36% 1,3 0,7 0,7
Global 35,1% 68% 25% 12% 1,9 0,7 0,4

100% (a)
80%

60%

40%

20%

0%
Mass Mo Ins Cu

Figura 13 Third and fourth cleaner modified circuit metallurgical results. Recovery (a) Enrichment ratio (b)

Molybdenum recovery in third and fourth cleaner modified circuit is 68%. Enrichment ratio is 1,9. as
shown in Table 12 and Fig. 13. This circuit show a higher molybdenum recovery and enrichment ratio
than design third and fourth cleaner circuit, because third cleaner tail does not returns to
conventional stage. Insoluble enrichment ratio and recovery is lowest obtained.

Comparison of results obtained during the pilot test and industrial level with
Maelgwyn cells third and fourth cleaner
The results obtained at G-Cell®-08 (G-Cell 02 and G-Cell 03) during the pilot test are comparable to
results obtained in Maelgwyn cells system, implemented at industrial level as third and fourth
cleaner.
Pilot circuit configuration consisted of a pneumatic cell as rougher that feeds two G-Cell®-08
(pneumatic cells) in series. The industrial circuit with pneumatic third and fourth cleaner cells, feeds
from conventional cells. However molybdenum grade at pilot circuit feeding is similar to the
industrial circuit, not the same with copper and insoluble (Table 13).

Table 13 Feed molybdenum grade at pilot and industrial circuits

Stream %Mo %Ins %Cu

Feed 3rd Cleaner 24,7% 43,4% 2,5%
Feed Pilot Cleaner 20,8% 18,4% 13,2%

13
 100%

Cumulative Passing (%)

(a)
80%

60% 3C Feed

40% Pilot Cleaner Feed

20%

0%
1 10 µm 100 1000

60%
60% (b) (c)
50% Mo Ins Cu
50% Mo Ins Cu
Metal Content (%)

40% 40%

30% 30%

20% 20%

10% 10%
0% 0%

Figure 14 Feed particle size distribution (a) Comparative particle size (b) Pilot cleaner feed grade
distribution (c) Industrial cleaner feed grade distribution

A similar particle size distribution is observed in both feeds (75-77 % -325#). However industrial
circuit feed has a higher content of molybdenum and insoluble at ultrafine fraction than pilot circuit,
which makes difficult their separation.
Third cleaner feed has a higher insoluble grade than pilot cleaner feed, because conventional
technology increases hydrophobic clays enrichment, sometimes even more than molybdenite.
In case of copper, pilot cleaner feed has a higher grade than third cleaner feed, but it is a low copper
grade for a rougher concentrate compared to conventional cells. It is important to consider that pilot
rougher stage was just one pneumatic cell (Table 14).

Table 14 Rougher grade at conventional cells compared with pneumatic pilot cell

Stream %Mo %Ins %Cu

Conventional rougher con 13,8% 19,4% 20,6%
Pneumatic pilot rougher con 20,8% 18,4% 13,2%

About the operation condition, there are some differences between pilot circuit parameters and
industrial operation parameters, the most important is feed pressure. The design G-Cell®-08
established 2 bar as optimal operation pressure. It was possible to get at pilot circuit, but at industrial

14
level the highest pressure obtained was 1,5 bar. Actually, it is under evaluation a new Aerator design
to be implemented.

Third and fourth cleaner comparison results from pilot test with G-Cell® -08
Molybdenum recovery at third cleaner design circuit (as one cleaner stage) is similar to recovery
obtained in G-Cell 02, from pilot circuit 79,1%. Lower recovery is obtained in third cleaner modified
circuit because third cleaner tail does not returns to second cleaner.
Molybdenum enrichment ratio at third cleaner design circuit is very low compared with a 1,6 from
G-Cell 02. It is because insoluble returns to conventional circuit and increase in recovery and grade
affecting molybdenum final grade. Third cleaner modified circuit enrichment ratio is higher to design
circuit but lower than G-Cell 02. The reason was explained before, the high insoluble content at
conventional second cleaner industrial (third cleaner feeding) impact in increase of insoluble
enrichment ratio.
Third cleaner mass recovery is higher in design and modified circuit. This is because a high recovery
of other elements (Cu, Ins included). Comparative third cleaner results are shown in Fig. 15.

1,8
100% (a) (b)
1,6
80% 1,4
1,2
60% 1
0,8
40%
0,6
Pilot
20% Pilot 0,4
3C Design
3C Design 0,2
3C Modify 3C Modify
0% 0
Mass Mo Ins Cu Mo Ins Cu

Figure 15 Comparative results from pilot G-Cell 02 and third cleaner circuits (a) Recovery (b) Enrichment ratio

(a) 2,4 Pilot (b)

100% Pilot
2,2 3C-4C Design
3C-4C Design 2 3C-4C Modify
80% 3C-4C Modify 1,8
1,6
60% 1,4
1,2
40% 1
0,8
20% 0,6
0,4
0,2
0% 0
Mass Mo Ins Cu Mo Ins Cu

Figure 16 Comparative results from pilot (G-Cell 02 y G-Cell 03) and different industrial cleaner results (Third
and fourth cleaner) (a) Recovery (b) Enrichment ratio

15
In the case of third and fourth cleaner (as two cleaner stages) molybdenum recovery in design circuit
is very low compared to 72.9 % obtained at pilot cleaner (G-Cell 02 and G-Cell 03). This is because of
the low flowrate that feeds fourth cleaner, and a small quantity of concentrate generated as final
concentrate. A high recovery of insoluble was obtained at this circuit. At modified circuit
molybdenum recovery increase and insoluble recovery decrease considerably from 44% to 25%.
Molybdenum enrichment ratio at third and fourth cleaner design circuit is lower than pilot cleaner
circuit, which is 2,1. Modified circuit increase molybdenum enrichment ratio and decrease in
insoluble enrichment ratio. Comparative third and fourth cleaner results are shown in Fig. 16.
Results presented above proved that are necessary two pneumatic cleaner stages as third and fourth
cleaner to obtain results in molybdenum final grade. The high content of hydrophobic clays in third
cleaner feeding implies not recirculate third cleaner tail to conventional circuit.
Finally, third and fourth cleaner modified circuit obtained most similar results compared to pilot
cleaner circuit in terms of molybdenum recovery and enrichment ratio.

Advantages and disadvantages of third and fourth pneumatic cleaner and it solution
The results presented above show an effective solution to main molybdenum Collahuasi plant
problem, which is separate molybdenum from insoluble. The pneumatic technology allows to
increase Mo grade in final concentrate, and send insoluble to the third cleaner tail, but this insoluble
generate a high recirculation to conventional stages and it cannot be eliminated through final tail.
Hydrophobic clays have properties as natural floatability even more than molybdenite. On the other
hand, a high content of molybdenite and clays are distributed in the same fractions of particle size
(ultrafine) at second concentrate making difficult the separation. That is the reason of high
molybdenum grade at third cleaner tail and molybdenum losses in final tail.
The solution for this problem is a project called “operational flexibility” which include
implementation of hydrocyclone cluster and Hydroseparator as a new technology designed to
operate feeding with different streams from molybdenum plant, depending of concentrate quality.
This technology is based in the difference between hydrophobic clays density and molybdenite
density. It will be able to process tails or hydrocyclone overflow to increase the final plant recovery
and molybdenum grade.

CONCLUSION
The metallurgical performance of G-Cell® technology implemented as third and fourth cleaner was
evaluated and compared from pilot test results. Pilot test obtained an effective increase of
molybdenum grade and decrease of insoluble grade at second cleaner concentrate.
It was found that design third and fourth cleaner circuit had a high recirculation effect of insoluble,
because third cleaner tail returns to conventional circuit (Rougher to second cleaner), therefore
metallurgical results were lower than pilot test (low molybdenum enrichment and recovery).
A modified circuit of third and fourth cleaner was performed to not return third cleaner tail to
conventional circuit. Results were an increase of molybdenum grade and decrease of insoluble grade
at fourth cleaner concentrate. These results were very similar to pilot test.
Modified circuit obtained a high enrichment ratio, but molybdenum recovery remains low. A
solution to this problem is the implementation of a new project that will treat third cleaner tail to
increase molybdenum recovery and eliminate definitely insoluble content at final tail.

16
ACKNOWLEDGEMENTS
We would like to thank to Metallurgical department and Process team from Collahuasi port for its
active role in the development of the project. Similarly, we would like to thank the company
Maelgwyn Mineral Services for the technology implemented.

NOMENCLATURE
MMS Maelgwyn Mineral Services Limited
Ro Rougher stage
Cl1 First cleaner stage
Cl2 Second cleaner stage
3°Cl Third cleaner stage
4°Cl Fourth cleaner stage
3C Third cleaner Maelgwyn circuit
3CC1 Third cleaner Maelgwyn cell 1
3CC2 Third cleaner Maelgwyn cell 2
3CC3 Third cleaner Maelgwyn cell 3
4CC1 Fourth cleaner Maelgwyn cell 1
4CC2 Fourth cleaner Maelgwyn cell 2
4CC3 Fourth cleaner Maelgwyn cell 3
4C Fourth cleaner Maelgwyn circuit
Con Concentrate
Cy Hydrocyclone
Mo Moybdenum
Cu Copper
Ins Insoluble
TK101 Feed tank to Maelgwyn cleaner
TK1501 Feed tank to third cleaner Maelgwyn
TK1504 Feed tank to fourth cleaner Maelgwyn
G-Cell 02 Pilot first cleaner pneumatic cell to test denomination
G-Cell 03 Pilot second cleaner pneumatic cell to test denomination

REFERENCES
Imhof R., Fletcher M., Vathvooran A., Singh A. (2007) Application of Imhoflot G-Cell centrifugal flotation
technology. The journal of the Southern African Insitute of mining and metallurgy.
Castro S., Castro Ingeniería Ltda. (2013) Mejoramiento metalúgico Planta Premo Collahuasi.
Battersby M., Fletcher M., Imhof R., Singh A., Puder F., (2015) The advantages of the imhoflot G-Cell
pneumatic Flotation Process with Centrifugal Froth Removal –Two case study
Castro S., Lopez-Valdivieso A., S. Laskowski J. (2016) Review of the flotation of molybdenite. Part I:
Surface properties and floatability

17
Farrokhpay S. , Bradshaw Dee. (2012) Effect of clay minerals on froth stability in mineral flotation: A review.
XXVI International mineral processing congress (IMPC) 2012 proceeding/ New Delhi. India 24-28
Sept 2012
Battersby M., Battersby R. M., Flatman S., Imhof R., Sprenger H., Bragado T.(2011). Recovery of Ultra
Fines Using Imhoflot Pneumatic Flotation – Two Pilot Plant Case Studies Recovering Nickel and Zinc from
Tailings Stream.
Gonzalez, L.R,. (2002). Pneumatic flotation study as a pre rougher cell at the molybdenum plant. Technical
report. Chuquicamata, Chile
Chander, S., Wie, J.M., Fuerstenau, D.W., 1975. On the native floatability and surface properties of
naturally hydrophobic solids. In: Somasundaran, P, Grieves, R.B. (Eds.), Advances in Interfacial
Phenomena of Particulate/Solution/Gas Systems; Applications to Flotation Research, AIChE Symp.
Series, 150, Vol. 71
Sanchez-Pino, S., Sanchez-Baquedano, A., Imhof, R.M., (2008). Imhoflot G-Cell Pneumatic Flotation
Technology for Fine Particle Applications. Procemin 2008 V International Mineral Processing Seminar,
22-24 October 2008, Santiago, Chile.)

18
Predicting the Energy Requirements of High Pressure
Grinding Rolls Using Piston Press Test Procedures
Zorigtkhuu Davaanyam, Bern Klein, Stefan Nadolski and Amit Kumar
University of British Columbia, Canada

ABSTRACT
The High Pressure Grinding Rolls (HPGR) is a relatively new comminution technology that has
been shown to bemore energy efficient compared to semi-autogeneous (SAG) mills. Full and fair
consideration cannot be given to HPGR-based comminution circuits in early stageprojects, because
presently pilot scale testing is required for assessing and sizing; pilot scale testing requires bulk
samples weighing more than one tonne. A CAMIRO research program was conducted at the
University of British Columbia to develop bench scale tests that would reduce the quantity of
sample required to assess HPGR.
Three test procedures were developed.
1. The Direct Calibration Methodology calibrates piston press test results to pilot scale test
results such that piston press tests can then be used to determine the energy–size
reduction relationship for a range of ore types.
2. The Database Calibrated Methodology, which is similar to the Direct Calibration
Methodology except that it calibrates piston press tests against a database of pilot scale
and piston press tests.
3. TheSemi-Mechanistic Methodology wasdeveloped to use piston press test results as an
input for an HPGR based energy-size reduction performance model.
Through utilization of these methodologies, the energy–size reduction performance of an HPGR
can be predicted with minimal sample requirement. In addition, piston press testing is far more cost
effective, faster, and accessible than pilot-scale HPGR testing.

1
INTRODUCTION
One could argue that in the future, decisions regarding selection of comminution circuits will be
made primarily on the basis of energy consumption. The following facts and predictions support
the argument:
 Comminution processes consume a significant proportion of energy expended in a
mining operation due to its inherent inefficiencies;
 Demand for energy is predicted to continue to rise, which willdrive the cost of energy
higher as a clear majority of energy sources are non-renewable;
 Coal is the main energy source for electricity generation; coal burning power plants
release greenhouse gases(GHG) and governments are starting to implement policies to
combat GHG generation such as the carbon tax introduced in British Columbia.
On average, comminution accounts for 55% of total energy consumption in metal mining
operations(U.S. Department of Energy, 2007). When the amount of energy expended in mining
activities is excluded, comminution accounts for 66-80% of the total electrical energy consumption
in milling processes (Abouzeid & Fuerstenau, 2009; Claffin, Leonard, Rogers, & Urritia, 2009).
Comminution is energy-intensive; its efficiency is in the range of 1-2% in terms of the amount of
energy required to generate new fracture surface area (Tromans, 2008).
The U.S. Energy Information Administration (2011) forecasts that global energy consumption will
grow 47% from 522 quadrillion Btu in 2010 to 770 quadrillion Btu in 2035.In 2010, 75% of global net
electricity was generated from non-renewable sources, which are predicted to account for 74% or
more of electricity generation until 2040. For example, coal is the source of 40% of net world
electricity generation and is predicted to remain the main source for electricity generation until
2040.Simple economic principle suggests that increase in energy demand will drive energy costs
higher.
Governments around the world started to introduce policies to reduce GHG emissions. An example
of this policy is the carbon tax in British Columbia that started at $10 per tonne of CO 2 in 2008 and
has been increased to $30 per tonne CO2 in 2014 (Beaty, Lipsey, & Elgie, 2014).
HPGR technology has been shown to be moreenergy efficient solution to preparing feed for ball
mills, compared to traditional semi-autogenous(SAG) mills. Yet the technology has not gained
wide industry acceptance. One of the reasons for slow adaptation of the technology is identified as
a lack of small-scale test for sizing and selection. Full and fair consideration cannot be given to the
technology until pilot-scale test work is completed on a large volumes of sample weighing 1 to 10
tonnes. The costs associated with collecting these large sample quantities are usually prohibitive for
early stage developers.
Driven by the objective of reducing the sample quantity required to generate HPGR performance
data, piston press test procedures have been developed and are presented in this paper.

2
EXPERIMENTAL

HPGR Testing
A KoeppernHPGR, located at the University of British Columbia, was used for HPGR testing. The
rolls of the unit are 750mm in diameter and 220mm in width. The standard procedure for sizing
and selection used by the manufacturer was followed for pilot-scale HPGR testing. Feed materials
were stage crushed down to a top-size of 32mm or less. In case of multiple tests, all feed materials
were homogenized prior to splitting the material into individual 200 liter drums for testing. Two
sub-samples (10-15kg) were taken: one is used for determining feed particle size distribution (PSD),
moisture content, and bulk density measurements and the other is used for piston press testing.
A data-logger connected to the fully instrumented HPGR was used to record roll speed, torque,
power draw, and operating gap.Subsequently, these machine data and sample weights were used
to calculate net specific energy consumption and specific throughput constant (m-dot) values.
Centre and edge products of HPGR were collected separately and sub-samples were taken from
each bulk product to carry out particle size analyses.

Piston Press Testing

HPGR feed subsample is stage crushed to -12.5 mm; the top size was nominated to minimize
boundary effects. For each piston press test a 240 ml volume of sample is weighted anplaced in
86 mm hardened steel die and press by MTS hydraulic press. The piston position is reset to zero at
the rim of the die and its movement is recorded as the hydraulic press applies seating load of
2.5 kN force to the sample. Once the seating load is reached hydraulic press ramps up at
200 kN/min up to 1400 kN, which is equivalent of 240 MPa pressure.
Total energy input was calculated by numerical integration of the recorded force-displacement
curve. Prior to testing, the mechanical strain of the piston press system was measured and
subtracted from test measurements. Specific energy input to the sample was found by dividing the
total applied mechanical energy by the weight of the sample contained in the piston die. For each
piston press test, 240 ml of sample was used. Following testing, pressed sampleswere taken out of
the die by removing the die base plate andused for wet and dry sieving to determine the resulting
particle size distribution.

RESULTS AND DISCUSSION

Direct calibration between HPGR and Piston Press tests

Based on numerous test results and published data, HPGR specific energy consumption is known
to linearly increase with specific pressing force, which is the total hydraulic force divided by roll
diameter and width — expressed in units ofN/mm2. In piston press tests, specific energy input
increases linearly with piston pressure. Figure 1 shows an example of specific energy linearly
increasing with both specific pressing force and piston pressure; it also illustrates how the two
piston press is calibrated using HPGR results. For example, in the figure it can be seen that when a
piston pressure of 190 MPa is used with the piston press and a specific pressing force of 4.0
N/mm2is applied with the pilot HPGR, both comminution devices result in a specific energy
expenditure of 2.0 kWh/t.

3
 Figure 1 Illustration of calibration between specific pressing force and piston pressure; such that both input
the same specific energy to the sample

Once calibration information is available for a composite sample, specific energy can be predicted
within ±10% as shown in Figure 2. This level of accuracy is considered to be excellent for scoping
and pre-feasibility level studies. Furthermore, coordinating smaller volume piston press tests,
rather than compositing samples for pilot testing, allows ore domains to be defined at a higher
resolution for geometallurgical mapping.

Figure 2 Specific energy predictions with ore-specific calibration information

As mentioned previously, HPGR tests were performed on -32 mm material and piston press tests
were conducted on -12.5 mm feed. Therefore, resulting reduction ratios could not be directly
compared for the following reasons:
 Multiple HPGR tests performed on feeds with 32 mm and 12.5 mm top sizes showed that
the reduction ratio achieved decreased with feed top size, even though the same specific
energy levelwas recorded, as shown in Figure 3;

4
  In the case where sample having the same feed size is used for both pilot HPGR and
piston press testing, the two products have been found to be similar, as shown in Figure
4. Particularly, when HPGR edge and centre products are mixed at weight proportions of
10% and 90% respectively. This ratio of edge and centre products is nominated based on
typical aspect ratios of larger HPGR rollers and is considered to represent the HPGR
product of an industrial scale machine.

Figure 3 Comparison of reduction ratio achieved in HPGR tests on feeds with varying top size

Figure 4 Comparison of products from HPGR tests and piston press tests

When reduction ratios achieved in HPGR tests and piston press tests are calibrated on the basis of
specific energy input, prediction of reduction ratio is within ±10%, as shown in Figure 5.

5
 Figure 5 Scaled-up reduction ratio of piston press test compared with reduction ratio of HPGR test

HPGR and piston press tests both produce products that can be normalized by their median size.
Figure 6a compares four HPGR test products which were conducted at four different pressing force
setpoints. As expected, the highest pressing force of 5 N/mm2produced the finest product.
However, once the PSDs are normalized by their respective median sizes, they overlap and can be
represented by a single fitted curve, as shown in Figure 6b. The resulting normalized product PSD
is considered to be a material characteristic and is a critical outcome of the testing regimen.
Equation 1 is functional form of the fitted curve suggested by Limet al.(1996).

( ( ) )
( ) ( ( ( ) )) (Eq. 1)

where
F is cumulative percent passing
x/X50 is the dimensionless particle size normalized by median size
A, m, and n are fitted parameters.

6
 Figure 6 Comparison of regular product PSDs and normalized product PSDs

Normalized product PSDs of HPGR tests on 32 mm and 12.5 mm feed samples, and a piston press
test on a 12.5 mm sample overlap as shown in Figure 7. The implication is that normalized product
PSDs that have beendetermined from piston press testing can be used to predict the normalized
product PSDs which would result should the same sample type be processed with an HPGR.
Therefore, a product PSD can be determined using equation 1 with parameters that have been fitted
to the normalized product PSDs.

Figure 7 Comparison of normalized product PSDs of HPGR test on 32mm and 12.5mm and piston press test
on 12.5mm

7
The Direct Calibration methodology has the following five steps:
1. Conduct pilot-scale HPGR tests and piston press tests on the composite feed.
2. Calibrate piston pressure against specific pressing force such that they result in an
equivalent energy input.
3. Calibrate the size reduction achieved inpiston press tests against the size reduction
achieved in HPGR tests.
4. Compare the normalized product PSDs and obtain a functional form by curve fitting
equation 1 to the normalized product PSDs.
5. Perform piston press tests on individual geo-metallurgical units and scale-up the results
to HPGR pilot-scale.

Database-Calibrated Methodology
The Direct Calibration Methodology was applied to 15 different ore types. Since the ore types were
considerably diverse, it was possible to model a statistically significant empirical relationship
between piston pressure and specific pressing force. The empirical model, shown in equation 2, is
used to calculate the piston pressure required to conduct piston press tests when HPGR test data is
not available for direct calibration.
Eq. 2 is used to calculate the required piston pressure to input the same specific energy to the
sample as in HPGR tests.
(Eq. 2)

where
P is the estimation of required piston pressure in MPa;
FSP is specific pressing force in N/mm2;
u is moisture content in %;
ρb is bulk density;
F50 are 50% passing feed particle sizes for HPGR and piston press tests
Figure 8 shows that conducting piston press tests at the modelled piston pressure provides a good
prediction of specific energies that are mostly within ±25% of the measured value. For projects in
early stages of development and for scoping level of studies it is an acceptable level of accuracy.

8
 Figure 8 Specific energy prediction conducting piston press tests at the modelled piston pressures

Equation 3, shown below, is used to scale-up reduction ratios achieved with piston press tests on
12.5 mm material to HPGR pilot-scale for a-32 mmtop size feed sample.
(Eq. 3)

where
RRH is the estimated reduction ratio of an HPGR test;
RRP is the reduction ratio of a piston press test;
u is the moisture content of feed;
F50 is 50% passing particle size of feed to HPGR and piston press tests.
The database-calibrated methodology involves the following four steps:
1. Using the empirical model for piston pressure, calculate the required piston pressure for
carrying out piston press tests;
2. Perform the piston press tests at the calculated piston pressures and obtain specific
energy and reduction ratio values;
3. Apply the empirical model for reduction ratio to scale-up the reduction ratios achieved
during piston press testing to HPGR pilot-scale;
4. Determine the resulting product size distribution by using the predicted P50,available
from the reduction ratio model, and the normalized PSD curve of piston press product.
Thus, piston press test provides prediction for both the specific energy requirement and the
product PSDs.

Simulation Methodology
The Direct Calibration and Database-Calibrated Methodologies are empirical in that they require
either HPGR test data for calibration or the empirical models for: (1) piston pressure calculation and

9
(2) scale-up of reduction ratios achieved in piston press tests. In their presented form, these
empirical methodologies cannot be used for simulation of closed-circuit operation of HPGR.
The Simulation Methodology is semi-empirical in that the energy–breakage relationship of the ore
is characterized by performing piston press tests on mono-size particles. Then the ore
characteristics information is used to simulate its energy–size reduction performance of HPGR at a
phenomenological level.
The method involves piston pressing five classes of mono-size particles at three different energy
levels, resulting in 15 unique tests. The data points are used to define the relationship between
specific energy input and the degree of breakage, t 10, defined asthe proportion of material which
passes 1/10th of the geometric size of the mono sized feed. Generally, higher specific energy input
results in a higher t10 value. In other words, a greater degree of breakage is achieved. The procedure
of testing multiple mono-size particle lots at multiple energy levels is analogous to the JK
DropWeight test method. A significant difference lies in the breakage mechanism being utilized for
the test methods:slow compression of a particle bed is carried out in a piston press test, whereas the
JK impact breakage of single particles is performed for the JK DropWeight test. The degree of
breakage via piston press, t10 is modelled as a function of specific energy and square root of particle
size, as in equation 4, which is similar to the model proposed by Shi and Kojovic(2007).
( ( √ )) (Eq. 4)

where
t10 is percentage passing 1/10th of the original mono-size particle and represents the degree
of breakage after energy input [%]
M is a fitted parameter and represents the maximum attainable value of t10 [%]
fmat is a fitted parameter and represents the material properties [t/kWh∙mm -0.5]
x is particle size [mm]
ESP is specific energy input [kWh/t]
The measured t10 values along with the fitted curve, plotted using equation 4, is shown in Figure 9.

Figure 9 Measured t10 values of piston press tests on a gold ore and the fitted curve

10
Other tn values, which represent the percentage passing 1/nth of the original particle size, can be
similarly found from product size distributions from piston press testing. Plotting t n values against
t10 values for a database oftested ores shows that master curves can be used to represent piston
press product size distributions, shown in Figure 10. An important implication of having a set of
master curves for breakage appearance is that a PSD can be reconstituted just with the knowledge
of t10; each value of t10 in Figure 10 represents a PSD. The master curve of compression breakage has
significantly lower t2 and t4 than suggested master curves for single particle impact breakage
(Morrison & Morrell, 1997).

Figure 10 Master curves of breakage appearance

In order to quantify the influence of fines being present in piston press feed on comminution, where
fines are defined as particles less than 0.5 mm in size, piston press tests were performed on mono-
size particles in the presence of fines. The proportion of fines ranged from 0 up to 75% by weight
for the coarsest -12.5+11.2 mm mono-size particles. Through piston press testing it was found that
equation 5 adequately predicts the t10 value when fines are included in the feed. The same values
for M and fmat that were model fitted to piston press tests conducted without fines could still be
used. However, the inclusion of fines in feed had a considerable impact on the maximum attainable
breakage, represented by M in equation 4. To account for this effect, equation 4 was modified as
shown in equation 5. Figure 11 shows the measured and modelled t10 values for equation 5.
( )( ( √ )) (Eq. 5)

wherec is a fitted constant to multiply the percentage of fines in the feed, qf.

11
Figure 11 Measured and modelled t10 when the piston press tests are conducted in presence of finesFigure
12 shows a schematic of the model used for simulating breakage in an HPGR. The feed to HPGR is
split into two fractions according to the pre-crushing critical size. Coarser particles contained in the
feed undergo pre-crushing while finer particlesavoid pre-crushing and combine with theprogeny of
the coarser feed particles.Subsequently, the combined material goes through a grinding stage.
HPGR feed

Pre-crushing
critical size

Coarser fraction Finer fraction

Epre-crushing Pre-crushing

Combiner

Egrinding = Etotal - Epre-crushing Grinding

HPGR product

Figure 12 Schematic of HPGR breakage model used in simulation

Figure 13 shows predicted P50 and P80 particle sizes from the simulation. In this case the fines,
which avoid pre-crushing, were defined as -1 mm. Based on the degree of fit, the model provides an
excellent prediction given that only 10 kg is required to conduct 15 piston press tests on mono-size
particles and define the energy–breakage relationship. In summary, the total specific energy input
during a piston press test is used with equation 5 to determine the t 10 for each feed size fraction.
Following this, the set of master t-family curves, presented in Figure 12,is used to reconstitute the
simulated product PSD.

12
 Figure 13 Simulation of 36 different tests

CONCLUSIONS
Three piston press test methodologies were developed:
1. The direct calibration methodology uses around 1000 kg of composite sample and less
than 10 kg of sample for defining each geo-metallurgical unit. Thus, the overall sample
requirements are considerably reduced and the methodology can be used from scoping to
feasibility level studies.
2. The database-calibrated methodology requires only 10 kg sample and it provides energy
prediction within ±25%, which is acceptable at scoping level studies.
3. The simulation methodology provides a means of simulating closed-circuit operation of
HPGR and it is applicable to all stages of metallurgical studies and even operations.
At the time of writing, only a few pilot HPGR units are setup globally at metallurgical facilities for
commercial assessment of HPGR amenability for a particular ore type. The developed piston press
methods can be carried out using hydraulic piston press equipment which is often used in rock
mechanics and materials testing laboratories. Therefore, in addition to the associated test benefits of
reduced sample quantities and turn-around time, a larger number of testing facilities will be able to
offer the test which significantly improves accessibility and opens the test up to industry regulated
standardization.

REFERENCES
Abouzeid, A. M., & Fuerstenau, D. W. (2009). Grinding of mineral mixtures in high-pressure
grinding rolls. International Journal of Mineral Processing, 93(1), 59–65.
doi:10.1016/j.minpro.2009.05.008
Beaty, R., Lipsey, R., & Elgie, S. (2014, July 9). The shocking truth about B.C.’s carbon tax: It works.
The Globe and Mail. Vancouver, BC, Canada.

13
Claffin, J., Leonard, R., Rogers, D., & Urritia, F. (2009). Concentrated solar powered mineral
processing. In D. Malhotra (Ed.), Recent advances in mineral processing plant design. Littleton, Colo:
Society for Mining, Metallurgy, & Exploration.
Lim, I. L., Voigt, W., & Weller, K. R. (1996). Product size distribution and energy expenditure in
grinding minerals and ores in high pressure rolls. International Journal of Mineral Processing, 44-45,
539–559. doi:10.1016/0301-7516(95)00064-X
Morrison, R. D., & Morrell, S. (1997). Evolution of AG/SAG Milling Technology - Current Status
and Future Directions. In S. K. Kawatra (Ed.), Comminution Practices.
Shi, F., & Kojovic, T. (2007). Validation of a model for impact breakage incorporating particle size
effect. International Journal of Mineral Processing, 82(3), 156–163. doi:10.1016/j.minpro.2006.09.006
Tromans, D. (2008). Mineral comminution: Energy efficiency considerations. Minerals Engineering,
21(8), 613–620. doi:10.1016/j.mineng.2007.12.003
U.S. Department of Energy. (2007). Mining Industry Energy Bandwidth Study. Office of Energy
Efficiency and Renewable Energy.

14
Jameson Cell Project Evaluation in the Cleaner
Circuit at Codelco Andina
Eduardo Morin1 and Virginia Lawson2
1. Andina Division, Codelco Chile
2. Glencore Technology, Australia

ABSTRACT
The Process Engineering Superintendent, visualizing the opportunity for improvement in the
recovery of Molybdenum in the stage of column flotation carried out during 2015 an experimental
campaign on a pilot scale, using flotation technology Jameson Cell in the concentrator plant Division
Andina.
It has been experimentally found that the columns tend to deconcentrate molybdenite, so Mo
recoveries at this stage are <25%. This phenomenon occurs by high Mo recoveries in step Scavenger
(> 90%). However, this stage is not able to recover more Mo, so is lost in the final tailings plant.
The purpose of these tests was to confirm the advantages of the technology on Jameson technology
currently used in the final cleaning stage flotation circuit.
The results show that the Jameson cell is capable of producing concentrates quality similar to that
generated in columns cells in terms of copper (27-29%) and 60-90% recoveries, but a clear difference
was obtained between the recovery of molybdenum obtained in columns (10-25%) compared with
the recoveries obtained in the Jameson (40-80%) cell. Regarding the law of molybdenum concentrate
obtained cell column values 0.6-1.0% compared to the Jameson cell with laws 1.0-3.0%
The main difference between the two technologies is the bubble size and intensity of mixing of the
particles, allowing smaller particles may adhere to bubbles and be collected as concentrate, which
does not occur in cells columns, where mixing intensity is lower and larger diameter bubbles are
difficult collection molybdenum particles in the bubbles.
In the quest to be regarding national level in the production and recovery of molybdenum Division
Andina economically evaluate the option of including the Jameson cell in the process.

1
INTRODUCTION
In the bulk copper molybdenum flotation plant at Codelco Andina Division it has been
experimentally found that the columns tend to "deconcentrate" molybdenite, so Mo recoveries at this
stage are <25%. This phenomenon is mitigated by some degree by high Mo recoveries in the
Scavenger (> 90%) however, this stage is not able to recover more Mo, so is lost in the final tailings
plant.

Figure 1 Plant flotation circuit layout

The flotation circuit at Andina Division is comprised of three lines of rougher flotation (lines A, B
and E). The rougher concentrate is sent to a regrinding step in vertical mills.
Subsequently reground concentrate is sent to 6 columns cells (1 stand by) producing the final
concentrate. Tailings from the column cells are sent to a scavenger stage. The scavenger concentrate
is sent to regrinding in vertical mills (in combination with the rougher concentrate) while the tailing
of scavenger stage is sent to final tailings thickeners, together with the tailings from the roughers (
lines A, B and E).
The superintendent of process engineering, visualizing the opportunity for improvement in the
recovery of Molybdenum in the final cleaner stage, contacted Glencore Technology (formerly Xstrata
Technology), to analyse whether its flotation technology, the Jameson Cell, could be a solution to
improve the recovery of molybdenum in the cleaning step as an alternative to the columns cells.
During 2013, Andina Division of Codelco, performed laboratory dilution tests (used to represent the
operation of the Jameson cell using conventional cells at laboratory scale), to determine whether the

2
Jameson Cell flotation technology may obtain improvements in performance compared to the
columns cells.
The results of these tests showed that the Jameson Cell could consistently recover both copper and
molybdenum, up to 85%, which in the case of molybdenum, was a significant improvement over the
base case where the plant columns were only recovering 15% of molybdenum. These results led
Andina Division to consider conducting a pilot plant on site using the Jameson Cell during 2015, in
order to validate the laboratory dilution test results with a plant based pilot.
Currently the concentrator achieves the following overall plant performance:

Table 1 Overall plant performance

Cu (%) Mo (%)
Feed 0,8 0,032
Final Concentrate 28,5 0,6-0,9
Recovery 88,5 71,5

As for column flotation cell, the stage recovery range corresponds to:

Table 2 Range of recovery of column cells

Cu (%) Mo (%)
Recovery 50-80 10-25

METHODOLOGY
In order to carry out the test plan industrially, the Jameson Cell pilot model L500 / 1 was installed, as
shown in Figure 2. The L500 / 1 unit was mounted on the floor of the flotation section D. The tests
were performed using a bleed of the slurry feed flow to the column cells taken from the pump box
feeding the Andina Division columns via a bredel pump. Segregation at the feed sample point
resulted in a coarser feed to the Jameson cell. The location was later moved to enable a better sized
sample however the feed grade was lower.
The pilot tests were conducted at the Andina Division concentrator plant in two stages: the first
between May and July 2015 where a range of different operating conditions of the Jameson cell were
tested in order to determine the grade vs recovery curve (Cu and Mo) for different operating
conditions.
The second stage was conducted between August-November 2015, in order to validate the best
operating conditions obtained during the first phase of testing.

3
 Figure 2 Jameson Cell concentrator plant installation

During the pilot with the Jameson Cell the following operational variables of the cell were varied:
froth depth, wash water flow and air flow, so as to generate different recovery conditions vs grade,
and get as many conditions operational possible for the cell. This intentional variation of parameters
produces results over a range of conditions to enable analysis of cause and effect. The tests were
designed to operate along the grade recovery curve over different days and conditions.
To improve selectivity the following operating conditions was used:
• Increased froth depth
• With water wash (on)
• Low airflow.
However, when operating conditions aimed at a greater recovery was explored the following were
used:
• Low froth depth
• No washwater (off)
• High airflow
This demonstrates the great versatility of Jameson cell technology, which allows it to operate in a
wide range of operating points, depending on the requirements that exist from the process.

Table 1 Jameson Cell range of operating parameters

Parameter Range
Air flow (m3/h) 9-16
Vacuum pressure (kPa) 7-16
Cell level (%) 60-90
Froth Depth (mm) 150-650
Washwater (m3/h) 0-1.7

4
CURRENT PLANT CIRCUITS
The main results of this experimental campaign are presented, calculated based on the fresh feed for
the Jameson Cell:

Testing on Column Feed stream

Upon completion of plant testing the results confirm the benefit of the application of the Jameson cell
observed at laboratory scale (dilution test). These results show that the Jameson cell is capable of
producing concentrates similar to that generated in the plant columns cells in terms of copper (grade
and recovery), but show significant improvement in recoveries and grades of molybdenum
compared to what is currently achieved in columns cells. Figure 3 shows the copper grade and
recovery obtained for both the Jameson Cell and the plant column cells.

Figure 3 Cu Grade Recovery curve

5
 Figure 4 Molybdenum Grade Recovery curve

With respect to molybdenum a clear difference between the results obtained for molybdenum
recovery in the columns (10 to 25 %, with a single value around 37.8 %), compared with the recoveries
obtained in the Jameson Cell with up to 90% recoveries observed. The main difference between the
two technologies is the bubble size and the intensity of mixing for contact of the particles and bubbles.
In the column cells, where mixing intensity is much lower and with bubbles of significantly larger
diameter molybdenum particles are difficult to recover. Figure 4 shows the recovery comparison for
molybdenum (Jameson Cell and plant column cells). Both the grade and recovery are significantly
improved.
Figure 5 shows the selectivity curve of copper with molybdenum, for the plant columns cells and for
the Jameson Cell. This graph shows a significant difference in the behaviour of both technologies and
reinforces that the molybdenum can be floated by true flotation and not only by entrainment, as is
often accepted by operations and reported in the literature (Gonzalez, 2015).
The column cell demonstrates more selective flotation behaviour of copper compared to
molybdenum. Copper recoveries ranging from 50 to 80%, while for molybdenum ranging from less
than 10 % to 25 %. Even at the maximum value of 37.8 % molybdenum recovery (80 % copper) the
performance is two thirds that achieved in the Jameson Cell of 60 % molybdenum.

6
 Figure 5 Copper Molybdenum Selectivity curve

Testing on column tails

Exploratory tests were also conducted on the plant column tail to identify if this may be an ideal
location for the Jameson Cell. The results show the same behavior on the column tail as was observed
on the column feed i.e. copper recoveries of 60% and recoveries of 50% for molybdenum.
Molybdenum concentrate grades were excellent. This confirms the floatability of the molybdenum
that is lost in the column.
The copper grades on this stream are lower, below 20% copper. This indicated that the copper in the
column tails is not liberated and that the Jameson cell is successfully floating composite materials,
probably from the coarser fractions with lower grade. This indicates that liberation is limiting the
concentrate grade as the Jameson Cell recovers more coarse material.
Figure 6 and Figure 7 illustrate the impact of a single variable, wash water, on the grade recovery
curve of copper and molybdenum. In the case of copper, some recovery is lost as particles with lower
hydrophobicity are removed and the concentrate grade is improved at a lower recovery. This
represents movement on the same grade recovery curve again confirming the presence of composite
particles. The molybdenum however shows that its grade is increased with very little change in
recovery demonstrating its recovery by true flotation and suggests that the froth stability in a
Jameson Cell with wash water is supporting its recovery.

7
 Figure 6 Effect of wash water on copper performance - column tails

Figure 7 Effect of wash water on molybdenum performance - column tails

8
CONCLUSIONS
Pilot plant testing confirmed the results, and the benefit, of the application of the Jameson cell
observed at a laboratory scale (dilution test). These results show that the Jameson cell is capable of
producing concentrate grades similar to that generated in columns cells in terms of copper (grade
and recovery), but show significant improvement in recoveries and grades of molybdenum in
comparison to what is currently achieved in the plant columns. A summary of the results can be
observed the following table:

Table 2 Comparison of Column Cells and Jameson Cells

Column Cells Jameson Cell

Recovery Cu % 50-80 60-90
Recovery Mo % 10-25 40-80
Maximum Grade Cu (%) 29.7 28
Maximum Grade Mo (%) 0.9 3.0

Due to the intense mixing and novel bubble generation in the Jameson Cell, it is possible to
simultaneously obtain good recoveries of both copper and molybdenum, compared to the columns
where copper is recovered in preference to the recovery of molybdenum. This shows that with the
Jameson Cell molybdenum is floated by true flotation and not as entrainment, as is often accepted by
operations and literature (Gonzalez, 2015).
These results confirm that the Jameson cell is a valid alternative to consider for replacing the column
flotation cells currently installed in the circuit, in order to improve the metallurgical performance of
the flotation cleaning circuit at Andina Division.

REFERENCES
Gonzalez, C. / Glencore Technology (2015). Pruebas Piloto Celda Jameson L500/1 – División Andina – Codelco
Chile.

9
Characterization of Laboratory Crushing Tests with
Chuquicamata Underground Mine Samples
Bruno Merello1, Álvaro Videla1, Edisson Pizarro2 and Jaime Sepúlveda3
1. Pontificia Universidad Católica de Chile
2. Chuquicamata Division, Codelco, Chile
3. J-Consultores, Chile

ABSTRACT
This publication examines the results obtained from an extensive laboratory jaw-crushing test
program conducted with seven different ore types, which were obtained from the Chuquicamata
Underground Mine project. In total, one hundred trials to evaluate two different Close Side Setting
(CSS) conditions were performed with 9kg samples. Measurements of specific energy consumption
(kWh/ton), feed (F80) and product (P80) sizes, crusher power (kW) and residence time (s) in the
crushing chamber were recorded for each test to collect sufficient data to apply different predictive
models for evaluation. As expected, results show that product size distribution (P80) and maximum
discharge size are mainly defined by the CSS. Specific energy consumption also depends mostly on
CSS configuration and ore type, while feed distribution impacts only slightly on the results. In this
particular case, it was observed that both alternative empirical models, proposed by Bond and Kick,
correlated well with the experimental data thus generated, with a determination coefficient larger
than 88%. Also, it may be observed that ores with similar geological and geotechnical characteristics
(UCS, Y, E, PLI, among others) behave significantly different with respect to specific energy
consumption, and therefore in the crushing capacity of a given installed power. The crushing
behavior of mixed ore types has also been evaluated, showing a non-linear relationship.

1
INTRODUCTION
Chuquicamata Underground project represents one of the severest technical challenges for the
Chilean mining corporation, CODELCO, in this decade. Underground extraction involves block
caving mining, which will be processed with inside crushing equipment to assure a maximum
particle size of 8 inches. These will then be transported on a conveyor belt 7 km up to surface. The
process is essential, but the crushing machines’ size has not yet been decided.
Previous geological and geotechnical studies have revealed several mineralization and geotechnical
units with different mechanical behavior. According to the current long-term plan for the mine, six
ore types have been picked for mining during the first 4 years of operation. Figure 1 displays ore
type distribution for every two years for the first six years.

Figure 1 Ore type distribution per year of operation

The current research has developed a laboratory test to characterize the crushability behavior of
these ores to identify potentially significant differences that could affect the crushing performance
during the first years of operation. We have installed an online power consumption control on a

2
laboratory jaw crusher to run 115 trials in total to evaluate the effect of ore type, close side setting,
feed size, and ore mixing.

METHODOLOGY
We have run 103 crushing trials in total for seven different rock types as shown in Table 1. In most
cases, duplicates were run to evaluate the variability of the results. We have identified the ore
material we received from the sampling campaign as normal particle size. Coarse and fine particles
were selected manually and sent to SGS for size characterization in a vibrating screen before testing.

Table 1 Number of Experiments per Trial

Equipment Description
A Laboratory Jaw Crusher 4”x6.4’’x10.5” with a 4 KW WEG-W42 TE1BFX0X power engine has been
used for the experiments. The machine was connected to an AC Power Monitor PQube 02-100,
which allowed detailed recording of information into an Excel spreadsheet. Registered variables
included Current, Voltage and Power per unit of time (seconds). The data was then downloaded
via a secure WIFI connection as displayed in Figure 2.

3
 Figure 2 Left: Jaw Crusher. Right: Schematical Diagram of the Monitoring Setup

Preparation of Sampling Ore

Ore Samples were collected from the extracted material from development and exploration tunnels.
2.5 tons in total were received at the university facilities in Santiago. This material was visually
identified and categorized as 11% PF, 11% PI, 23 PIR, 26% Qs, 10% VMO, 11% ZT and, additionally,
8% GFO. The latter was not part of the original mine plan for the first four years, though, but
included in case it becomes unexpectedly important.
Manual selection of rock samples, coning and quartering was applied to prepare ore material for
one crushing test. Material was conned and quartered until samples of around 10 kg each emerged.

Test Protocol
Each sample was weighted and located on the feed crusher. The Jaw crusher was started under
shook-feed conditions and run until empty. Data was collected during the entire process, and the
net energy consumed during the test was divided by mass, in order to calculate the Specific Energy
Consumption (CEE) for the condition under analysis.
The feed particle size distribution was obtained in an ATM Arrow 12 inches vibration screen, and
the product particle size distribution was obtained with a WS Tyler RX 29 10 rot-up desktop
machine.

RESULTS AND DISCUSSION

Figure 3 shows the product particle size distribution for each test at different Close Side Settings
per rock type. As expected, the CSS significantly impacts the final product definition.

4
 Figure 3 Product Particle Size Distribution

Bond’s Law Applicability

General Charles’ Law of particle size reduction establishes a relationship between product, feed
and SEC as in Equation 1. A regression analysis applied to the set of data reveals that Bond’s Law
fits the set best, compared to Kick’s and Rittinger’s Laws. However, the differences between all
regression coefficients remain small, reaching respectively 83.3%, 89.7% and 89.5%. Figure 4 shows
the Bond’s Law fit line as applied to the data set .

æ 1 1 ö
SEC = K çç n-1 - n-1 ÷÷ (1)
è dp dF ø

æ 1 1 ö
SEC =10Wi ç - ÷ (2)
è P80 F80 ø

5
 Figure 4 Relationship between the Real SEC and the expected SEC as obtained
from application of Bond’s Law with the best-fit Work Index

The Bond’s Crusher Work Index (Wi) has been calculated by application of Equation 2 to each set of
data collected in each trial. Table 2 shows respectively the average Wi per ore for the set of
experiments and the coefficient of variation.

Table 2 Average Crusher Work index and coefficient of variation per sample

6
The table reveals that PIR and QS ore, the most common types that will be mined during the first
years, will have an expected Crusher Wi between 14 and 17 kWh/ton, respectively. PF ore type
denotes the toughest rock sample family tested, although it represents only a minor percentage of
the ore extracted during the first years of operation.
We have run additional tests to measure the Crusher Bond’s Work Index as obtained from an ore
mixing, according to the mine plan in Figure 1. Figure 5 shows a comparison between a mass
weighted average and the real Wi as obtained from the laboratory test. The mass weight average
has been obtained from weighting Wi values from Table 2 and mass percentage according to the
mine plan in Figure 1. 5% or less ore material in the mix has not been taken into account for the
averages or laboratory tests.

Figure 5 Comparison between a weighted mass estimation of the mixed-ore Wi

according to mine plan v/s the actual value measured in the experimental test for the same
mixed-ore

The results reveal a low variation between real and estimation values. Differences between both
values are below 10%; specifically 6%, -3% and 10% for 2019, 2021 and 2023 respectively, which is
acceptable for conceptual studies.

CONCLUSION
The current example shows an example of a characterization method as used to evaluate the
crushability of different ore types in the Chuquicamata Underground Project. The evaluation
proceeded carefully by monitoring energy consumption online and at low CSS. The data sets
obtained are showing a high variability between the seven ore types under analysis. Despite the
high variability, experiments performed with mixed ore are showing that a mass weight average
between the evaluated ore types is a good approximation to estimate the real laboratory Crushing
Work Index Value of the mix.

7
ACKNOWLEDGEMENTS
We acknowledge the kind help and support of CODELCO Chuquicamata and appreciate their
permission to publish this information.

NOMENCLATURE
SEC Specific Energy Consumption
K Proportionality Constant
dp Characteristic Product Size

dF Characteristic Feed Size

Wi Work Index
P80 Particle Size with total product mass below 80%
F80 Particle Size with total feed mass below 80%

REFERENCES
Gutiérrez, L. and J. E. Sepúlveda (1986). Dimensionamiento y Optimización de Plantas Concentradoras
Mediante Técnicas de Modelación Matemática, CIMM-Chile Publication.
Araya, K and Videla A. (2015).’ Practical Evaluation of an HPGR as Pebble Crusher to Increase
Energy Efficiency in the Overall Grinding Circuit and Promoting Technological Substitution‘, 11th
International Mineral Processing Congress.
Achiardi, C and Videla A. (2014). ‘Technical-economical evaluation of product size grinding on copper
suphide flotation circuit”, XXVII International Mineral Processing Congress, pp 752.

8
Bubble Surface Area Flux Measurement in
Laboratory Flotation Cells: Two- And Three-Phase
Operation
Cesar Gómez and Jaime Álvarez
Universidad de Concepción, Chile

ABSTRACT
The most used particle separation technology to produce concentrates of a mineral of interest is
flotation. The process is based on the capture of particles containing the valuable mineral on the
surface of bubbles, and therefore, metallurgical performance is expected to be strongly related to
the characteristics and flow of the interacting particle and bubble populations. This interaction is
characterized in part through variables related to the rising flow of bubbles along the cell
(superficial gas velocity and gas holdup). The characteristics of the bubble population generated by
gas dispersion, in the case of lab mechanical cells, depends on the selected impeller type, rotation
speed, cell size, and operating variables (gas flow rate, frother type and concentration). Bubble
surface area flux (Sb), which integrates bubble size and superficial gas velocity effects, is considered
the most promising variable to relate gas dispersion and metallurgical performance. Calculation of
Sb requires values of superficial gas velocity and an average bubble size. In the case of lab flotation
testing, the superficial gas velocity is known and calculation of Sb requires only measurement of the
bubble size distribution, which can be reliably accomplished using a sampling-for-imaging
technique. Two-phase gas dispersion characterization in lab flotation machines is relatively simple
and inexpensive; measurements in a variety of cell types and sizes have shown a wide range of
bubble sizes obtained for different combinations of the operating variables. A review of two- and
three-phase bubble size measurement exercises performed in the same laboratory machine under
similar conditions showed small differences. Knowing that for a set of conditions the same bubble
size distributions are generated in the presence of frother solutions or mineral pulps, a two-phase
gas dispersion characterization of a flotation machine makes possible the educated selection of test
program conditions for ore characterization and performance optimization efforts.

1
INTRODUCTION
Mineral flotation is a process for selectively separating particles by attachment to the surface of
bubbles. Once the particle size distribution and liberation characteristics are defined in the grinding
stage, the efficiency of the flotation process is a function of size distribution of the generated bubble
population, which depends on the installed gas dispersion technology and its operation. Gas
dispersion is characterized through gas holdup, superficial gas velocity and bubble size
distribution, but bubble surface area flux (Finch et al., 2000), which integrates bubble size and
superficial gas velocity effects, is considered a more inclusive variable to relate gas dispersion to
metallurgical performance.
Bubble surface area flux (Sb), defined as the bubble interfacial area generated per unit time and cell
cross sectional area, continuously rises through the cell determining bubble-particle collision
frequency; as a result a correlation with metallurgical performance is expected. Have been
demonstrated relationships between Sb and recovery in pilot units (Gorain, Franzidis & Manlapig,
1997; Gorain et al., 1998; Hernandez, Gomez & Finch, 2002) and laboratory cells (Gomez, Castillo &
Alvarez, 2014; Gomez, Mesías & Alvarez, 2016). Bubble surface area flux is calculated from two
measurable gas dispersion variables: superficial gas velocity and average bubble diameter
(estimated from the bubble size distribution). In the case of lab flotation testing, the superficial gas
velocity is known (delivered gas flow rate and cell cross sectional area are defined), and Sb
calculation requires only measurement of bubble size. Have been proposed several techniques to
measure bubble size in industrial flotation cells (Jameson and Allum, 1984; Randall et al., 1989;
Gorain, Franzidis & Manlapig, 1997; Yianatos et al., 2001; Grau & Heiskanen, 2002; Gomez & Finch,
2007); reproducible measurements in laboratory flotation units have been demonstrated (Girgin et
al., 2006; Gomez et al., 2013; Alvarez, Gomez & Wen, 2014; Gomez, Castillo & Alvarez, 2014).
Are certainly affected tests run in laboratory mechanical cells by changes in Sb. Selection of impeller
type (Agitair vs. Denver) and operation (rotating speed), cell volume, and operating variables (gas
flow rate, frother type and concentration) will have an effect on bubble size and Sb (Gomez et al.,
2013). Variations because of poor or no control of some of these variables during the collection of
concentrates (i.e., gas flow rate, frother concentration), will probably go unnoticed and ensuing
changes in performance will be attributed to variations of other process variables. It has been
demonstrated in two-phase tests that the presence of surface-active agents has a significant impact
on bubble size in lab flotation columns (Azgomi, Gomez & Finch, 2007), and in lab and pilot
mechanical cells (Gomez, Castillo & Alvarez, 2014; Nesset, Gomez & Finch, 2007). Characterization
of gas dispersion in lab flotation machines is relatively simple and inexpensive when it is done
using frother solutions. One concern, however, is whether bubble sizes obtained in two-phase
measurements can be used for the selection of conditions in tests with mineral particles included.
Tests to measure the critical coalescence concentration (CCC) in two and three phases, as proposed
by Cho & Laskowski (2002), showed for several frothers minor bubble size differences when
performed in frother solutions and mineral pulps. This communication reviews experiences
involving two- and three-phase bubble size measurement exercises performed in the same
laboratory machine under similar conditions.

BUBBLE SURFACE AREA FLUX: IMPACT IN FLOTATION AND MEASUREMENT

The selective separation of particles of valuable minerals in flotation machines is accomplished by
creating and maintaining a collection and a froth zone. For a mineral species, flotation performance

2
is modeled using a collection zone recovery (Rc), defined as the fraction of mineral mass entering
the zone that leaves on the surface of bubbles, and a froth zone recovery (Rf), defined as the fraction
of mass entering the froth on the surface of bubbles that reaches the concentrate. An interaction
between zones, in the form of a drop-back stream formed by drained pulp and captured particles
that detach or are released by bubble bursting, is also considered. The overall recovery achieved in
the machine (Roverall), defined as the fraction of mass that reaches the concentrate relative to that in
the incoming fresh feed to the unit, is obtained from a total mass balance:
(1) R f Rc
R overall =
R f R c +(1-R c )

This equation indicates that the maximum possible overall recovery is that obtained in the
collection zone when all particles collected on the surface of bubbles reach the concentrate (Rf=1).
Therefore, operation of laboratory or industrial size cells, focuses on creating conditions that
maximize the formation of bubble particle aggregates in the collection zone, and minimizes bubble
bursting in the froth zone.
Calculation of the collection zone recovery assumes, for a given mineral, that the mass in particles
captured on the surface of bubbles follows a first order process:
(2)
dM
- = kM
dt

where M is mass of the selected mineral in the particles suspended in the collection zone after a
flotation time t, and k is a proportionality constant known as flotation rate constant. This flotation
rate constant has been frequently related to bubble size, but bubble surface area flux (Sb) is currently
considered the logical variable to relate gas dispersion to particle collection, as integrates bubble
size and superficial gas velocity effects (Finch et al., 2000). It is also recognized that the formation of
bubble-particle aggregates depends on the hydrophobic character of the particles containing the
valuable material, a factor incorporated through a constant (floatability P), which leads to:
(3)
dM
- = PS b M
dt

Integration of Equation 3 to calculate the collection zone recovery requires to take into account the
conditions existing in the flotation machine (a lab mechanical cell in this case), as they define Sb
variations in time and space, and the characteristics of the particle population, that dictate
floatability values. Gas dispersion into small bubbles produces every second interfacial areas that
are between one and two orders of magnitude as large as the cross sectional area of the cell. The
consequence of this large interfacial area, which rises continuously through the cell, is that the
fraction of bubble surface covered with particles is negligible and Sb can be considered constant and
equal to that generated through gas dispersion. On the other hand, the value of floatability is
constant only for liberated particles of the same size, but for multi-mineral particles, depends on the
fraction of external area covered by the mineral of interest. In this case, a floatability distribution is
expected, as particle populations obtained from grinding ore samples have their own size and
liberation distributions.
Measuring performance in lab flotation tests, once the machine factors and chemistry have been
selected, leaves gas dispersion and flotation time as the only factors determining recovery.
Recovery calculation by integration of Equation 3 requires Sb values, which can be obtained from

3
values of superficial gas velocity (Jg=Qg/Acell) and bubble size, following its definition (bubble surface
area generated per unit time and cross sectional area of the cell). For a bubble of diameter d, its
volume and surface area are πd3/6 and πd2, respectively, and for same size bubbles, the total
number of bubbles (Nb) generated per unit time by injection of a gas flow rate Qg is:
(4) Qg 6 Qg
Nb= 3
=
πd /6 πd 3

The surface area generated by unit time and cell cross sectional area (Sb) is then:
(5) N b (πd 2 ) 6 Q g πd 2 6 J g
Sb = = =
A cell πd 3 A cell d

In flotation machines, gas dispersion results in bubble populations with a distribution of sizes. The
bubble size is measured, in our case, on a sample of bubbles collected from the cell using a
sampling-for-imaging technique (Gomez and Finch, 2007). The total external area and volume of
the sampled bubbles, Asample and Vsample, respectively, are calculated from the individual bubble
diameters (di) measured on sampled bubbles collected close to the top of the collection zone:

(6) A sample =  πd i2  
(7) 
V sample =  πd i3 /6 
The Sb is calculated from the total external area of the sampled bubbles per unit cross sectional area
of the cell times the ratio between the gas flow rate and the total volume of the sampled bubbles, as
indicated in the following equation:

Sb =
 2
 πd i Qg
=
Qg 6  d i2
 
(8) 3
A cell 3
 πd i /6 A cell  di

This equation is valid for gas flow rates and bubble diameters both corrected to the conditions
existing at the bubble collection point. As the Sauter mean diameter D32 is equal to the ratio Ʃdi3/
Ʃdi2, then Sb is calculated as:
(9) 6 Jg
Sb =
D 32

This equation indicates that high gas flow rates and conditions leading to smaller bubble sizes, will
result in higher Sb values, and increases in collection zone recovery should be expected. If the
increase in Sb can be achieved without reducing the froth zone recovery, then the recovery of the
machine will increase. Values of Sb measured in industrial units close to the top of the collection
zone, are in the range 10 to 90 s-1, with most cases between 20 and 60 s-1 (a value of 50 s-1, for
example, means that the external surface area of the bubbles generated every second is 50 times the
cross sectional area of the cell).

4
EXPERIMENTAL ASPECTS
Measurements of performance and bubble surface area flux in tests run in laboratory mechanical
cells, at selected operating conditions and in the presence of frothers, require protocols for ore
sample preparation, grinding and pulping, frother addition, bubble size measurement, concentrate
sampling and analysis, and selection of a flotation kinetics model.

Laboratory mechanical cell

The laboratory cell utilized in our work has been used and described before (Gomez et al., 2013;
Alvarez, Gomez & Wen, 2014; Gomez, Castillo & Alvarez, 2014). The unit, an automated forced-air
mechanical flotation machine (Edemet), allows programming, continuous tracking, and control of
operating conditions and electrochemical variables. Features used in our work were impeller speed
control, the internal compressor providing clean (filtered) oil-free air, and the automated paddles
for concentrate removal in three-phase flotation tests. Air flow is manually controlled and
monitored using one of two rotameters (ranges 0-10 L/min and 0-20 L/min). Two types of impellers
are available: Agitair for cell volumes of 1.5, 2.7, and 5.0 L; and Denver, for a cell volume of 4.5 L.

Preparation of mineral samples and flotation pulps

The ore samples, collected after crushing, were provided by a local concentrator. The ore samples
were crushed in three stages using a jaw, a cone and a roll crusher; the product from the roll
crusher was in close circuit with a 10-mesh (2 mm) screen. This product was separated in one-kg
flotation samples; two of these samples were used for chemical analysis (total Cu, soluble Cu, S,
and insoluble material), and five were used to determine ore density by water displacement. The
one-kg samples were ground in a 5.4-L ball mill, Marcy type, using a balanced charge of steel balls.
Collector, diesel, calcium oxide and a volume of frother solution are added to have a 70% pulp
density. The grinding time to have a product with the desired 80% passing size was determined.
The grinding product was charged to the flotation cell with the amount of frother solution
necessary to have a 30% pulp density. The pH is adjusted, if necessary, with calcium to a value of
10.

Preparation of frother solutions

For every frother, preparation of test solutions at several concentrations were necessary to model
the bubble size variation as a function of frother concentration. The preparation procedure
considers that frothers, in many cases, have limited solubility and are slow to dissolve. At the same
time, frothers may be volatile, and may decompose at room temperature if exposed to daylight. A
large volume (about 12 L) of frother solution is necessary to fill both the cell and the bubble size
measuring device. The test solutions are prepared by dilution of a 100-ppm (µg/g) stock solution,
made by dissolving 2.3 g of frother in 23 kg of water. The mixture was maintained with a gentle
agitation to ensure complete frother dissolution, for 4 and 12 hr in the case of analytical-grade and
commercial frothers, respectively, and then sealed and stored in darkness in the same container. An
effort was made to run the tests for one frother on the next day the stock solution was prepared, to
minimize concentration changes by volatilization and decomposition, which would go unaware.

5
Bubble size measurement
Bubble size measurement was accomplished using the sampling-for-imaging technique developed
at McGill (Gomez & Finch, 2007), which is based on the processing of bubble images collected using
backward illumination, in a chamber designed to expose the collected bubbles on a single focus
plane with little or no overlapping. This chamber, which is permanently installed above the
flotation cell, connects to the cell through a vertical sampling tube closed at its bottom with a rubber
stopper. A steady flow of rising bubbles is established after a few minutes following the stopper
removal, and the collection of 200 to 300 images at a rate of one every second (for avoiding imaging
of the same bubbles twice), is initiated. A digital camera (Nikon model D5100) fitted with a macro
lens (Nikon 60-mm) produces high magnification (4928x3264 pixels) and resolutions (between 170
and 180 pixels/mm). Images are processed using the program Image J, which find area and pixel
value statistics for every object (group of neighboring pixels darker than a threshold) in the image;
area is used to calculate bubble size as the diameter of the circle of the same area, and circularity is
used as criterion to accept objects as bubbles (larger than 0.8). Bubble clusters, if present, show
circularities lower than 0.8 and are eliminated. Sizes of the individual bubbles are used to calculate
the average Sauter mean diameter D32 (same volume and interfacial area) for the distribution.

Procedure for measuring bubble size and metallurgical performance

Practical considerations make necessary to perform bubble size and metallurgical performance
measurements independently. The bubble size measurement is initiated by filling the bubble
imaging device with the selected frother solution and the cell with the frother solution (two-phase)
or the test flotation pulp (three-phase). A low conditioning impeller speed (600 rpm) is set up for a
few minutes to ensure proper mixing, and then the selected gas flow rate and impeller rotation are
established. After a few minutes to stabilize bubble formation, bubble sampling is initiated by
removing the stopper closing the bubble collection tube, followed by collection of bubble images as
a steady flow of rising bubbles rapidly develops. Measurement of metallurgical performance
involves the collection of consecutive samples of the aerated pulp overflowing the cell. Sampling
frequency is specific to the ore tested based on its flotation response. After conditioning and bubble
formation stabilization, the rotating paddles are activated to remove and collect samples of the top
of the froth. The fastest froth paddling rate (about 1 rotation/second) is normally used, with a
frequency that depends on the flotation response (available range is once every 1 to 10 seconds).
Concentrate samples (and the residual material left in the cell) are weighed, filtered, dried and
processed for obtaining representative fractions for chemical analysis.

RESULTS AND DISCUSSION

Determining flotation performance (Equation 3), involves measurement of the size distribution of
the bubble population rising through the cell, for calculating Sb. Work designed to characterize gas
dispersion in two-phase operation of forced-air lab flotation cells, run at numerous combinations of
impeller type, cell size, rotation speed, gas flow rate, and frother type and concentration, involved
bubble size measurement (Alvarez, Gomez & Wen, 2014). The results demonstrated a significant
effect of the cell and conditions chosen on bubble size and Sb. Bubble size measurements obtained
in a 2.7-L Agitair unit as a function of frother concentration for constant gas flow rates and constant
impeller speeds, which are of interest as this unit has been also used to measure bubble sizes in
three-phase tests, are included in Figures 1a and 1b, respectively. The results showed the expected

6
rapid decrease at frother concentrations below the CCC, and a negligible effect above the CCC, with
larger values as the higher gas flow rates were used, and a minor no consistent effect of impeller
speed (it is believed that differences are a consequence of poor gas flow rate setting).
Corresponding Sb's (Figures 2a and 2b) showed rapidly increasing values below the CCC, with no
much effect of gas flow rate as increases in bubble size compensated increases in gas flow rate, and
above the CCC, larger Sb's values as increases in gas flow rate had no effect on bubble size. There
was no consistent effect of impeller speed on Sb.
Two- and three-phase bubble size measurements were performed to establish the effect of solids on
the determination of frother critical coalescence concentrations (Grandon, Alvarez & Gomez, 2014).
The tests were run following the procedures described in the previous section, using the 2.7-L
Agitair cell operated at 1,000 rpm, using 10 L/min (J g=0.72 cm/s) of air, and the fastest paddling rate

Figure 1 Two -phase bubble size measurements as a function of frother concentration at several:

a) gas flow rates at 1,000 rpm; and b), impeller speeds at 10 L/min

7
 Figure 2 Two -phase bubble surface area fluxes obtained as a function of frother concentration at several: a)
gas flow rates at 1,000 rpm; and b), impeller speeds at 10 L/min

Of once every 5 seconds. Analytical grade samples of 1-heptanol and MIBC (Sigma-Aldrich), and
commercial samples of MIBC (Celanese) and Flomin 500 (SNF), were used in the tests. One-kg
samples of a copper ore with addition of collector (22 g/ton), diesel (8 g/ton), 0.35 g of calcium oxide
and water to have a pulp density of 70%, were ground in a Marcy ball mill. A grinding time of 18
minutes resulted in a product with 80% passing 150 µm. The grinding product was diluted to have
a 30% pulp density and the pH adjusted to a value of 10.
Although no differences in CCC were detected in measurements in two and three phases,
consistent smaller sizes were obtained when solids were present (Figure 3), a result previously
obtained by O'Connor, Randall & Goodall (1990). However, at frother concentration above the
CCC (over 5 ppm), bubble sizes in three-phase measurements were on average 4.6% smaller (Figure
4a), while at concentrations below the CCC (2 and 5 ppm) bubbles were on average 8.8% smaller
(Figure 4b). It is not clear whether these lower values in the presence of solids were the
consequence of a more effective preservation of bubble size or just the effect of slightly different
operating conditions (i.e., lower gas flow rate or collector effects, when solids are present).
A program to demonstrate a recovery-Sb relationship also was performed on the 2.7 L Agitair cell
(Gomez, Mesias & Alvarez, 2016), which was operated at impeller speeds of 1440 rpm using the
fastest froth paddling rate once every 8 seconds, and a frother blend (MIBC, F-810 and F-1202, all
provided by Mathiesen, at 55, 40 and 5% by weight, respectively). One-kg samples of copper ore
with addition of collector (38 g/ton), diesel (15 g/ton), 0.35 g of calcium oxide, and the volume of the
selected frother solution to have a 65% pulp density, were ground in a Marcy ball mill. A grinding
time of 13.5 minutes, predetermined to have a product with 80% passing 150 µm, was utilized. The
grinding product was diluted to 30% pulp density in the flotation cell with the corresponding
volume of frother solution and the pH was adjusted to 9.5. Variations of Sb were achieved through
changes in frother concentration at constant gas flow rates, but also through variations in gas flow

8
rate at the same frother concentration. Measured bubble sizes and corresponding Sb's obtained at
the different gas flow rates, as a function of frother concentration, are included in Figures 5 and 6,
respectively; the same trends observed in the two-phase experiments are evident in these results.
The expected rapid bubble size decrease at frother concentrations below the CCC, and the
negligible effect above the CCC, with larger values as the higher gas flow rates were used, are
evident in both figures. Corresponding Sb's showed rapidly increasing values below the CCC, with
no much effect of gas flow rate and, above the CCC, larger Sb's values as increases in gas flow rate
had no effect on bubble size.

CONCLUSIONS AND FUTURE WORK

The works reviewed demonstrated that bubble size measurements performed in the same cell, in
two- and three-phase operation, have minor differences: sizes in the presence of solids were on
average 4.6% smaller at concentrations above the CCC and 8.8% smaller at concentrations below
the CCC. These differences are small and it is possible that they are a consequence of slightly
different operating conditions and not of a more effective preservation of bubble size when solids
are present.
The demonstration that, for a set of conditions, the same bubble size distributions are generated in
the presence of frother solutions or mineral pulps has enormous potential to make educated
selection of conditions for tests to measure ore flotation response. Future work in the form of a test
program specifically designed to measure bubble size in frother solutions and mineral pulps of
different characteristics, is necessary.

Figure 3 Comparison of bubble size measurements in two and three phases for analytical grade and
commercial frothers

9
 Figure 4 Comparison of bubble size measurements in two and three phases for several frothers: a) above
CCC, and b), below CCC

Figure 5 Three-phase bubble size measurement variations: a) as a function of frother concentration at constant
gas flow rates; and b) as a function gas flow rates at constant frother concentrations

10
Figure 6 Three-phase bubble surface area fluxes obtained: a) as a function of frother concentration at constant
gas flow rates; and b) as a function gas flow rates at constant frother concentrations

ACKNOWLEDGEMENTS
The authors want to recognize funding from CORFO Chile through the Sustainable Minerals
Institute – International Centre of Excellence – Chile, Project 13CEI2-2184413CI2-21844.

REFERENCES
Alvarez, J., Gomez, C.O., & Wen L. (2014) 'Gas dispersion characterization of a laboratory
mechanical flotation cell'. Proceedings of XXVII International Mineral Processing Congress IMPC 2014,
IMPC 2014 Organization, Santiago, Chile, Volume 1, Chapter 5, Paper 188
Azgomi, F., Gomez, C.O. & Finch, J.A. (2007) 'Correspondence of gas holdup and bubble size in
presence of different frothers', International Journal of Mineral Processing, vol. 83, pp. 1–11
Cho, Y.S. & Laskowski, J.S. (2002) 'Effect of flotation frothers on bubble size and foam stability',
International Journal of Mineral Processing, vol. 64, pp. 69–80
Finch, J.A., Xiao, J., Hardie, C. & Gomez, C.O. (2000) "Gas dispersion properties: Bubble surface
area flux and gas holdup", Minerals Engineering, vol. 13, no. 4, pp. 365–372
Girgin E.H., Do S., Gomez C.O. & Finch J.A. (2006) 'Bubble size as a function of impeller speed in a
self-aeration laboratory flotation cell', Minerals Engineering, vol. 19, pp. 201–203
Gomez, C.O. & Finch J.A. (2007) 'Gas Dispersion Measurements in Flotation Cells', International
Journal of Mineral Processing, vol. 84, pp. 51–58

11
Gomez, C.O., Castillo, P., Rebolledo, M. & Alvarez, J. (2013) 'Frother characterization in a laboratory
flotation cell', Proceedings of the Eighth International Copper Conference 2013, Instituto de Ingenieros de
Minas de Chile, Santiago, Chile, pp. 223–231
Gomez, C.O., Castillo, P., & Alvarez J. (2014) 'A frother characterization technique using a lab
mechanical flotation cell'. Proceedings of XXVII International Mineral Processing Congress IMPC 2014,
IMPC 2014 Organization, Santiago, Chile, Volume 1, Chapter 2, Paper 71
Gomez, C.O., Mesias, J., & Alvarez J. (2016). 'Bubble surface area flux and performance in
laboratory flotation testing' To be presented at the XXVIII International Mineral Processing Congress
IMPC 2016, Quebec City, Canada
Gorain, B.K., Franzidis, J-P. & Manlapig, E.V. (1997) 'Studies on impeller type, impeller speed and
air flow rate in an industrial scale flotation cell. Part 4: Effect of bubble surface area flux on flotation
performance', Minerals Engineering, vol. 10, no. 4, pp. 367–379
Gorain, B. K., Napier-Munn, T. J., Franzidis, J-P. & Manlapig, E. V. (1998) 'Studies on impeller type,
impeller speed and air flow rate in an industrial scale flotation cell. Part 5: Validation of the k-Sb
relationship and effect of froth depth’, Minerals Engineering, vol. 11, no. 7, pp. 615–626
Grandon, F., Alvarez J. & Gomez, C.O., 'Frother dosage in laboratory flotation testing', 8th
International Mineral Processing Seminar Procemin 2011, W. Kracht, R. Kuyvenhoven, S. Lynch-
Watson and G. Montes-Atenas, Eds., Gecamin Ltda., Santiago, Chile, 2015, 223-231
Grau R.A. & Heiskanen, K. (2002) 'Visual technique for measuring bubble size in flotation
machines', Minerals Engineering, vol. 15, pp. 507-513
Hernandez, H., Gomez, C.O. & Finch J.A. (2003) 'Gas dispersion and de-inking in a flotation
column', Minerals Engineering, vol. 16, pp. 739–744
Jameson, G. J. & Allum, P. (1984) 'A survey of bubble sizes in industrial flotation cells', report to
Australian Mineral Industries Research Association (AMIRA), pp. 32–37
Nesset J.E., Finch, J.A. & Gomez, C.O. (2007) 'Operating variables affecting the bubble size in
forced-air mechanical flotation machines', Ninth Mill Operators' Conference 2007, The Australasian
Institute of Mining and Metallurgy, Victoria, Australia, pp. 55–65
Randall, E.W., Goodall, C.M., Fairlamb, P.M., Dold, P.L. & O´Connor, C.T. (1989) 'A method of
determining the sizes of bubbles in two-and three-phase systems', J. Phys. E. Sci. Instrum., vol. 22,
pp. 827–833
O´Connor, C.T., Randall, E.W. & Goodall, C.M. (1990) 'Measurements of the effects of physical and
chemical variables on bubble size', International Journal of Mineral Processing, vol. 28, pp. 139–149
Yianatos J.B., Bergh L.G., Condori P. & Aguilera, J. (2001) 'Hydrodynamic and metallurgical
characterization of industrial flotation banks for control purposes', Minerals Engineering, vol. 14, no.
9, pp. 1033–1046

12
CHAPTER 6

Modeling, Simulation
and Design of Mineral
Processes
How to Avoid Errors in HPGR Product
Characterization
Felix Heinicke 1, Harald Günter 1 and Andre Kamptner 2
1. Köppern Aufbereitungstechnik, Germany
2. UVR-FIA, Germany

ABSTRACT
HPGR have been very successfully implemented in the mineral processing for grinding of various
ores. Currently the worldwide interest in pellet feed preparation of iron ore concentrates (re-
grinding) is rising as well. In addition to the known low specific energy consumption, high
availabilities of the grinding system and less environmental load (no water consumption), the
improvement of the pelletizing process is an additional bonus of the re-grinding HPGR.
It is known that the HPGR product can differ from ball mill product. One example is the possible
formation of agglomerates (flakes) by the HPGR grinding process depending on the ore properties.
The above mentioned advantages of the HPGR can only be verified if the applied analytical and de-
agglomeration methods for the fineness analysis are comparable. With regard to the above
mentioned possibility of agglomeration in the HPGR the fineness measurement methods need to
pay attention to this effect. This paper details the need of careful sample preparation for iron ores in
view of the results of laser particle size analysis, wet and dry screening as well as the measurement
of the specific surface according to Blaine. Several test results will be used and the error impact on
the industrial practice will be addressed.

1
INDRODUCTION
During recent years a worldwide trend towards comminution installations using High Pressure
Grinding Rolls (HPGR) has been established (Klymowsky, 1998). This comminution practice was
made famous in the 1980s by the research work of the German Professor Schönert. The technology
had its foundations much deeper in the past (Tundermann & Singer, 1968) (Schönert, 1966).
Köppern, as one supplier of HPGR, has worked for almost 120 years on roller presses, starting with
the formation of coal briquettes between two rollers. Subsequently, other fields of application, e.g.
for fertilizer compaction, emerged. The first HPGR grinding solutions were used in the cement
industry where the advantage of energy savings was very welcome on this medium cost product
(Klymowsky et al., 2002).
Nevertheless, it was a difficult step to transfer this grinding principle to minerals applications due
to the much higher wear caused by the abrasive materials. One of the first applications started in
iron ore concentrate grinding and were established for hard rock with improved wear lining of the
roller surfaces. Examples can be found for several plants (Odenwald, Oliveira & Goulart, 2005)
(Ehrentraut & Kudremukh, 2001). HPGR are available in a wide size range. Depending upon the
required plant throughput, grinding tests are necessary to predict the operational behaviour of the
machine for a certain material (Hees, 2005) (van der Meer & Maphosa, 2012) (Forsmo, 2007)
(Kortmann & Kurghardt 1973). The most important parameters are the specific throughput and
specific energy consumption which enable to size the HPGR for the required product fineness
(Eisele & Kawatra, 2006) (van der Meer & Oenol, 2014) (Heinicke, Günter & Lieberwirth, 2016). The
product fineness itself is defined by the transfer size to the downstream application such as
flotation, magnetic separation or pelletizing and will vary with the ore characteristics (McNab et.al.,
2009).
Due to the high applied pressure in the HPGR grinding process and depending on the ore
characteristics it is possible to create agglomerates (flakes) in the HPGR product. One example for a
gold ore is given in Figure 1.

Figure 1 example of a flake made of gold ore

Normally those flakes are easy to deagglomerate and will rupture during the transfer from HPGR
to the following belt (destroyed by drop impact). Nevertheless it is questionable if small
agglomerates could survive the transport regime and can find their also way to the laboratories.
There they could affect the fineness analysis.

2
The above mentioned advantages of the HPGR can only be verified if the applied analytical and
deagglomeration methods for the fineness analysis in the laboratories are comparable. The
following chapters therefore detail the need of careful sample preparation for iron ores in view of
the results of laser particle size analysis, wet and dry screening as well as the measurement of the
specific surface according to Blaine.

METHODOLOGY

Dry deagglmomeration
Both standard HPGR lab scale test or industry measurement deliver high amounts of samples.
Those samples contain a certain moisture range. For iron ore concentrates the value of 0-12 % can be
found on several worldwide installations with HPGR. Due to the high amount of sample of >100 kg
the material has to be representably splitted. It is common to dry the material before analysis as
well. During the drying process it is possible that the material creates additional packing on the
drying plates. Those agglomerates have be to dissolved.
To dissolve the agglomerates of the HPGR for dry analytics a lab scaled ball mill was used (Figure
2). With a length of 80 mm and a diameter of 90 mm approximately 100 g iron ore concentrate can
be deagglomerated. The deagglomeration speed is set to 80 rpm. To make the deagglomeration
more efficient special rubber balls are used. Thereby no additional comminution and good
deagglomeration is ensured.

Figure 2 lab scaled deagglomeration ball mill

Laser diffractometer measurement

The used device is a SYMPATEC Helos H2023 with selectively wet (SUCELL) and dry (RODOS)
dispersion unit (Figure 3). The technical information is presented in Table 1.

Figure 3 Laser Diffractometer SYMPATEC Helos H2023

3
Table 1 Details for laser diffractometer measurement

Producer Sympatec GmbH

Type HELOS H2023

Dry dispersion unit RODOS 1 – 4 bar dispersion pressure

Wet dispersion unit SUCELL 2 agitators; 60 watt ultrasonic bath

Typical measuring fluids air, water, isopropanol, ethanol

The dry dispersion is realized by the transport of particle in a pressurized airstream in two steps. In
the first step, particle – particle and particle – wall contacts will break up the agglomerates.
Furthermore in the second step an abruptly pressure decreasing at the transition from pressurized
airstream through the nozzle into the measuring chamber will eliminate the rest of the
agglomerates. This pressurized airstream can range from one through four bar.
In the wet dispersion unit, the breakage of the agglomerates is a combination of an ultrasonic bath
with adjustable ultrasonic energy and agitation of the suspension with adjustable number of
revolutions. The agitation makes sure that the suspension gets homogenized.
A big difference between wet and dry dispersion unit is the way to measure the particles. In dry
dispersion the sample is measured one time with a higher necessary sample mass. In wet
dispersion a circulation system is used which allows to lower the sample mass.
To make sure that the used laser is not blocked by too big particles a protection screening on
0.315mm is necessary.

Air jet screen

The air jet screening was done according to DIN 66165. The termination condition was set within
one minute to 0.1 % of the mass change based on the feed mass. The screen cuts for these
investigations are 20 µm, 32 µm, 45 µm, 63 µm, 100 µm and 200 µm.
The deagglomeration of flakes in the air jet screen is a combination of two effects. First, the contact
between air jet and particle which causes a takeoff from the screen. Second, caused by the negative
pressure, an acceleration towards the sieve and with this a contact between particle and screen
surface.
For any test, the negative pressure was set constantly to 60 mbar. The used air jet screen is shown in
Figure 4.

4
 Figure 4 air jet screen

Wet screen
For wet screen analysis, a vibrating screen machine from Retsch type AS 200 was used (Figure 5).
To ensure the comparability between wet screening and air jet screening, the same screen cuts as in
air jet screening are used. The screen time was set to 20 minutes and the vibrating amplitude to 2
mm. The water flow set constant to 0.5 l/min.

Figure 5 vibrating screen machine from Retsch type AS 200

Blaine measurement
Any determination of the specific surface according to Blaine was done according to DIN 66126-2.
All investigated products have been pre-screened at 0.315 mm. The undersize material was then
tested in Blaine device. For some material with residues on 0.315 mm the Blaine value was
corrected according to its residue mass. The time for each test was stopped with a stop watch. One
Blaine value is based on three times of measuring one material with three times of measuring one
sample.
As the Blaine device is not able to deagglomerate particles on its own physical measurement a
separate deagglomeration with the above mentioned lab scaled ball mill is essential.

5
 Figure 6 Manually Blaine device

RESULTS AND DISCUSSION

Dry laser diffractometer measurement

Figure 7 shows the particle size analysis of dry laser diffractometer for various deagglomeration
times with the lab scale ball mill. All curves are similar and do not vary. The same effect can be seen
by varying the pressure of the dispersion unit RHODOS according to Figure 8. It is to be considered
that the deagglomeration ability of the Rhodos unit is replacing any missing deagglomeration of the
ball mill unit. Therefore it is not necessary to use additional deagglomeration before which saves
lab time.

Figure 7 differences in dry laser diffractometer measurement for various deagglomeration times

6
Figure 8 differences in dry laser diffractometer measurement for various dispersion pressures of Rhodos unit

The reproducibility of the above mentioned laser diffractometer measurements was tested
according to ISO 13320. A very small error could be found according Figure 9. This emphasizes that
the dry laser diffractometer measurement is a good method for analysis of iron ore. While the time
for one analysis is quite short with about five minutes only the purchase price is very high for a
small industrial lab in comparison to other analysis methods. For comparing industrial results with
lab scale test work it has to be remarked that different units might produce different results. This
influence has to be acknowledged in industrial guarantee contracts.

Figure 9 accuracy of dry laser diffractometer measurement after 10min deagglomeration

Wet Laser diffractometer measurement

Figure 10 shows the particle size analysis of wet laser diffractometer for various deagglomeration
times with the lab scale ball mill as well as measurements with and without ultrasonic
deagglomeration supply of the SUCELL unit. Also an additional measurement with isopropanol is
given to show the variance of several measurement fluids. The particle size curves for wet laser
diffractometer measurements show higher variations. Also the reproducibility (ISO 13320) is out of
the range (Figure 11).

7
 Figure 10 differences in wet laser diffractometer measurement by use of ultrasonic and Isopropanol

Figure 11 accuracy of wet laser diffractometer measurement after 10 min deagglomeration

The error range of wet laser diffractometer measurement was measured with up to 10%. The client
is normally aiming at a certain fines degree in the product. If a lower amount is measured the
HGPR would have to be bigger in size depending on the installed flowsheet and the possible higher
amount of recirculation load. A difference in 10% can easily result in an increase of 500.000 € in the
CAPEX of a HPGR.
More variations have been found when measuring Magnetite samples in wet and dry ambience.
The shapes of the curves are very different. It was considered that due to the magnetic valves in the
SUCELL unit Magnetite particles might be astrayed. As the content of Magnetite and Hematite can
vary depending on the mine conditions the wet diffractometer measurement might product non
comparable results. Thereby the wet laser diffractometer measurement method was not found to be
useful during this test series with iron ore concentrates.

8
 Figure 12 differences of wet and dry laser diffractometer measurement for Magnetite ore

Air jet screen measurement

Figure 13 and Figure 14 show the particle size analysis and reproducibility of an air jet screen for
various deagglomeration times with the lab scale ball mill. With remarks to Figure 7 and Figure 9
the method can be classified as also producing representable results as the laser diffractometer
measurements.

Figure 13 differences in air jet screen measurement for various deagglomeration times

9
 Figure 14 accuracy of air jet screen measurement

The challenge with this method is to be seen in the time consumption. Figure 15 show the
importance to act in accordance with the norm DIN 66165 in view to the screen time. The
cumulated percent passing is varying for more than 15% from five to 40 minutes in analysis time.
Thereby this method is not useful for high amounts of sample analysis. As different results of
various machines are not known to the authors this method is useful for comparing industrial
results with lab scale test work in relation with single guarantee values.

Figure 15 influence of screening time in air jet sieving for 0,020mm screen cut

Wet screen measurement

Figure 16 shows the particle size analysis and reproducibility of a wet jet screen for various
deagglomeration times as well as measurements with different fluids (water and isopropanol) and
additional ultrasonic support. Again the reproducibility is very good (Figure 17). Even if the time
consumption of up to four hours is remarkable higher than above mentioned analytical methods
this option is good for an industrial application as it is commonly used on mine sides.

10
Figure 16 differences in wet screen measurement for various deagglomeration times and different fluids and
ultrasonic impact

Figure 17 accuracy of wet screen measurement

Blaine measurement
Figure 18 shows the particle size analysis according to Blaine method for several material and
various deagglomeration times. The method shows high differences in results of up to 400 cm²/g
depending on the preparation method while the representability was very good for the single
measurement. Thereby the exact method has to be fixed during contract negotiations if used for
guarantee purposes. As the measurement device is commonly used in iron ore application it is still
a good option to get an overview of industrial performance of HPGR in iron ore concentrate
grinding.

11
 Figure 18 results of Blaine measurement for various deagglomeration times and different materials

CONCLUSIONS
This paper presents different results of laser diffractometer, air jet screen, wet screen as well as
Blaine measurements. The particle size distribution results vary depending on the analysis device
but also depending on the deagglomeration method. Differences of up to 15% or 400cm²/g have
been found with several materials. This might lead to an increase of the size of HPGR which has to
be installed in industry application. Due to the nature of heavy duty machinery this will increase
the OPEX of the machine.
There is no wrong measurement methodology nor is there the best method if compared to each
other. So the is no instruction manual on how to avoid errors but it can be stated that the methods
need to be depending on the production goals of the client. There is no systematic error which can
be avoided but the method for analysis including deagglomeration procedure has to be discussed
with the client. In view of evaluation of test work and industrial performance of HPGR the
procedure has to be fixed before deviation between industry and test results are forming claim
battles.

ACKNOWLEDGEMENTS
This paper summarizes the work of the bachelor thesis by Tony Lyon at the Technical University
Bergakademie Freiberg in 2016 (Lyon, 2016). This work was funded by Köppern and UVR-FIA. The
authors would like to thank Tony Lyon for the huge amount of practical work and good theoretical
input and Köppern as well as UVR-FIA for the right to publish this paper.

REFERENCES
Ehrentraut, Kudremukh: Experience with a roller press in the pellet plant of Kudremukh iron ore
company Ltd. Aufbereitungstechnik No.10 (2001), S. 469-476
Eisele, Kawatra: Design of iron ore comminution circuits to minimize overgrinding. Advances in
Comminution. Society for mining metal (2006), S. 309-320

12
Forsmo: Influence of green pellet properties on pelletizing of magnetite iron ore. Doctoral Thesis.
Lulea Universität of Technology. 2007
Hees: Untersuchungen zum Verhalten von Eisenerzkonzentraten bei einer Gutbettbeanspruchung.
Diplomarbeit. RWTH Aachen, 2005
Heinicke, Günter, Lieberwirth: Modelling of HPGR Edge recycling with progressive grinding data.
Proceedings of Comminution in Cape Town, 2016
Klymowsky: Roller press installations in iron ore. Report from KHD to World Mining Equipment
(1998), S. 1-7
Klymowsky, Patzelt, Knecht, Burchardt: Selection and sizing of High Pressure Grinding Rolls. SME
Handbook 2002, S. 1-40
Kortmann, Burghardt: Qualitätsmerkmale von Eisenerzpellets und Möglichkeiten der
Beeinflussung bei verarbeitung hämatitischer Erze. Aufbereitungstechnik Nr. 12 (1973), S. 803-810
Lyon, T.: The influence of deagglomeration on particle size distribution analytics. Bachelor Thesis.
Technical Universaty Bergakademie Freiberg, 20016
McNab, Jankovic, David, Payne: Processing of Magnetite iron ores - comparing grinding options.
Iron ore conference in Perth (2009), S. 277-286
Odenwald, Oliveira, Goulart: HPGR applications in iron ore pellet plants in Brazil. Paper at Randol
Pacific conference on innovative metallurgy in Perth (2005), S. 1-6
Schönert: Einzelkorn-Druckzerkleinerung und Zerkleinerungskinetik. Untersuchungen an
Kalkstein-, Quarz-, und Zementklinkerkörnern des Größenbereiches 0,1-0,3mm. Dissertation. TH
Karlsruhe, 1966
Tundermann, Singer: The flow of iron powder during roll compaction. Powder Metallurgy Vol.11
(1968) No.22, S.261-293
van der Meer, Maphosa, W.: High pressure grinding moving ahead in copper, iron, and gold
processing. The Journal of The Southern African Institute of Mining and Metallurgy Vol 112 (2012),
S. 637-648
van der Meer, Oenol: Pellet feed grinding by HPGR. Proceedings of Comminution 2014 in Cape
Town, S.1-18

13
Simulation of the Flotation Process in a Hybrid Cell
with the OpenFOAM CFD Framework
Gerhard Holzinger1, Stefan Pirker 2 and Lukas Petzold3
1. K1MET, Austria
2. Johannes Kepler University Linz, Department for Particulate Flow, Austria
3. Primetals Technologies, Austria

ABSTRACT
The process of flotation is widely used in many fields, e.g. mineral processing, wastepaper deinking
or wastewater treatment. The aim of this work is to implement a method for simulating the
macroscopic flotation process within the open source CFD framework of OpenFOAM.
In computational fluid dynamics (CFD) the flow domain is discretized into individual finite
volumes. The flow equations are solved for the discretized domain. The two-phase flow within the
flotation apparatus (slurry and gas) is modelled with an Eulerian two fluid model. A model for the
flotation kinetics is implemented according to established models from literature. The models for
the sub-processes of flotation (collision, attachment and stability) are evaluated locally for each
finite volume. Due to the modular design of the CFD framework, the implemented flotation model
is extendable and flexible in use.
The CFD model is applied to a hybrid flotation cell which combines a pneumatic flotation cell and a
flotation column. The local nature of the model proves particularly useful for the hybrid cell, which
combines two regimes in one apparatus.
Modelling flotation within the frame of CFD allows for studying the impact of the local flow on the
flotation performance. This might provide a useful tool to improve the design of flotation columns
and cells.

1
INTRODUCTION
Flotation is a physical process used to separate materials by utilizing differences in their surface
properties. In order to achieve separation, air bubbles have to be introduced and brought into
contact with the slurry.
Flotation columns and pneumatic flotation cells have been gaining increasing importance in the
beneficiation of minerals due to their low operating cost (Wills & Napier-Munn, 2006). Flotation
columns are generally operated in counter-current flow, i.e. the slurry enters the vessel at the top
whereas air is introduced at the bottom. In pneumatic flotation cells slurry and air brought into
contact outside the main vessel by a contactor. The contactor creates flow conditions with a high
turbulence intensity to bring slurry and bubbles in intense contact.
Macroscopic flotation model have long been used to investigate the flotation process and to predict
flotation efficiency. However, as the flow conditions within the flotation apparatus boil down to
single numbers to be inserted into an expression for the flotation rate constant, this approach
becomes more and more problematic for flotation devices with complex geometries and operating
conditions.
Computational fluid dynamics (CFD) has been increasingly used to study the conditions within
flotation machines (Koh et al., 2003) or the flotation process itself, either individual sub-processes
(Wierink, 2012) or the whole process (Karimi et al., 2014).
In this work the flotation rate are computed from local properties of the two-phase Eulerian flow
simulation. The flotation process is modelled by solving a population balance for a finite number of
particle size-classes. Local bubble size is determined by solving a simplified population balance for
bubble sizes. Thus, bubbles and particles do not have to be assumed mono-disperse.

FLOTATION MODELLING

Flotation Kinetics
Flotation is widely modelled as a first order kinetic process (Deglon, 1998). Eq. (1) shows the
simplest possible model for the flotation process. This model, however, considers only collection of
particles. Bloom and Heindel (2003) derived a kinetic model, which also considers detachment of
already collected particles from its bubble. Thus, the entity of floatable particles needs to be divided
into the populations of free and attached particles. Equations (2) and (3) describe the evolution of
the populations of the free and attached particles in terms of the number densities and . The
bubble load factor is defined as the ratio of the number density of bubble-particle aggregates
and the total bubble number density .

(1)
(2)
(3)
(4)
(5)
( )

2
The rate constant , or the rate constants and , respectively subsume all information computed
by the models for the sub-processes of flotation. The flotation rate constant is computed from a
collision frequency between particles and bubbles and a collection probability. The collection
probability accounts for non-ideal collisions, attachment and stability (Deglon, 1998). The rate
constant accounts for detachment of already collected particles depending on a detachment
frequency and the probability of stability .

(6)
( ) (7)

The probabilities of collision , attachment and stability are computed by models for the
respective sub-processes. Also the frequencies of collision and detachment need to be
computed by a model.

Sub-process models
The probability of collision between a particle and a bubble was computed by the model of Yoon
and Luttrell (1989). This model is based on a semi-analytical stream function describing
intermediate flow around the bubble, by combining solutions for Stokes (creeping) flow and
potential flow. The collision efficiency is computed from the bubble Reynolds number and the
diameters of bubble and particle involved.

( )( ) (8)

Bubble-particle attachment is governed by the time a particle is in contact with a bubble after
collision. The time needed for the liquid film to drain and to establish the three-phase contact is
referred to as induction time . If the particle is longer in contact with the bubble than the induction
time, attachment will occur. The induction time is modelled using the model of Koh and Schwarz
(2006), which depends on the particle diameter and the contact angle .

( ) (9)

( )
( )

3
 ( )

The stability model is based on a balance of attaching and detaching forces on an attached particle,
which is determined via a modified Bond number (Schulze, 1984). Bloom & Heindel (2003)
proposed an empirical constant which assumes a value between 0 and 1. Among others, the
forces acting on a particle attached to a bubble are related to gravity , surface tension and the
acceleration experienced by the aggregate.

[ ( )] (10)

( ) ( ) ( )
| ( )|

( ) ( )
⁄
⁄
( )

The collision frequency was calculated by the simplified model of Schubert & Bischofberger
(1979) proposed by Duan, Fornasiero & Ralston (2003) was used. The collision frequency is
computed from the dissipation of turbulent kinetic energy , the viscosity of the pulp and the
densities of the liquid phase and the bubbles.

⁄ ⁄
( ) (11)
⁄

The detachment frequency was computed by the model proposed by Bloom & Heindel (2003).
They suggested an empirical model constant .

⁄ ⁄
√ ( ) (12)

THE NUMERICAL MODEL

For the numerical simulations the open source CFD framework OpenFOAM has been employed
(Weller et al., 1998). The standard two-phase Eulerian solver twoPhaseEulerFoam was used to
simulate the two-phase flow within the flotation device (Rusche, 2002). Equations (13) and (14) are
the continuity and momentum equation for the phase i with the volume fraction , the velocity

4
 ,.the stress tensor , the density , the pressure , the gravitational acceleration and the
momentum exchange term .

( ) (13)
( ) ( ) (14)

By default, Eulerian two-phase flow models are mono-disperse, i.e. only one bubble size is
considered. With the interfacial area transport equation (IATE) (Ishii et al., 2005), OpenFOAM
allows for the consideration of the two bubble sizes within the hybrid flotation cell. Instead of
solving for two gas phases, one with small and one with large bubbles, one additional transport
equation for the interfacial area is solved. From this the local Sauter mean diameter of the bubbles
can be evaluated. This model offers a compromise between considering the gas phase to be
monodisperse and the large computational cost added by fully considering an additional gas phase.
The flotation model was based on the general species transport equation for multi-phase flows.
Equation (15) shows the transport equation for the mass fraction with a source term on the
right hand side. This source term is computed by the flotation model as particles are collected of
lost by air bubbles.

( ) ( ) (15)

In the case of the flotation process there are two species: the free particles suspended in the liquid
phase and the collected particles attached to the bubbles.
As the flotation process is strongly size dependent, we consider a number of size classes for the free
and attached particles. Thus, the size distribution of the floatable particles is discretized into a finite
number of size classes. For each size class, the flotation model computes the respective flotation rate
constants.
Equations (16) and (17) show the transport equations for the free and attached particles of the size
class . The source terms on the right hand sides are equal but of opposite sign, as particles are
neither created nor destroyed as they are collected by bubbles. Particles are simply transferred from
the population of the free particles to the population of the attached.

( ) ( ) (16)
(17)
( ) ( )
(18)

The implementation in OpenFOAM makes heavy use of the object oriented programming
paradigm to allow for an easy extensibility of the model.

5
The source terms and on the right hand sides of equations (16) and (17) are taken from the
equation (2), which is rearranged in terms of the size class’s mass fraction instead if its number
density .

( ) (19)

THE HYBRID FLOTATION CELL

The hybrid flotation cell combines the principles of a pneumatic flotation cell and a flotation
column (Krieglstein & Grossmann, 2010). The term hybrid flotation refers to the combination of the
pneumatic and columnar operating principle. In literature, the term hybrid flotation is also used to
describe other flotation cells combining two or more operating principles, e.g. the combination of
flotation and microfiltration in wastewater treatment (Peleka et al., 2006).
Figure 1 shows a schematic representation of the hybrid flotation cell. The central cylindrical
section represents the columnar flotation part of the flotation device, referred to as the columnar
section. The outer, upper section of the apparatus represents the pneumatic flotation part of the
device, referred to as the pneumatic section. Here, a slurry and air are mixed in the ejectors which
are connected to the pneumatic section. The pneumatic section joins the columnar section roughly
midway in vertical direction. Thus, the columnar section can be divided into a lower columnar
section and an upper columnar section. The ejectors join the main body tangentially to induce a
rotating flow in the pneumatic section

Figure 1 The configuration of a hybrid flotation column; the blue lines schematically indicate the mean
trajectory of air bubbles introduced; the grey lines indicate the slurry flow

Due to the different flow conditions in the ejectors and the lower columnar section, the hybrid
flotation cell can operate with different bubble sizes, i.e. with fine bubbles generated by the ejectors
and larger bubbles stemming from the aerators within the lower columnar section.

6
NUMERICAL SIMULATION
Figure 2 shows a cross section of the two-phase flow within the hybrid flotation cell. The bubble
plumes exiting the ejectors at the top and emanating from the aerators at the bottom can be clearly
seen. Also the interaction between the flow and the bubble plume in the intersection region
between pneumatic section and columnar section can be seen.
The flotation simulation was initialized by the solution of a precursory two-phase flow simulation
to have an already established flow present. This means, however, that the flotation process
magically kicks in, in the already operating flotation cell. Since, we are interested in the steady-state
performance of the flotation apparatus, this flawed transient behavior at the start of the simulation
is negligible.

Figure 2 Full-sectional view of the interior of the FTC 32 flotation column; the iso-volume is coloured
according to the value of the mean volume fraction field of the gas phase

In order to determine the separation efficiency, two zones were defined in the finite volume mesh
of the hybrid flotation cell. All cells neighboring an inlet formed the inlet region. Analogously, all
cells at the outlet formed the outlet zone. By logging the volume-averaged mass fraction per size
class within the inlet and the outlet zone, the relevant data for examining separation was gained.
Equation (20) shows the definition of the recovery , which is based on the mass fractions of a
particular size fraction i on the inlet and outlet of the hybrid cell.

(20)

Figure 3 shows the normalized particle masses over time. As expected the flotation behavior is
strongly size-dependent. Furthermore, we see the establishment of a steady-state. Since the hybrid
flotation cell is continuously operated, the mass of the floatable particles never reaches zero as in a
batch process.
Figure 4 shows the recovery , which was computed as the ratio of the inlet mass fraction and the
outlet mass fraction averaged over the last 10 seconds of simulated time.

7
Figure 3 Normalized particle mass in a control Figure 4 Recovery for particle diameters from 7.5
volume at the outlet of the hybrid flotation cell µm to 120 µm

CONCLUSIONS
A flotation model has been implemented within the CFD framework of OpenFOAM. The flotation
model was demonstrated on the example of the hybrid flotation cell of Primetals Technologies.
The simulation results showed the expected size-dependent separation behavior as expected.
However, further work is needed for validation.
A CFD-based flotation model can be used not only to get a look into the flotation apparatus, it
could also be used to determine parameters for a zone-based macroscopic model.

ACKNOWLEDGEMENTS
The authors acknowledge the support provided by Primetals Technologies. They thank all partners
for the scientific discussion and the productive co-operation.

REFERENCES
Bloom, F. & Heindel, T. J. (2003) ‘Modeling flotation separation in a semi-batch process’, Chemical
Engineering Science, vol. 58, pp. 353-365
Deglon, D. A. (1998) A hydrodynamic investigation of fine particle flotation in a batch flotation cell, PhD
Thesis, University of Cape Town
Duan, J., Fornasiero, D. & Ralston, J. (2003) ‘Calculation of the flotation rate constant of chalcopyrite
particles in an ore’, International Journal of Minerals Processing, vol. 72, pp. 227-237
Ishii, M., Kim, S. & Kelly, J. (2005) ‘Development of interfacial transport equation’, Nuclear
Engineering and Technology, vol. 37, pp. 525-536
Karimi, M., Akdogan, G. & Bradshaw, S. M. (2014) ‘A computational fluid dynamics model for the
flotation rate constant, Part I: model development’, Minerals Engineering, vol. 69, pp. 214-222

8
Koh, P. T. L., Schwarz, M. P., Zhu, Y., Bourke, P., Peaker, R. & Franzidis, J. P. (2003) ‘Development
of CFD models of mineral flotation cells’, Proceedings of the Third International Conference on CFD in
the Mineral and Process Industries
Koh, P. T. L. & Schwarz, M. P. (2006) ‘CFD modelling of bubble-particle attachments in flotation
cells’, Minerals Engineering, vol. 19, pp. 619-626
Krieglstein, W. & Grossmann, L. (2010) ‘A newly developed flotation technology for increased
recovery – especially in the finest particle fractions’, Proceedings of Cu 2010, vol. 7, pp. 2793-2806
Peleka, E. N., Lazaridis, N. K., Mavros, P. & Matis, K. A. (2006) ‘A new hybrid flotation-
microfiltration cell’, Separation Science and Technology, vol. 41, pp. 3229-3243
Rusche, H. (2002) Computational fluid dynamics of dispersed two-phase flows at high phase fractions, PhD
Thesis, Imperial College, London
Schubert, H. & Bischofberger, C. (1979) ‘On the optimization of hydrodynamics in flotation
processes’, Proceedings of the 13th International Mineral Processing Congress, vol. 2, pp. 1261-1284
Schulze, H. J. (1984) Physico-chemical elementary processes in flotation: an analysis from the point of view
of colloid science including process engineering considerations, Elsevier, Amsterdam, New York
Weller, H. G., Tabor, G., Jasak, H. & Fureby, C. (1998) ‘A tensorial approach to computational
continuum mechanics using object-oriented techniques’, Computers in Physics, vol. 12, pp. 620-631
Wierink, G. A. (2012) A computational framework for coupled modelling of three-phase systems with soluble
surfactants, PhD Thesis, Aalto University
Wills, B. A. & Napier-Munn, T. J. (2006) Mineral processing technology, Elsevier Science & Technology
Books

9
Comparative Analysis of Grinding Efficiency and
Classifications Curves Using JKSimMet
Miguel Becerra
El Soldado, Anglo American, Chile

ABSTRACT
El Soldado is an operation of Anglo American Copper located in the Valparaiso region, Chile,
which processes copper sulfides at a rate of 8.1 mt/year according to the installed capacity at the
concentrator plant. This plant has conventional (ball mills) and SAG grinding technologies (Single
Stage circuit). In the case of the former, there are primary, secondary and regrinding circuits
operating with different ball sizes and processing different feed sizes distributions depending on
the location in the circuit plant. As classification equipments El Soldado has different types of
operating hydrocyclones in differents stages: flat bottom, conical bottom and ReCyclones in
different dimensions, operating with different feed flowratess and particle sizes distributions.
From surveys conducted on each circuit and using JKSimMet software, all the mentioned
equipment was modeled and respective grinding efficiency curves were obtained (represented by
the R/D vector in conventional grinding and ri in SAG circuit) and classification efficiency (derived
from the Nageswararao model parameterization for each hydrocyclone).
This article presents a comparative analysis of these curves, providing an important bibliographical
guide with industrial information that also allows evaluating optimization strategies quickly
implemented whose qualitative effects than are also presented. Finally, the differences and causes
in results between similar circuits are analysed considering design and operational strategies.

There is no full article associated with this abstract.

1
Reduction and Fracture Analysis of a Tungsten Ore
and its Use for Fundamental Liberation Modelling
Lorena Guldris, Magnus Bengtsson and Magnus Evertsson
Chalmers University of Technology, Sweden

ABSTRACT
Nowadays there are numerous tungsten applications and uses, due to this the global demand for
tungsten has increased the extraction and production of this rare earth mineral. The liberation of
tungsten is achieved by using comminution equipment, e.g. crushers, there is a lack of knowledge
how tungsten is liberated by means of compressive breakage.
Even though it is possible to use find comminution models for explaining the size reduction, there
is also need to implement a mineral liberation model into theses comminution models. This paper
will focus on a fundamental model that combine and integrate the size reduction with the mineral
liberation in early stages of the breakage process.
This work focuses on analysing the liberation characteristics that occur by means of interparticle
breakage. The work includes two parts, firstly the integration of the measured data into a size
reduction model, and secondly a reflexion and understanding of the liberation and fracture
mechanism during the breakage.
Results from the work show that in order to address multi component modelling it is necessary to
focus on controlling factors such as dominating quantities of different minerals, breakage
mechanism in different bounded minerals and how this is connected with the liberation of them. It
also was observed that the model could be used for determining the choice of the compression ratio
in crushing for optimising liberation.

1
INTRODUCTION
An ore is a naturally occurring mineral or aggregate of minerals that contains a valuable or useful
substance that could be extracted, in general metals. In order to release or liberate the valuable
minerals, it is necessary to reduce all the particles to small sizes. All these particles consist of metal
rich grains and coarser/gangue minerals. The procedure to achieve this reduction is known as
comminution and it includes complex processes such as crushing and grinding. The mining
operation for comminution requires high amount of energy, that represents 30% to 50% of the total
energy consumption of the entire mining process (La Nauze, 2002; Tromans, 2008), which means
that it is the principal operational cost. All materials, including the gangue and the minerals of
interest is typically processed and reduced in size. The amount of energy consumed and its
associated cost can be reduced by an earlier separation of the minerals or by removing the non-
valuable gangue.
In order to accomplish an earlier separation by removing of the non-valuable gangue, one needs a
thorough understanding of the how the liberation occurs, taking into consideration the breakage
mechanisms, the presence of micro fractures (or micro cracks), mineral associations and the
behaviour of different minerals in the earliest stages of the process, i.e. the crushing stage.
One important concept is the breakage mechanism. An understanding of the breakage mechanism
will open up new possibilities regarding the design and optimization of comminution circuits. In
order to understand the breakage mechanism, the size distribution, shape distribution and
liberation characteristics play an important part in every stage of the reduction process. Fandrich
(1996) established different types of breakage of ore, which include random and selective
behaviour. In the random breakage, the fracture will occur randomly, without considering the
different phases present in the ore, allowing the composition of all particles to remain constant. In
contrast, the second type is selective breakage, in which the fracture is preferential. There are two
forms of non-random breakage in particle bed breakage: interfacial breakage (grain/phase
boundary fracture) and preferential breakage. Both types have different breakage characteristics,
different rates. In interfacial breakage, the fracture occurs between different grains or between the
phase boundaries, allowing the preferential liberation of the phases. Moreover, preferential
breakage occurs when the characteristics of the phases are different, which generates breakage at
different rates due to the fact that the ore contains soft minerals and a hard minerals, in which case
the soft minerals will break more rapidly than the hard minerals, generating a different
concentration in the composition of the fine fraction and of the coarse fraction.
The presence of micro-cracks in the solid rock is significant in the liberation process. Nielsen and
Malvik (1999) showed the importance of the micro-cracks: fragments with a high density of micro-
cracks will be crushed more easily and hence allow an earlier liberation of minerals. Micro-cracks
are formed either simultaneously or before the fragments forming macro-fractures. The
heterogeneity of the rock and the presence of different minerals will have a considerable impact on
the presence of micro-cracks. Different minerals have different mechanical properties. This means
that the behaviour of each mineral will be different in the presence of compressive forces. Minerals
with less strength will be linked to the locations of the micro-cracks due to the creation of an
elevated number of fractures in the applications of strain rate loading, as explained by Djordjevic
(2014).
The surrounding minerals play an important role in the liberation of particles. Djordjevic (2014)
also proposed that, due to the large difference in mechanical properties of valuable minerals and
gangue, the ore matrix could exhibit preferential breakage. The elastic properties of each mineral

2
(valuable minerals and gangue) will be significant in terms of deformations that will result in the
development of micro-cracks along the minerals boundaries.
Guldris (2016) proposed an empirical model that classify and structure how the liberation occurs
during the application of compressive forces, considering the surrounding minerals (major and
minor minerals). This model shows that, after the application of a compression force, a particle size
distribution will be generated. Through the observation of each size fraction, it was possible to
structure associations and identify the presence of micro cracks.
The aim of this study is to evaluate and classify the mechanism of particle liberation, for tungsten
mineral, during the early steps of the process where the liberation is minor but the initial size
reduction and the chosen breakage mechanism could affect the liberation performance in later
stages. One of the motives for understanding how the breakage mechanisms influence liberation is
that if, for example, gangue could be removed early in the process, and the energy consumption
may decrease dramatically. One hypothesis is that the applied compression ratio may influence the
composition of liberated ore and gangue particles. The ability to selectively liberate minerals from
surrounding phases would be a significant advancement in mineral processing technology.

METHODOLOGY
The experiments used in this paper were conducted in a laboratory and are divided into two parts.
The first part is laboratory compressive crushing involving interparticle breakage as Evertsson
(2000) presented in his research. The other part of the experiment focuses on mineral
characterization using scanning electron microscope (SEM) and geochemical analysis (bulk
chemical analysis). The material used in these experiments is a tungsten ore from a mine located in
southern Europe (Leucogranite ore hosted in metamorphic rocks).
The mechanical characterization of the samples includes compressive breakage tests for
interparticle breakage (IPB) or bed breakage. The mineralogy characterization includes an SEM-
EDS scan for high-density mineral and geochemical analysis (bulk chemical analysis).

Compressive breakage test (Interparticle breakage)

The compressive breakage tests were performed in a hydraulic press (Instron 400RD) using a piston
and die test procedure originally presented by Evertsson (2000).
The basic concept of compressive breakage is to analyse the rock behaviour in terms of the
breakage, particle size distribution and energy consumption after compressive forces. A specific
amount of material is placed in a cylinder to apply compression forces. The bed height will be
generated through the space or thickness occupied by the material. During the compression test,
the displacement of the piston and the compression ratio are fixed values, which will allow
information such as the generated particle size distribution and the energy used to generate the
breakage to be understood and extracted. The definition of the compression ratio is presented in
equation (1).
(1)
⁄

Where:

b = Bed height s = displacement of the piston s/b = compression ratio

3
The particle size distribution is obtained by sieving the resulting material. The series of breakage
applied in the experiment was conducted according to the test plan developed by Evertsson (2000).
A schematic illustration of IPB is shown in Figure 1.

Figure 1 Illustration of Interparticle breakage before and after the compression

The samples used for each test were approximately 350 particles corresponding to a size fraction of
11.2 - 8.0 mm.
The test was performed with different compression ratio values. For each compression value, the
samples were compressed and sieved a number of times to analyse the particle size distribution as a
function of the compression ratio. The tested values for s/b were 0.10 (the test was repeated 5
times), 0.15 (4 times), 0.20 (4 times), 0.25 (4 times), 0.30 (2 times with one added compression of
0.15) and 0.35 (2 times). In Figure 2, the test procedure for the experiment is described.

Figure 2 General test procedure for Interparticle Breakage (IPB)

4
Mineralogical analysis

SEM-EDS
For the study of mineralogy, scanning electron microscopy (SEM-EDS) with energy-dispersive x-
ray spectroscopy analysis was used. The analysis was performed using a Hitachi s-3400n scanning
electron microscope coupled with an oxford instrument x-max EDS.
The evaluation of the mineral liberation was performed by the study of samples with different size
distributions (4.00, 1.00, 0.25, 0.063 mm) that were previously compressed using different
compression ratios (10%, 20% and 30%).
The SEM analysis offers a qualitative result, where it is possible to see the minerals that the rock
consists of, and the different associations between them. Nevertheless, it is necessary to perform a
quantitative analysis to be able to measure the quantity of each element, in this case, tungsten. The
quantitative data was obtained through the geochemical analysis (also called bulk chemical
analysis).

Geochemical Analysis
A geochemical analysis or bulk analysis was performed in the samples with a different size
distribution (4.00, 1.00, 0.25, 0.063 mm) that was compressed previously using 30% of compression
ratio.
The rock analysis performed corresponds to fused bead, acid digestion and ICP-AES and fused bed,
acid digestion and ICP-MS. These tests offer analytical data regarding composition and content of
tungsten, or other element, in the samples that were broken using different compression ratios and
with different size distributions.

RESULTS AND DISCUSSION

Compressive breakage test

The particle size distributions, cumulative and frequency for the interparticle breakage test product
obtained with a 20% compression ratio are shown in Figure 3. As expected, the particle size
distribution is dependent of the compression ratio and the number of cycles.

5
 Figure 3 Product particle size distributions for the IPB 20% test series (cumulative and frequency)

Mineralogical analysis
The ore is a granite dominated by mayor minerals as quartz (SiO2), albite (NaAlS3O8), muscovite
(KAl2(AlSi3O10)(OH)2), biotite (K(Mg,Fe)3(AlSi3O10)(F,OH)2), k-feldspar (KAlSbO8) and cholorite
((Mg,Fe)3(Si,Al)4O10(OH)2·(Mg,Fe)3(OH)6). The ore present a heterogeneous composition with
presence of veins rich in some minerals (specially the main ore minerals).
Main ore minerals are wolframite ((Fe, Mn)WO4) and scheelite (CaWO4). The accessory minerals
present are apatite (Ca5(PO4)3(F,Cl,OH)), zircon (ZrSiO4), titanite (CaTiSiO5), monazite ((Ce,La)PO4)
and rutile (TiO2). The mineral phases identified by SEM*EDX analysis are shown in Figure 4.

Figure 4 SEM micrographs of the bulk ore showing morphologies and mineralogy

The figures presented below correspond to the observation though SEM to different samples after
compression. Figure 5A shows particles of scheelite in associations with other minerals. In SEM
pictures, the tungsten minerals (scheelite or wolframite) are recognized because are the lighter and
brighter comparing with other minerals. Generally, scheelite was found together with muscovite,

6
K-feldspar, quartz, apatite and albite. Normally an entire area, surrounded by other minerals that
have more/less presence of fractures or micro-fractures. Considering the definition of “the locking
chart of Amstutz” regarding the types of textures, the scheelite present in this ore correspond to a
simple locking with rectilinear or gently curved boundaries, resulting in an easy liberation.
Likewise, it was observed that k-feldspar and albite have more microfractures and fractures in
comparison with quartz and other minerals. This could be observed in the figure 5A, where the k-
feldspar (light grey) exhibit fractures when is surrounding the half of the scheelite grain, which
suggests an easy liberation in following steps of the process. It was found totally liberated scheelite
particles at 0,25 and 0,063 mm, as shown in Figure 5B.

Figure 5 A- Particle integrated by scheelite mineral in albite and quartz observed through SEM after a
compression ratio of 30% considering the size fraction particle of 1 mm. B- Liberated grain of tungsten
(Scheelite mineral) observed through SEM after a compression ratio of 10% considering the size fraction
particle of 0.063mm

The other main ore mineral found is wolframite. One of the differences between these two minerals
is the content of tungsten and purity. It was observed that the wolframite is mixed with others
minerals rich in iron such as for example iron oxide and/or iron silicates. Figure 6 A shown a
composite particle of wolframite, K-feldspar, quarts, albite, iron silicates, muscovite, and biotite.
The appearance of the grains of wolframite is totally different to the scheelite. The wolframite is not
an entire grain, contrary to that is an extension that crosses different minerals and grains, like as a
veins or branches that are occupying a crevice or fissure in the particle. According to the types of
textures, the wolframite corresponds to veins-like and network type, which suggests that a
complete liberation would be fairly difficult. Generally, when one particle has wolframite in it, all
the particle contains small proportions of it blended in forms of small grains or veins that cross the
particle. Some examples of this formation are shown in Figure 6 where the wolframite crosses other
minerals as k-feldspar, muscovite, and/or albite. In other cases, it is placed in the grains boundaries.
This analysis indicates that the liberation of wolframite is difficult; it will require the reduction of
the particle in very small sizes in order to be able to separate the different minerals.

7
 Figure 6 A- Composite particle of wolframite, quartz, k-feldspar and albite observed through SEM after a
compression ratio of 30% considering the size fraction particle of 1mm. B- Composite particle of wolframite in
k-feldspar, quartz and albite observed through SEM after a compression ratio of 10% considering the size
fraction particle of 1mm

Comparing the two tungsten minerals present in the ore, scheelite and wolframite, it is evident that
different behaviour of different minerals will generate the need for different approaches to create an
upgrade in the process, the same applies to the creation of an empirical liberation model.
As explained before, an ore is a heterogeneous material composed of different minerals. All of these
minerals have different mechanical properties, as for example, hardness and Young’s modulus. The
consequence of this will be a difference in the behaviour during the compression in each mineral. It
is possible to observe in Figure 5A and 5B that some minerals show more fractures or presence of
micro-cracks. In general terms, the muscovite due to its crystalline structure (laminate or sheet)
showed more inner fractures and generation of small particles after the compression test, in
comparison with others minerals.
The geochemical analysis results are presented in Figure 7, which shows the content of tungsten in
the sample as a function of the compression ratio and size fraction. It is important to note that the
tungsten quantity is related to the element (%W) and not to the mineral (wolframite or scheelite),
which means that it could be in different mineral forms. The important information to note from
this test is where the tungsten is located in terms of particle size.
1000
679
W Content (PPM)

271 256 297

100 222 205

27 27
10

1
rest 0,063 0,125 0,25 0,5 1 2 4
Size particle (mm)

Figure 7 Tungsten content in different size fraction considering 30% of compression ratio

8
Evaluating the results of the geochemical analysis and combine it with the observation through
SEM, is possible to deduce that the tungsten content after 0.5 mm was continuous, regardless of
mineral which comes (SEM observation). In the observed samples, was found scheelite mineral in
the big size fraction evaluated.
Combining all the techniques, it was notorious that some associated minerals, as for example
muscovite and albite, are softer or brittle minerals (due to their crystalline structure), so less energy
is required to complete the size reduction of its particles, and therefore to achieve liberation. It was
observed that some minerals had more fractures than others (or the breakage occurred before than
in other minerals in the ore), generating smaller sizes of particles and good liberation.
These results open the possibility of an early separation of the material as a function of size. It was
observed that particles larger than 1 mm have a low concentration of tungsten, and smaller
particles are composite particles and liberated tungsten particles, this means that it could be
possible use selective comminution techniques to divide the material and to reduce the total
amount of needed energy through the reduction of total material.

CONCLUSIONS
The liberation of tungsten ore in coarse comminution was investigated. Particles of scheelite and
wolframite mineral were totally liberated after the sequences of compression breakage, which were
applied to imitate or simulate what occurs during the crushing stage in industrial devices such as a
jaw crusher or cone crusher. This indicates that from early stages in the process, it is possible to
achieve liberation.
In order to analysed the content of tungsten in the ore, was necessary a qualitative analysis, SEM
analysis. This test offer information about the minerals in the ore, as well as the associations
between them. But also was necessary a quantitative analysis, geochemical analysis, to be able to
measure the quantity of tungsten (element).
Comparing the two tungsten minerals present in the ore, scheelite and wolframite, it is evident that
different characteristics of these minerals will generate the need for different approaches to create
an upgrade in the process, the same applies to the creation of an empirical liberation model. It was
observed a simple locking structure for the scheelite, which suggests an easy liberation and a
possible pre-concentration formulation. On the other hand, wolframite showed veins-like and
network type, which indicates that its complete liberation is fairly difficult and a necessity of fine
reduction of sizes of the particles.

ACKNOWLEDGEMENTS
This work is based on research supported by the OPTIMORE, project number 642201, funded by
the European Union for the H2020 project.

REFERENCES
Amstutz, C., Geometric classification of basic intergrowth patterns of minerals. Amer. Geol. Inst.
Data Sheet 21, 1960.

9
Djordjevic, N., Recovery of copper sulphides mineral grains at coarse rock fragments size. Minerals
Engineering 64 (2014) 131–138. June 2014.
Evertsson M. Cone Crusher Performance. Chalmers University of Technology; 2000.
Fandrich, R.G., Bearmani R.A., Boland J., Lim W., 1996. Mineral Liberation by Particle Bed
Breakage. Minerals Engineering, 10(2), pp. 175-187.
La Nauze, R.D., Temos, J., 2002. Technologies for sustainable operation., CMMI congress. AusIMM,
Cairns, Australia. pp. 27–34.
Tromans, D., 2008. Mineral comminution: energy efficiency considerations. Minerals Engineering
21 (8), 613–620.

10
Geometallurgical Models, their Interaction with Plant
Models and the Importance of Final Reconciliation
Francisco Soto1, Javier Pizarro1, Romke Kuyvenhoven1 and Orlando Rojas2
1. Empírica Consultores, Chile
2. GeoEstima, Chile

ABSTRACT
The concept of geometallurgy has been extensively mentioned and used in the mining industry in
the past years. However, each mining firm has a different understanding of what geometallurgy
means, and its actual value varies significantly from site to site. Even in the case of properly
constructed geometallurgical models, it is fairly common to observe a sub-utilization of the model
in the actual operation, because of its low production forecasting capability on the short term. This
disconnect between the geometallurgical model and operations is typically caused by a lack of
tuning between the original model and the actual plant operation. Also, the absence of procedures
and other standards for reconciliation, affects the confidence and the predictive quality of the
model. The geometallurgical model should consist of three different levels: (I) adequate capture of
information, (II) data analysis and modelling, and (III) development of the block model. Once the
geometallurgical in-situ model is finalized, a process model should be developed of the plant at
industrial scale, in order to scale up the in-situ geometallurgical parameters from their base or
standard condition, to the actual operating reality of the plant. Finally, the stage of reconciliation
will close this process, assuring that any deviation between the geometallurgical model and the
actual plant operation is not due to lack of tuning, but due to operational issues such as unattended
maintenance requirements or poor process control in the operation. The methodology as described
in this paper has been applied at several large mining sites in Chile. The paper presents the
characteristics of the proposed methodology, and quantitatively describes its benefits with
examples from actual operations.

1
INTRODUCTION
It probably is fair to say that geometallurgy became widely used in greenfield projects in the Andes
region during the mining boom of the 90’s, when massive orebodies were studied and developed,
and upfront metallurgical characterization proved its use for accurate economic evaluation
including estimates of plant throughput, energy consumption, final product quality and overall
recovery. A few examples of such greenfield studies for concentrator plants in the 90’s are Minera
Los Pelambres (Antofagasta Minerals, Chile), Minera Escondida, Laguna Seca (BHP Billiton, Chile),
and Rio Blanco (then Montterrico Metals, Peru). Large size hydrometallurgical operations such as
Cerro Colorado, Quebrada Blanca and Lomas Bayas (all in Chile), also recognized early on the
importance and value of relating ore mineralogy, geological characterization and metallurgical
behavior.
The use of geometallurgy in actual operations is of more recent date, and with mixed results so far
(Spichiger, 2011). One of the challenges in actual operations is that geometallurgical
characterization of the orebody is of real use only when the mine production does not present large
deviations from the original mine plan. However, in practice, significant deviations between mine
plan and real production – at least for short time spans such as shifts, days or even weeks – seem to
be more the rule than the exception. According to the outcomes of Amira project P-1110
(Kuyvenhoven, 2013) in which several large mining companies participated (AMSA, Anglo
American, Codelco, Barrick, Collahuasi, Goldfields, BHP Billiton and Xstrata), some sites have
reported up to 100% deviation between mine plan and actual production on shift basis. Re-running
the geometallurgical model in these cases, to update the product forecast based on the new data, is
not a common practice. This paper will not analyze the causes in detail or possible solutions for
decreasing these deviations. However, it is important to keep in mind that when little credit is
given to geometallurgical models, then little effort will be made at operations to incorporate the
geometallurgical information from the block model in the decision making process.
Another relevant and even more important challenge for the use of geometallurgy at operational
level that has been poorly addressed to far, is that even if the deviation between the mine plan and
actual production is low, then still the data from the geometallurgical block model are of use only
(1) if the geometallurgical model itself is of good quality, and (2) if a proper plant model is available
that correctly scales up in-situ data for energy consumption (kWh/t), recovery and grade (%) to
actual plant values. Several services have been developed that address this issue (Wolff et al,
2012GeoMetsoTM, presented by Lynch-Watson et al, 2013; Integrated Geometallurgical Simulation –
IGS by SGS). However, the Chilean mining industry has been slow in recognizing the importance of
this aspect and in practice shows low utilization of geometallurgy at operational level.

PROBLEM STATEMENT
Orebodies nowadays are characterized by ever decreasing head grades that put profit margins
under pressure. High energy prices and low commodity prices even further decrease operating
margins. A comprehensive understanding of the orebody based on geometallurgical
characterization, is required to predict how the orebody will respond to a specific process or plant
configuration. Geometallurgical models are a necessary tool in understanding and controlling
variability of both the orebody and the actual operation, because they can contribute to a more
stable and more profitable operation. However, in order to be of value for the actual operations,
these models have to meet the following requirements:

2
 1. The “proper or adequate” samples, the “right” metallurgical testwork, and the “right”
QA&QC should be used to generate the base data for the model. These data should be
adequately analyzed to define the geometallurgical units throughout the orebody.
Finally, the block model should provide a correct 3D representation of the predicted
metallurgical behavior of the different geometallurgical units. This first issue of
“rightness” is not unique for the use of geometallurgical models at operational level, but
clearly has to be addressed.
2. A plant model is required that correctly scales up the in-situ variables to actual plant
data. The absence of a good plant model, limits the use of a geometallurgical model to a
predictive tool only. It will quickly loose its utility at operational level as the predicted in-
situ variables will be considered “incorrect and not applicable” by the operations.
3. Final reconciliation between the geometallurgical model and the plant model, is required
to fine-tune the plant model and to assure that deviations between production forecast
and actual plant response are not caused by a poor geometallurgical model or an
inaccurate plant model, but by pure operational reasons such as lack of maintenance or
poor process control.

METHODOLOGY
The objective of the geometallurgical model, correctly tuned with a plant model, is to provide
valuable process information ahead of time to employ a proactive rather than reactive attitude with
respect to plant operation. Operational variability can be anticipated and timely adjustments can be
made to either mine plan or plant operation in order to mitigate this variability, allowing for
stabilization and eventually optimization of both throughput and final product quality.

Step I: Construction of geometallurgical in-situ model

The first step is the construction of a geometallurgical model. It should contain three separate
levels: (I) capture of information (data), (II) data analysis and modelling, and (III) development of
the block model. Figure 1 shows schematically how the geometallurgical model should be built up
from step (I) that forms the basis, through step (II) that identifies the geometallurgical units to step
(III) that gives the final block model populated with metallurgical parameters from the testwork
program.

3
 Figure 1 Construction of geometallurgical model through pyramidal approach

This pyramidal approach has been presented in a previous publication (Rojas et al, 2014). It goes
beyond the scope of this current paper to analyze in further detail each of the triangles of this
pyramid, however it is important to mention that on many relevant issues there is no industry-
consensus, such as is the case for:
 The number of samples considered adequate for the size and in-situ variability of the
orebody: It might be evident that a total number of samples should be used to arrive at an
acceptable standard deviation within the range of values that are obtained from chemical
and metallurgical characterization. However, what is exactly an “acceptable standard
deviation” in this context?
 The way in which non-additive variables – which is the case for most of the metallurgical
parameters – should be handled.
 The chemical elements that should be considered, apart from the main species of
economic interest, for instance deleterious elements. By means of example, in the case of a
IOCG (iron ore copper gold) deposit: apart from the obvious assays for Fe, Cu and Au,
what are the elements that should be tested and modelled? Which trace elements should
be considered that could affect the economic value of the orebody?

Step II: Plant model to scale up in-situ data to actual plant values
The second step consists of scaling up the in-situ metallurgical parameters that form part of the
block model, to actual plant values. This is done through the development and use of a plant model
that is unique for the operation and translates the in-situ data to actual plant values. This plant
model can convert – by means of example – Bond Wi values that are used to populate the block
model, to specific energy consumption; likewise, rougher recovery and concentrate grade from the
block model can be translated into metallurgical response of the orebody to the actual flotation
process.

4
 Figure 2 Geometallurgical model and its relation to the actual plant model

The plant model at industrial scale, should be capable of predicting throughput rate, recovery,
settling velocity, to name a few, starting from the in-situ parameters (ore hardness such as BWi or
SPI, rougher recovery, concentrate grade, etc.) and in-situ testwork standards (grind size, reagent
dosage, etc.) from the geometallurgical model. The plant model at industrial scale should respond
to any changes in the actual operation, such as grind size, solids percentage, residence time,
reagents dosage, etc. Figure 2 demonstrates how the plant model relates to the original
geometallurgical model.

Step III: Reconciliation between geometallurgical model and plant model

The reconciliation between the geometallurgical model and the plant model, is required to close the
process and assure accurate prediction from the geometallurgical model to actual plant values.
Reconciliation will allow us to evaluate whether deviations between the model forecast and actual
plant data are due to inadequate forecasting by the model, or due to pure operational reasons such
as poor process control or lack of maintenance, that could affect the capacity of either the grinding
or flotation circuit.
Figure 3 shows the complete and final picture of how a geometallurgical should relate to a plant
model, and which role reconciliation plays:

5
 Figure 3 Geometallurgical model, plant model and reconciliation

CASE STUDIES

Case study 1: SAG – ball mill circuit

Summary of plant characteristics and project objective

The most relevant data of the SAG circuit that was studied in this project are summarized below:
 Plant capacity: 1700 tph per production line
 Current plant configuration: 2 production lines, each with 1 SAG mill and 2 ball mills
 Location: Chile, II region
 Objective: Development of methodology for the accurate estimate of SAG circuit
throughput

Information used for the study

Production data of a total of 991 production days were considered for review and validation of the
methodology currently used. After application of QA&QC criteria, 534 days were selected for
further data review of the first production line, and 522 days for the second production line. By
means of example, production days for which no data were available for Bond Wi or SPI, or for
which total power draw either in the SAG mill or the ball mill was less than 65%, were not
considered for construction of the model.

6
Analysis of available data
A clear correlation was found between alteration and breakage characteristics as defined by Bond
Wi and SPI tests (Figure 4). The different colors indicate the different types of alteration.

Figure 4 Cumulative frequency of SPI and Bond Wi values per alteration type

Relation between geometallurgical model and actual plant model

Figure 5 shows how the actual plant data relate to the geometallurgical data. The plant model
considers actual equipment characteristics, F80 feed size to the plant (or K80 in case of ball mill
circuit), power draw and specific energy consumption as observed in the plant. These data are then
correlated to the SEC (specific energy consumption) data from SPI or Starkey test in this case (for
the SAG model), or Bond Wi data for the ball mill model.

Figure 5 SAG and ball mill plant model and correction factors

7
Of all available data, 80% was used for the construction of the plant model (as identified in Figure
4), while the other 20% was used for model validation, according to the following (here simplified)
equation for SEC:
SECplant (kWh/t) = SECmodel (kWh/t) x correction factor

Final model forecast

The model was applied to the existing data to demonstrate its validity for the four main types of
alteration. The results of this exercise are illustrated in Figure 6 showing the comparison of the
actual operational data and model outcome for production line 1. Production line 2 had very
similar data and is not shown here. The numbers on the x-axis are grouped per type of alteration,
and the total number per alteration indicates the actual amount of production days that were
considered after the QA&QC was applied to the available data at the beginning of the study.

Figure 6 Comparison of actual operational data and model forecast, production line 1

The overall error, in absolute terms, is between 8 and 9% on hourly basis, and decreases
considerably when larger amount of data are available for model construction and validation. Note
that the data shown in figure 5 are simplified and shown as daily production data (tpd), whereas
the actual forecast was done on hourly basis.

Case study 2: Flotation Plant

Summary of plant characteristics and project objectives

The most relevant data of the flotation plant that was studied in this project are summarized below:
 Plant capacity: 2250 tph
 Location: Chile, II region
 General objective: Increase the accuracy of the short term production forecast for the
flotation plant, considering the main operating variables such as alteration, lithology,
mineralization, head grade, flotation feed size, solids %, reagent dosage and circuit
configuration
 Specific objectives:

8
 o Development of a predictive plant model for primary recovery at industrial scale
o Validation of the predictive model

Information used for the study and analysis of available data

A total of 506 geometallurgical tests, which had been performed prior to the starting date of this
project, were reviewed to select only those samples that met the previously defined QA&QC
requirements. The 506 samples represented a total of six (6) geometallurgical units. The review
resulted in the selection of a total of 347 tests, from which the following preliminary conclusions
were drawn for rate constant k (1/min) and infinite flotation recovery R inf (%), as shown in Table 1.
As can be seen, there is not one geometallurgical unit that is “outperforming” all others, and the
highest values for flotation constant k and Rinf vary from one UGM to another depending on the
actual element being studied (Cu, Ag or As).

Table 1 Preliminary conclusions for flotation kinetics

Element Highest value of Observed Decrease in Rinf(%) for

k for each 10 µm P80
Cu 1.105 UGM 3 0.3
Ag 0.851 UGM 5 0.6
As 1.098 UGM 4 0.4
It was observed that laboratory scale kinetics adjusted better to the Klimpel model, whereas the
plant data adjusted better to the García-Zuñíga model. For the plant data, the Klimpel model
tended to overestimate flotation kinetics and underestimate Cu recovery.

Plant surveys
Two plant surveys were performed as part of the scope of work of this project. Subsamples were
taken of the flotation feed for performance of laboratory scale flotation tests in order to link the
plant results with the metallurgical outcome at lab scale. The outcome of this part of the study was
the actual predictive plant model.

Validation of the model

During the validation stage of the plant model, which considered a time span of a total of 6 months,
it became clear that copper and silver recovery were overestimated by the model by approximately
2% and 4.5%, respectively. However arsenic recovery was underestimated by 9% for this same
period. A clear tendency was observed towards lower deviation at higher plant throughput. These
deviations could be due to the different analytical techniques used for As determination for the
plant and for the geometallurgical testwork. Also, in case of Ag, the actual grades observed in the
head ore and flotation products are close to the detection limits of the analytical method employed.
Figure 7 shows the estimated and actual copper recoveries for the 6 months’ validation period, and
also indicates the Cu head grade for that same period.

9
 Figure 7 Model forecast and actual plant recovery for 6-month validation period

Final model forecast

Once the model was validated, a short term forecast was made for 2017, using the model and the
available geometallurgical data. As ca be seen in Figure 8, recovery will be below 90% for the first
five months of the year (January to May). These values can be explained by the presence of copper
oxides species that are characteristic of UGM2, that accounts for up to 30% of the ore sent to the
plant in the first semester of the year 2017.
As the presence of UGM2 decreases in the feed to the plant (minor tan 10% of total throughput), the
primary recovery will increase to values over 90%. It has to be considered that the P80 is assumed
to be fixed at 212 μm (Figure 8).

Figure 8 Expected Cu recovery for 2017

10
CONCLUSIONS
A plant model is required to translate in-situ metallurgical parameters from the block model to
actual plant values. The plant model, apart from using ore characteristics, considers equipment
specification and circuit configuration. Plant surveys are typically required to calibrate the plant
model, and only once validated, it can be used for production forecasts.
Operating sites typically manage geometallurgical databases of significant size. However, when
applying strict QA&QC requirements, a significant part of those data can turn out to be unusable. It
is important that only data that are trustworthy are used for model purposes. When models are
constructed this way, it is likely to observe clear relations between intrinsic ore characteristics and
metallurgical behavior, which is typically not the case when unfiltered databases are used.
A correctly validated plant model uses geometallurgical data for short and longer term production
forecasts, and can identify required circuit modifications or specific equipment restrictions that
need to be addressed timely, in order to meet or correct projected throughput, energy consumption
and overall metallurgical behavior such as grade and recovery. Only this way, geometallurgical
characterization will contribute significantly to the accurate economic evaluation of current and
future operations.

ACKNOWLEDGEMENTS
This paper was first published in the proceedings of the XXVIII International Mineral Processing
Congress – IMPC 2016, Quebec City, Canada. ISBN: 978-1-926872-29-2.

REFERENCES
Lynch-Watson, S., Valle, R., Duffy, K., La Rosa, D., Valery, W. (2013). GeoMetsoTM: A site specific
methodology to optimize production and efficiency over the life-of-mine, Proceedings of the 10th
International Minerals Processing Conference – PROCEMIN 2013 (page 157-158), Santiago, Chile.
Kuyvenhoven, R. (2013). Internal AMIRA report, project P-1110 – Geometallurgy Roadmap Report
of Interviews in Chile
Rojas, O., Soto, F., Pizarro, J. (2014). Systematic review of geometallurgical model, Proceedings of the
2nd International Seminar on Geometallurgy – GEOMET 2014 (page 83-84), Santiago, Chile.
SGS website, 2016. http://www.sgs.com/en/mining/production-and-plant-
services/geometallurgy/igs-integrated-geometallurgical-simulator#cta-download
Spichiger, S. (2011). Geometallurgy: Reducing risk and assuring the highest value of the mining
business, Proceedings of the 2nd International Seminar on Geology for the Mining Industry – GEOMIN
2011 (page 10-11), Antofagasta, Chile.
Wolff, R., Dunham, S., Jackson, S., Coward, S., Moayer, S., Bye, A., Jackson, J. (2012). Transfer
functions to enable more realistic evaluations from driver based geometallurgical models
(page 80-81), Santiago, Chile.

11
Analysis of Autoregressive Model with Exogenous
Variables to Predict the CEE SAG Mill as a
Geometallurgical Answer of Porphyry Copper
Deposits
Nicolás Miranda1 and Antonio Candia2
1. Andina Division, Codelco, Chile
2. Universidad de Concepcion, Chile

ABSTRACT
This work aims at identifying the significant variables that affect the specific consumption of energy
of SAG mill and find a new way to make calculations with a better tendency; this is because the
actual model is not able to anticipate making operational decisions in a timely manner. For this
work, the model considers the CEE of SAG mill an input, and it needs several variables to
determine the impact of one of these. The process of research includes statistical analysis of the
operational data collected from the year 2011 through 2015 and the results of metallurgical test of
the open pit characterization conducted by the Department of Geometallurgy of Codelco Chile
Andina Division.
The mathematical autoregressive model with exogenous variables needs a pre-selection of the
variables we need to use in the model, this due to its complexity. Based on the former and with the
analysis of historical information, we’ve found eight main variables.
The throughput of the mill, percentage of ore under one inch size of feed, the Work Index, lithology
of GDCC, DIOR and mineralogy proportion of Magnetite-Hematite, Titanite-Sphene, Calcite and Y-
Anhydrite, these variables have a substantial impact on the CEE of SAG mill and they are used to
generate the model.
The model predicted the CEE of SAG mill with an error below 2%; it is recommendable to study
more variables aimed at getting a stronger model.

1
INTRODUCTION
Energy means one of the highest costs in mineral processing, and grinding equipment such as SAG,
rod, and ball mill and all those machines that use up large amounts of energy inefficiently. Due to
this companies seek optimize the energetic consumption produced by these equipments as only a
small percentage of energy consumed is actually used to grind mineral; while the rest is used to
move the mill and the grinding devices, plus some energy waste associated.
At the plant, specific consumption is defined as the relation between kilowatts per hour consumed
to grind a ton of a certain mineral. This so defined variable, CEE, is used to evaluate the energy
throughput of the plant.
Thus it seems coherent to implement operational strategies to get a specific consumption of
controlled energy facing the increase of mineral hardness and mineral treatment capability.
Una de estas estrategias es realizar un análisis del comportamiento del molino SAG tomando en
consideración distintos escenarios de trabajo a partir de la información del tipo de mineral, la data
operacional y test de molienda. Posteriormente generar modelos matemáticos que aproximen y
predigan el valor del consumo específico de energía.
One of these strategies is to make an analysis of SAG mill behavior taking into account different
work settings based on the information of the type of mineral, operational data and laboratory
grinding test, with all these information we’ll be able to generate mathematical models to predict
the C.E.E.
The main objectives of this research based on the aforementioned are:
 Statistic study and analysis of operational historical data of SAG mill, geological data
(lithology and mineralogy) of SAG mill input mineral besides grinding laboratory index.
 To relate obtained data and determine most significant variables in the process of SAG
mill and those that have a relation with specific energy consumption.
 Build and implement an autoregressive model to predict specific consumption of SAG
mill based on historical information, geological data and grinding laboratory index (work
index, SAG power index, Drop weight test, etc.)

METHODOLOGY

Study and identification of main variables in the SAG grinding

To identify all the possible variables that will be part of the mathematical model, all the input and
output SAG mill flows of information must be identified, provided they are available and have a
reliable quality.
There are two sets of data; they are identified as operational variables and geological variables.
Operational variables are data such as mill input tons, power, speed, particle size distribution, etc.
and they are shown on chart n. 1. Geological variables consist of lithology, thorough mineralogy
SAG mill input, besides the information provided by grinding tests such as Bond Test, SAG Milling
Test, and Drop Weight Test.

2
Table 1 Operational variables of SAG mill circuit

Operational variables of SAG circuit

C.E.E [kWh/t] Throughput belt A9 (SAG) [t/h]
SAG Power [kW] Throughput belt A6 [t/h]
MB1 Power [kW] Throughput belt A3 [t/h]
MB2 Power [kW] Throughput belt 52 [t/h]
Speed of rotation [rpm] Particle size distribution belt A9 [%]
Torque [Nm] Particle size distribution belt A3 [%]
Current [A] tc belt A9 [%]
Pressure [psi] tc belt A3 [%]

Operational variables are carried out a statistical study based on histograms and box diagrams to
check behavior data, distribution and frequency. This with the purpose of getting representative
data of SAG mill plant and getting rid of the atypical system data.

Choice of main variables to generate the predictive model

Representative data consists of operational variables, geological variables and metallurgical tests.
Once generated, a study must be carried out about each one of these areas in specific energy
consumption in SAG mill as this variable is the one that should be predicted. For this, it is necessary
a statistical and mathematical handling of representative data. To do this, correlation matrixes and
variance analysis ANOVA for every variable are made to finally decide which variables are to be
included in the mathematical model.

Methodology to build the predictive mathematical model

An autoregressive mathematical model with exogenous variables will be built (ARX) to predict the
specific energy consumption in SAG mill, which is described by the following equation of linear
differences:
𝑖=𝑗 𝑖=𝑗 𝑖=𝑗 𝑖=𝑗 𝑖=𝑗
(𝑗+1) (𝑖) (𝑖) (𝑖) (𝑖)
𝑦 = 𝑎0 + ∑ 𝑎𝑖 𝑦 (𝑖) + ∑ 𝑏1,𝑖 𝑥1 + ∑ 𝑏2,𝑖 𝑥2 + ⋯+ ∑ 𝑏𝐾−1,𝑖 𝑥𝐾−1 + ∑ 𝑏𝐾,𝑖 𝑥𝐾 (1)
𝑖=𝑝 𝑖=𝑞1<𝑗 𝑖=𝑞2<𝑗 𝑖=𝑞𝐾−1<𝑗 𝑖=𝑞𝐾<𝑗

Where:

𝑎0 : Threshold model (average autoregresive)

𝑎𝑖 : Parameters of endogenous variables 𝑦 (𝑖) , 𝑖 = 1, … , 𝑗
(𝑖)
𝑏𝐾,𝑖 : Parameters of exogenous variables 𝑥𝑘 , 𝑖 = 𝑞𝑘 , … , 𝑗 , 𝑘 = 1, … , 𝐾

(𝑖)
The exogenous variables (𝑥𝑘 ) represent all the variables of input of the model that are not the
output variables. 𝑦 (𝑗+1) .

3
As variables of input, those that have the greater influence on the output variable of the model
(C.E.E SAG mill), will be used. Apart from C.E.E of SAG mill the steps below should be followed:
 Create an autoregressive model (AR)
 Include the exogenous variables previously chosen
 Iteration of endogenous and exogenous variables with different order to establish the
performance of the model, measure the error and show trends.
 Training, validation and prediction of the model
 Analysis of variation of the parameters of the model
Besides, this model will be compared to others to evaluate results and highlight the use of ARX.

RESULTS AND DISCUSIONS

Variables of mathematical model

Variables to be included in the mathematical model will be determined by the correlation degree
between some of the operational and geological variables as an analysis of ANOVA variance.
Regarding the correlation matrix, variables having a correlation coefficient bigger or equal to 0.2
will be accepted. Besides this condition and in relation to the analysis of variance with a level of
significance of 0.05 (𝛼 =0,05), variables having a value p less than 5% (p-value<0,005 ) will be
accepted in the model.
The independent variable that meets both conditions, correlation coefficient bigger that 0.2 and
value p less than 5%, will be a variable that will be included in the mathematical model. Table 2 is
shown as a summary showing independent variables that meet the aforementioned conditions.

Table 2 Indpendent variables of mathematical model, correlation coefficient and p-value

Indpendent variables correlation coefficient p-value

Power 0,20 0,01
Throghput -0,84 1,51E-58
%ore -1'' belt A9 -0,27 0,02
Work Index 0,30 5,00E-03
%GDCC 0,20 0,03
%DIOR 0,20 0,04
%Magnetite-Hematite 0,21 0,01
%Titanite -0,30 0,01
%Calcite -0,24 0,04
%Gypsum-Anhidrite 0,22 0,02

Order of variables in the mathematical model

The variables that will be in the auto-regressive model are defined; the next step is determine the
order of the variables. The auto-regressive model has two types of input variables, endogenous and

4
exogenous variables. The exogenous variable is the dependent variable, and the exogenous variable
the independent.
The ARX model to predict the specific consumes of energy of SAG mil need data of the endogenous
and exogenous variables in previous periods of time. The order of a variable will be how many
periods behind will take in conderation in the model to predict the variable of output. This is show
in the Chart 1.

Yi-1 [CEE SAG]

i-1, i-2, …, i-n

X1, i-1 [Mill power] ARX model of CEE SAG Yi [CEE SAG]
X2, i-1 [Throughput]
X3, i-1 [%Ore -1’’]
X4, i-1 [Work index]
X5, i-1 [%GDCC]
X6, i-1 [%DIOR] i-1, i-2, …, i-n
X7, i-1 [%Magnetite-Hematite]
X8, i-1 [%Titanite-Sphene]
X9, i-1 [%Calcite]
X10, i-1 [%Gypsum-Anhydrite]

Figure 1 Scheme of ARX model to predict C.E.E of SAG mill

Therefore different order cases must be studied for all variables involved in the model. This is
shown in chart 2, where the average relative error for different orders of endogenous variable is
evaluated.

6
Average relative error [%]

0
1 2 3 4 5 6 7
Order of endogenous variable (C.E.E)

Figure 2 average relative errors as a function of the order of the endogenous variable C.E.E

The evolution of the error as a function of the order of all variables can be seen in chart considering
all the variables, endogenous and exogenous and the order of each one.

5
 8
7

Averge relative error [%]

6
5
4
3
2
1
0
12 13 14 15 16 17 18 19 20 21 22 23 24 25
Order of variables

Figure 3 Average relative error as a function of the order of variables

Therefore, the order of the model will be 15 considering all the variables, as the error is less than in
other cases. In addition when the order of the model increases, the error begins to stabilize in values
near 5.5%.

Predictive power of ARX model to C.E.E of SAG mill

The predictive power refers to the number of predictions that ARX model can make with the
constant parameters 𝑎𝑖 𝑎𝑛𝑑 𝑏𝑘,𝑖 that do not exceed the admissible error. This number of predictions
will be called degree of prediction. Then, a maximum error permissible of 5% like the predictive
power of the model will be considered. Chart 4:
12
11
10
Relative error[%]

9
8
7
6
5
4
3
2
1
0
0 1 2 3 4 5 6
Degree of prediction

Figure 4 Predictive power of ARX model to C.E.E of SAG mill

6
Then, only up to 3 predictions can be made with the constant parameters 𝑎𝑖 𝑎𝑛𝑑 𝑏𝑘,𝑖 before the error
exceeds 5%. This means that with the parameters of the last prediction we can predict the C.E.E of
SAG mill for the next 3 weeks with a minimum percentage of error.

Difference between models

One way to identify and state that the ARX model of C.E.E of SAG mil is a good model is
comparing the difference of the results, errors and trends to the actual model. For this, the ARX
model will compare to a model of mobile means (k=4) and to the model of CODELCO Andina to
predict the C.E.E of SAG mill, Chart 5:
8,00

7,50
C.E.E [kWh/t]

7,00 C.E.E. operacional

ARX
6,50 Media móvil
DAND
6,00

5,50
1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33

Figure 5 ARX, movil mean and CODELCO Andina model to predict the C.E.E of SAG mill

When we compare the three models, the better tendency of C.E.E of SAG mill is achieved by the
ARX model. This model is more complex, but it is capable of predicting the specific consume of
energy with less error and better trends considering operational and geological variables.

CONCLUSIONS
The most relevant variables in the C.E.E of SAG mill energy consumption are: Throughput, power,
porcentaje of ore under one inches belt A9, Work Index, %GDCC, %DIOR, %Magnetite-Hematite,
%Titannite, %Calcite and %Gypsum-Anhidrite.
It is posible to build an ARX model to predict the C.E.E of SAG mill with an average error near 2%,
and a good trend. The model achieves a prediction for the next three weeks with an error of 5%.
When compared the actual model to the one of CODELCO Division Andina, the model ARX has a
better predictive power, because the error of actual model has values near 12-15%.

7
GLOSARY
C.E.E : Specific consumption of energy
SAG : Semi Autogenous Mill
tc : Ore between two and four inches
GDCC : Cascade granodiorite
DIOR : Diorite

REFERENCES
A. Aguado Behar & M. Martinez Iranzo.(2003). Identificación y control adaptativo. Prentice Hall:
Pearson Educacion S. A.
J. Douglas Hamilton. (1994). Time Series Analysis (1st ed.). Princeton University Press.
L. Ljung. (1987). System Identification theory for the user. New Jersey: Prentice Hall.
A. Jankovic, H. Dunbar, & R. Mehta (2010). Relationships between comminution energy and product size
for a magnetite ore. The Journal of The Southem African Institute of Mining and Metallurgy.
L. G. Austin & F. Concha A (1994). Diseño y simulación de circuitos de molienda y clasificación. Chile.
D. C. Montgomery & G. C. Runger (1996). Probabilidad y estadística aplicadas a la ingeniería (4th ed.).
Mexico: Mcgraw-Hill.
John Starkey & Glenn Dobby, Application of the Minnovex Sag Power Index at Five Canadian
Plants.
D. Baeza, L. Magne & G. Titichoca (2003). Internalización en CODELCO Chile división Teniente del test
de Starkey para la determinación de consumo especifico de energía en la molienda semiautógena. Chile.
J. Starkey. Accurate, economical grinding circuit design using SPI and Bond. Lakeshore Road East,
Oakville, Ontario.
P. Amelunxen, P. Berrios & E. Rodríguez (2014). The SAG grindability index test. Minerals
Engineering International, vol 55, 42-51.
P. Berrio y P. Amelunxen (2012). Procedimiento AMINPRO SGI. Chile.
Dr. C. Goñi A. (2015). Apuntes aplicaciones de software a la ingeniería, Chile: Universidad de
Concepción.
Dr. C. Goñi A. (2010). Apuntes Control Automático de Procesos Metalúrgicos. Chile: Universidad de
Concepción.
Hogg, R. and Fuerstenau, D. W. (1972). Power Relation for Tumbling Mills. Trans. SME-AIME.

8
Process Development for the Mina Justa Project
Adam Johnston¹, Ian Ritchie2 and Daniel Yataco2
1. Transmin Metallurgical Consultants, Peru
2. Marcobre, Peru

ABSTRACT
Marcobre is developing the Mina Justa copper mine project, located in the Ica region of Peru. The
Mina Justa IOCG ore body comprises of a sulphide mass overlain by an oxide cap with a sharp
transitional zone between the sulphide and oxide masses.
The primary target metal is copper, but there are minor amounts of gold and silver. The oxide will
be treated by crushing and vat leaching, whereas the sulphide will be treated by HPGR and froth
flotation.
Vat leaching was chosen over heap leaching and other processes after detailed trade-off studies found
it to be the most attractive, primarily due to the high acid consumption of the ore. The sulphide
flotation produces a very high grade concentrate. The oxide and sulphide rocks are extremely hard
and competent. These factors and others have contributed to a unique process selection for this
resource.
The process selection process for oxide and sulphide, and an overview of the technology and
equipment selection will be presented.

1
INTRODUCTION
Marcobre is developing the Mina Justa Project which is located in the Ica region of Peru about 500
km southeast of Lima and 35 km southwest of the provincial capital of Nazca, in the province of
Nazca. The site is about 25 km northeast of the city of San Juan de Marcona.

Figure 1 Project Location Plan

Deposit Description
Mina Justa is an Iron Oxide Copper Gold (IOCG) type deposit comprised of sulfide and oxide
mineralization separated by a discontinuous zone of mixed material. Copper is the most important
commodity (~91% of project value) with subordinate silver (~8% of the value) and very little gold (1%
of the value). Mineralization is predominantly bornite–chalcocite and chalcopyrite in sulfides and
chrysocolla and atacamite in oxides. The reported mineral resource contains 180 Mt of oxide ore
(0.2%CuAS cut-off grade) and 182 Mt of sulfide ore (0.2%CuT cut-off grade) potentially mineable by
open pit methods, and 27 Mt of mineral resources within optimized stope shapes (0.55%CuT cut-off
grade) potentially mineable by underground methods.

Project History
Mina Justa is part of the Marcona Copper Property, owned by Marcobre S.A.C.. Marcobre was
incorporated in Peru in 2004, with 70% ownership by Chariot Resources Ltd and the remaining 30%
stake owned by Kores Resources (15%) and LS Nikko (15%). Chariot undertook a Definitive
Feasibility Study in 2009 but was unable to proceed with the project due to the global financial crisis.
In 2010, CST Mining Group Limited (CST) acquired Chariot Resources Limited and its 70% interest
in the Marcona Copper Property and the Mina Justa Project. CST evaluated the development of the
Mina Justa Project based on the execution strategy outlined in the DFS. In June 2012, Grupo Breca’s
subsidiary, Cumbres Andinas, acquired the 70% stake in the Mina Justa Project owned by CST
Mining Group Ltd. The remaining 30% stake ownership is unchanged. Minsur proceeded to study
different project configurations, culminating in a comprehensive project alternative study with
AMEC in 2015. All of the project alternatives for mining methods, throughput and processing
options were evaluated. In early 2016 a positive prefeasibility study was delivered by Ausenco based

2
on the preferred alternative. A feasibility study is currently underway to further develop the project
engineering with a view of putting the project into production by 2020.

PROCESS ALTERNATIVES EVALUATION

Oxides
Early testwork focused on the heap leaching technologies that have typically used for copper oxide
deposits since the introduction of solvent extraction technology in the 1980’s. It was found the that
acid would not penetrate the heap adequately, resulting in high acid consumption and low copper
recoveries, even with short stacked heaps of 1 to 2 m and high irrigation rates.
It was found that leaching performance would improve as the irrigation rates increased. High
solution flowrates through heap leaches has practical limitations due to beaching, channeling and
decreasing PLS copper tenor. The copper leaching kinetics were fast, but the acid consumption was
slow and persistent. Vat leaching with high solution flowrates and short leaching time was compared
to on-off heap leaching. Vat leach testing is not common, so the test procedures were developed and
refined as the tests progressed. Based on limited data, the 2009 feasibility study was developed using
vat leaching as it was able to achieve a better economic return than heap leaching.
Additional vat leaching variability testing programs were completed, adding confidence in the
projected recoveries and acid consumptions. Since vat leaching was considered to be an unusual and
expensive approach, a thorough alternative analysis was conducted in 2014 and 2015. Several process
alternatives were considered at the conceptual level, including:
i) copper oxide froth flotation

ii) permenant pad heap leaching

iii) on-off pad heap leaching

iv) vat leaching

v) ammonia leaching

vi) ore sorting preconcentration

vii) stirred tank leaching

Processes that were found to have fatal flaws were eliminated from consideration quickly. Viable
technologies were studied further until sufficient detail on costs, recoveries and risk had been
developed to make a comparison to the vat leaching base case.
The copper oxide mineralogy is split between chrysocola and atacamite at Mina Justa. Since
chrysocolla is a silicate mineral, it is not amenable to either oxide flotation or ammonia leaching. Both
of these technologies underwent some scoping level tests before being eliminated from further
consideration.
Successful ore sorting, using color or x-rays, requires high contrast between gangue and pay minerals
at a coarse particle size. Testing showed that the thin copper oxide veins did not have enough
concentration intensity to get sufficiently high recovery to justify the ore sorting process. After a

3
financial analysis, it was decided that the ore sorting process may be more attractive for lower grade
material at the end of the mine life, but that it would not benefit the treatment of the higher grade ore
at the start of operations. Therefore, this option was eliminated from consideration for the base case,
but may be studied again during operations.
The high acid consumption was the key factor in deciding to eliminate a permanent heap leach pad
from further consideration. Heap leaching testwork showed that acid consumption would continue
to as much as 100 kg/t over time in a permanent pad leach. Interlift liners were estimated to be more
expensive than an on-off pad operation, and so the permanent pad leaching option was eliminated.
Finally, the on-off pad and vat leaching options were found to be the most attractive. These options
were compared using more detailed testwork, engineering and economic sensitivity
analysis. Finally, it was determined that the vat leaching process provided the most attractive option
for the project, based on the lower acid consumption and higher copper recovery, despite the higher
capital cost of the vats
As the process selection process advanced, the project team also made a throughput rate optimization
analysis between 3 Mt/a to 20 Mt/a. It was found that 12 Mt/a of oxide ore treatment would be optimal
for the overall project. The oxides overlay a high grade sulfide resource and by treating the oxide ore
at a high rate would expedite its exploitation.

Sulphides

Comminution Circuit
Sulphides ore at Mina Justa has the following characteristics:
 Moderate fracture frequency (RQD) and coarse crushing parameters (e.g. CWI, UCS);
 Very high competency (A*b 20-30) as measured by JKMRC and SMC procedures;
 High Bond Ball Mill Work Index values (>20kWh/t);
 Very low Abrasion Index values; and
 Lower competency and hardness as the magnetite content increase.
A number of flowsheet options were considered for crushing and grinding of the sulphide ores; these
included conventional SAG-ball mill circuits, HPGR circuits as well as hybrid circuits such as
secondary crushing and SAG milling. Because of the high competency and, also, the hardness of the
ore, combined with the low abrasion characteristics, a circuit using a HPGR as a tertiary crusher prior
to conventional ball milling was considered as the most appropriate, both from a financial as well as
technical risk perspective. A hybrid circuit suggested some benefits but the technical risk was
considered too great for a greenfields project.

Sulphide Recovery
The Mina Justa ores generally exhibit good flotation characteristics with fast kinetics, low clay or
similar content and high recovery to a clean, marketable concentrate. Therefore studies have been
focussed on optimisation of flotation flowsheets and the use of seawater for flotation rather than
examining alternative approaches.
A recent testwork programme has confirmed that seawater can be successfully used in the processing
of Mina Justa ores.

4
Flowsheet development has recently focussed on making use of the liberation and flotation
characteristics to minimise recovery losses in the cleaning and regrind circuits. This has included the
use of a cleaner-scalper stage using a contact type cell to remove easily liberated copper prior to
regrinding and reducing the impact of high grade copper on the sizing and operation of the
downstream conventional regrind and cleaning circuit.

PROCESS DESIGN OPTIMIZATION

Oxides

Figure2 Simplified Oxide Process Flow Diagram

A significant part of the capital and operating costs for the vat leaching process is the four stage
crushing circuit to achieve 100% passing 8 mm. Limited testwork on the use of HPGR to prepare the
vat feed has shown that a higher acid consumption and lower copper recovery would be achieved,
which outway the decreases in capital and operating costs. This option is still being investigated
further as it would likely be much easier to operate than a crushing circuit.
Crushed ore is contacted with a strong sulphuric acid solution in an agglomerating drum to
neutralize the initial strong acid consuming minerals before being fed to the vats. This is done to
avoid local high pH points in the vat that might lead to copper precipitation and lower recoveries.
The vats will be arranged in a single row and solution will flow in countercurrent to the order that
the ore is loaded into the vats. Consequently, a high tenor, 8 g/L Cu PLS is anticipated from the first
vat. This will be clarified before the copper is recovered by solvent extraction. The design was
updated to include a series-parallel arrangement whereby an intermediate PLS from the vats is fed
to an additional solvent extraction stage, thus generating a raffinate with lower acid and lower copper

5
than a conventional solvent extraction circuit could achieve. This reduction in acid and copper in the
leach at the end of the leach cycle will reduce soluble copper losses with the ripios and will also
reduce acid consumption overall.
Considerable engineering effort was required to optimize the vat dimensions for loading, unloading,
structural costs of the vats and cranes and piping arrangements.
The vats will be loaded by an automated tripping conveyor that will load 0.6 m lifts up to a 7 m
charge in each vat. Two automated clamshells on different bridges will be used to remove the ripios
(residue) from each vat once it has leached for 144 h and drained.
The ripios will be sent by conveyor to a permanent ripios storage dump.
It is anticipated that about 55 kt/a of high grade copper cathodes will be produced at Mina Justa each
year.

Sulphides
The overall plant flowsheet is shown in the figure below.

Figure 3 Simplified Sulphide Process Flow Diagram

Once the decision was made to use HPGR in the tertiary crushing duty, the focus moved to
optimising the design of this circuit and in particular how to configure the equipment to ensure
maximum utilisation in a relatively small circuit. The majority of HPGR circuits have been
implemented in “large” concentrators with capacities of over 20Mt/a which entails multiple parallel
units of crushers, screens and HPGR’s, thus providing the facility to more readily undertake

6
maintenance on a single unit whilst still keeping feed to the remainder of the plant. For a 6Mt/a
concentrator such as Mina Justa, the throughput can in theory be achieved with single units at all
stages with the problem that maintenance on any of these will shutdown the entire plant unless
sufficient surge capacity or standby facilities are provided.
The design work has closely benchmarked existing operations to assess appropriate maintenance
downtime periods and this data has been used in Discrete Event Modelling to assess downtime due
to maintenance issues and how this can best be mitigated. Key decisions included provision of
sufficient surge capacity at appropriate points in the circuit, use of a jaw crusher rather than a
gyratory crusher and identification of the secondary crushers as being the most sensitive part of the
circuit requiring the most downtime; therefore, the only standby equipment provided in the circuit
is an additional secondary crusher.
The sulphide concentrator, which consists of a ball mill circuit together with downstream flotation,
tails thickening and concentrate filtration, is relatively conventional but it has been carefully designed
to take advantage of the topography to minimise pumping stages and to provide a very compact and
cost-effective layout.

Figure 4 Oxide and Sulphide Plant Layout Overview

CONCLUSIONS
Mina Justa ore is quite different to many copper deposits in the region. As a consequence, the
planned treatment routes for both oxide and sulphide ores have novel flowsheets that may serve as
useful alternatives for hard ores and high acid consuming ores in the future.

7
ACKNOWLEDGEMENTS
The authors thank Marcobre for giving permission to publish this work.

REFERENCES
Schlitt, W J and Johnston, A, The Marcobre Vat Leach System: A New Look at an Old Process,
Proceedings, Copper 2010, Vol. 5, J. Harre, ed., June 6-10, 2010, Hamburg, Germany, pp. 2039-2057.

8
CHAPTER 7

New Processing
Technologies
Current Developments in Preconcentration Using
Dense Media Separation
Mauritz Lundt1 and Ish Grewal2
1. Sepro Mineral Systems, Canada
2. Met-Solve Laboratories, Canada

ABSTRACT
Dense medium separation (DMS) is one of several preconcentration methods used for early waste
rejection from run of mine ores at relatively coarse particle sizes prior to additional milling and
beneficiation. DMS has been used extensively in the coal, diamond and iron ore industries but lab
scale and pilot test work has demonstrated successful separation and upgrading of other minerals
such as lead, zinc, copper and lithium based ores using DMS.
This paper reviews historical dense media systems, results from the pilot scale Condor DMS system
as well as associated laboratory heavy liquid separation (HLS) test work on various minerals. The
mineral systems tested included a lithium bearing ore, fluorspar and a lead/silver bearing
polymetallic sulfide ore.
Pilot plant test work on a fluorspar showed that 41.6% of the mass of the feed sample could be
rejected to the tailings stream while 95.3 % of the Calcium Fluoride (CaF2) was recoverable to the
product stream. On a spodumene sample that was tested, it was possible to reject 55 % of the mass
to tailings while in excess of 90 % of the lithium reported to the product.
In recent market conditions and with the decreasing grades in existing ores, preconcentration might
be a consideration to maximize the value of certain operations.

1
INTRODUCTION
Interest in preconcentration of run-of-mine material, after minimal attrition and prior to fine
comminution, has been increasing over recent years. The advantages of ore preconcentration
provide opportunities to not only lower operating costs but also lower capital costs by reducing the
size of the downstream beneficiation circuit. In some cases, it may be possible to produce a
marketable concentrate directly. The benefits in operating costs arise from reduced mass flow to the
grinding circuit and the downstream beneficiation process. Often, the preconcentrated minerals are
of lower hardness than the gangue minerals hence the grinding circuit also benefits from lower
energy requirements for equivalent feed mass and significantly reduced wear and media
consumption. Preconcentration can also assist in producing a more consistent feed grade to the
downstream milling circuit which could improve overall circuit performance.
Preconcentration is achieved by exploiting a variety of differences in mineral properties such as
optical characteristics, magnetic susceptibility, density, radioactivity and conductivity. This paper
focuses on preconcentration of minerals based on their density using dense media separation
(DMS).
Recently, Met-Solve Laboratories, in conjunction with Sepro Mineral Systems, has conducted a
number of pilot scale test programs utilizing the Condor centrifugal, multi-stage, dense media
separation system. The mechanical design, operational advantages and findings from pilot plant
tests and heavy liquid test results are presented herein. The use of heavy liquid for bench scale
simulation and results are also discussed.

Bench Scale Simulation – Heavy Liquid Separation

Although not economically feasible in large scale operations, bench scale heavy liquid separation
(HLS) test work is the typical preliminary step to pilot scale DMS testing. HLS testing simulates the
separation of the minerals under ideal conditions. This information can then be used to identify the
target specific gravity cut-points for pilot and full scale operation.
Past practices of HLS test work utilized heavy organic liquids such as tetrabromoethane,
diiodomethane, and bromoform (Meyer & Craig, 2010). All three solutions are known to have
toxicity issues and require careful handling and extra safety precautions during testing. For this
reason, most industries have moved away from using these solutions and are instead utilizing
newer tungstate based liquids. The development of these tungstate based solutions such as lithium
metatungstate, sodium polytungstate, and lithium heteropolytungstate have made the bench scale
testing safer and consequently easier to carry out. The solutions are made of inorganic compounds
which dissolve readily in water. They are non-toxic, non-corrosive and reported to be ecologically
safe.

Media
While heavy liquids are used at bench scale, it is not practical to use them at full scale. Thus an
artificial heavy liquid is created by suspending a fine high density powder in water. This mixture is
called the heavy medium.
In order to create a stable and efficient suspension of heavy media that can be used in DMS, the
solid particles must have specific properties. They must be dense and finely ground to be able to be
easily suspended to create a fluidized dense medium slurry. They must be resistant to corrosion

2
and degradation to prevent viscosity increases during operation and they must be easily
recoverable from a washing circuit after exiting the vessel (Wills & Napier, 2006).
Currently the most commonly used materials for creating dense medium slurries are ferrosilicon
and magnetite. Both materials can produce efficient dense suspensions that can be used in DMS and
have the benefit of being easily recoverable by magnetic separation. Ferrosilicon has a higher
specific density (6.8 g/cm3) than magnetite (4.5 g/cm3) and can therefore produce a higher range of
relative medium densities. However, ferrosilicon is more expensive and finds use primarily when
the higher separation densities are required.

Dense Media Separators

Dense media separators can be categorized into two groups (Wills & Napier, 2006): gravitational
(static) and centrifugal (dynamic). In gravitational units, the feed and medium are introduced into
the vessel and the mixture is gently agitated to maintain a fluidized bed. The less dense minerals
float to the top of the liquid level and are removed by overflow or a paddle, while sink removal
methods vary depending on the type of separating vessel. The most common gravitational units
include the Wemco Drum, Drewboy and Norwalt (Wills & Napier, 2006).
Centrifugal units utilise high speed and tangential pumping to create a vortex within the vessel.
Any mineral with a higher density than the medium will be subject to greater centrifugal forces and
be pulled to the outer edge of the vortex, while any lower density mineral will remain at the center
of the vortex. The centre vortex generates a suction that extracts the less dense particles. The
different products are removed through separate discharge lines. There are many different
centrifugal vessels used in the mineral beneficiation industry however, the two most common
vessels are the Dutch State Mines (DSM) cyclone and the Tri-Flo type multi-stage dense media
separator. The Condor is based on the Tri-Flo design but with an involute medium inlet.

THE CONDOR DENSE MEDIA SEPARATOR

The Condor Dense Media Separator is a dynamic cylindrical separator.
In the Condor DMS unit, the dense media is pumped from the lower end of the cylinder (medium
inlet 1 and 2) as shown in Figure 1. Utilizing gravitational and centrifugal forces, a involute inlet
forces the media to spiral alongside the wall, creating a high pressure vortex that transports the
dense media upwards. The medium inlets and sink headers have been placed in an involute fashion
to decrease turbulence (Burton et al., 1991). By controlling the pressure and density of the heavy
media, the cut point for the Condor unit can be adjusted continuously.
At the top of the unit, feed material is gravity fed through the hopper directly into the vortex of the
rotating medium (Wills & Napier, 2006). The floats tube at the bottom of the Condor unit serves as
the outlet for less dense (floats) material as well as an opening to create a lower pressure at the
center of the vortex. This pressure difference allows lighter material to travel down the internal
surface of the media into the core, eventually being drawn out. The centrifugal forces created by the
vortex transports the sinks (material that is heavier than the cut point) away from the centre axis
and up into the sinks header boxes.

3
 Figure 1 Internal operations of a 2-stage Condor DMS unit

Cyclones vs. Condor Separator

While the Condor and DSM cyclones are both effective at separating minerals based on density,
there are some important differences between the devices that must be recognized.
The main advantage of the Condor multi-stage separator is the ability to produce multiple sink
products and the ability to fix the density cut points of each individual stage. Ores that contain
multiple minerals with small density differences can therefore be readily separated in a single
multi-stage unit. This essentially reduces the need for multiple medium circuits to a single unit.
Alternatively, the device can be operated with both stages having matching separation densities. In
this case the two stages can be considered as a rougher/scavenger system and has shown to
provider sharper separation efficiencies.
Another significant difference is that the feed ore is introduced by gravity into the Condor
separator as a dry product. In a DSM cyclone the medium and the ore reports to a mixing box and
then the mixture is pumped into the cyclone. This means the all ore particles will have to be
pumped and can contribute to high pump wear as well as degradation of the particles. In the
Condor separator only the medium is pumped. This offers the advantage of having the option to
dry feed the ore using a conveyor. Without having to pump the ore and the media, pump life and
performance is improved and the pump size is significantly reduced.
Fluctuations in the ore characteristics (e.g. grade in the heavy products) can have a major influence
in the operation of the separator. The required mass spilt of material can change. Compared to the
cyclone which is limited by the crowding effect in the apex, the multi-stage separation in the
Condor unit provides larger sinks product capacity which helps to accommodate the fluctuations in
the ore characteristics.
The primary advantages of using a dense medium cyclone arise when only a single stage separation
is required as it allows for a simpler circuit and reduced cost (i.e. a dense medium cyclone is
generally less expensive).

4
INDUSTRIAL APPLICATIONS
Dense media separators are commonly used for preconcentration of coal and diamonds but less
commonly for other mineral systems such as base metal sulfides and oxides, spodumene, fluorspar,
tin, tungsten and rare earth minerals. The response of upgrading low grade sulphide based nickel
ores using dense media cyclones has been reported on samples from Tati Nickel (Denysschen &
Wagner, 2009) and the Nkomati Nickel Mine (Pillay et al., 2011). The increase in grades and the
potential mass reductions were also discussed briefly. Altun et al. (2013) not only reported the
upgrading of mafic ore containing nickel with recoveries up to 97% and mass rejections ranging
from 38-53% but also identified the metallurgical benefits that arise from the rejection of
magnesium bearing gangue such as talc.
Burton et al. (1991) and Ferrara et al. (1994) present numerous examples of operations and pilot
programs utilizing the Tri-Flo for preconcentration. The treatise is quite extensive as it covers the
response of sulphide materials, tin, fluorspar, barite, bauxite, chromite, lithium, potash, phosphate
and nickel laterite ores.
The bigger the density difference between the valuable mineral and the waste mineral is, the easier
the DMS separation should be. Even if there is the slightest difference (density difference <0.2)
between the valuable mineral and the gauge mineral. As long as good liberation can be achieved at
the required size fraction the separation can be achieved with DMS.
If the value of the calculation above is >2.5 then it will be an easy separation. Dense medium
separations are possible at a density differential as low as 1.2 eg in the potash industry where DMS
is used to separate sodium chloride (NaCl, density 2.16 g/cm3) from potassium chloride (KCl,
density 1.98 g/cm3). The efficiency of the separation decreases with the decreasing of the value of
the above equation.

BENCH SCALE AND PILOT CAMPAIGN RESULTS

A series of test programs were performed using a single stage Condor dense media separator in
conjunction with bench scale heavy liquid test work. The general approach of the test program as
well the types of material tested are presented:
 A lead and silver bearing sulphide mineralized rock supplied by Benton Resources from
their Cape Ray project in southwest Newfoundland. The program consisted of bench scale
HLS test work and a pilot Condor DMS test.
 A fluorspar ore sample from Turkey. The program consisted of HLS and pilot Condor DMS
tests on the samples.
 A lithium bearing spodumene ore supplied by Nemaska Lithium from their Whabouchi
project site in Quebec. The program consisted of bench scale HLS test work and pilot Condor
DMS tests.

Lead Bearing Mineral Sample – Cape Ray Deposit

A sample, supplied by Benton Resources from their Cape Ray project containing lead, copper, silver
and gold was subjected to both bench scale HLS test work and pilot scale DMS test work. The
sample was crushed to 10mm followed by fines removal at a particle size of 1.18mm. The results
from the HLS testing demonstrate that 78.4% of the lead was recovered into 7.5% of the feed mass
at a specific gravity of 2.85. Recovery increased to 87% when the specific gravity cut point was

5
reduced to 2.75. The mass yield to the sink fraction at this lower cut point is only 12.2% indicating
that more than 87% of the mass can be rejected at this coarse crush size.
The lead grade of the heavy sink fraction dropped from 16.4% to 2.41% in the individual fractions
with weighted cumulative grade of 11.2% at the lowest cut point presented in Table 1.

Table 1 Bench Scale Heavy Liquid Test Results - Cape Ray Deposit Sample

Specific Sink-Float Grade Sink-Float Fraction Cumulative Distribution

gravity Fraction Pb Cu Pb Cu Pb Cu
of fraction (Wt %) (%) (ppm) (%) (%) (%) (%)
> 2.85 7.5 16.37 13,862 78.4 56.3 78.4 56.3
2.80/2.85 1.7 3.62 6,138 4.0 5.7 82.4 62.0
2.75/2.80 3.0 2.41 4,787 4.6 7.8 87.0 69.8
> 2.75 12.2 11.16 10,552 87.0 69.8
< 2.75 87.8 0.23 638 13.0 30.2 100.0 100.0
Total 100.0 1.57 1,852 100.0 100.0

This sample was also subject to separation using the Condor DMS pilot plant at specific gravity cut
points of 2.83 and 2.93 at a top particle size of 19.0 mm. The results from the pilot test, presented in
Table 2, show a similar response to that of the heavy liquid test work. Note that the material was
passed twice through the DMS unit, at identical specific gravity cut points, to scavenge any material
that may have short circuited in the first pass.

Table 2 Condor Pilot Plant Test Results - Cape Ray Deposit Sample (-19 mm, +1.18 mm material)

2-Stage DMS Weight Grade Distribution (%)

(Rougher-Scav) (%) Pb (%) Cu (%) Au (ppm) Ag (ppm) Pb Cu Au Ag
Sinks 1 (SG = 2.93) 6.2 20.0 1.5 37.3 268.0 66.2 45.7 29.6 52.5
Sinks 2 (SG = 2.93) 0.7 9.3 1.1 32.3 168.0 3.3 3.3 2.7 3.5
Total Sinks 6.8 19.0 1.5 36.8 258.4 69.5 49.0 32.3 56.0
Floats 93.2 0.6 0.1 5.7 14.9 30.5 51.0 67.7 44.0
Head 100.0 2.0 0.3 7.8 31.6 100.0 100.0 100.0 100.0
2-Stage DMS Weight Grade (%) Distribution (%)
(Rougher-Scav) (%) Pb (%) Cu (%) Au (ppm) Ag (ppm) Pb Cu Au Ag
Sinks 1 (SG = 2.83) 8.9 15.1 1.4 35.2 234.6 72.7 53.4 38.4 64.1
Sinks 2 (SG = 2.83) 1.8 6.3 0.7 22.6 109.0 6.0 5.3 4.9 5.9
Total Sinks 10.7 13.7 1.2 33.1 213.7 78.7 58.7 43.3 70.0
Floats 89.3 0.4 0.1 5.2 11.0 21.3 41.3 56.7 30.0
Head 100.0 1.9 0.2 8.2 32.7 100.0 100.0 100.0 100.0

The results from both the HLS and DMS are plotted in Figure 2 to present the comparative response
of the material from the test work. Generally, the response of the material demonstrates that
preconcentration can be effective for the low lead levels present in this material.

6
 100
2.75
90 2.80

Cumulative Pb Distribution (%)

2.85
80
70
60
50 HLS
40 2-Stage Rougher & Scav. DMS (SG = 2.93)
30 2-Stage Rougher & Scav. DMS (SG = 2.83)
20
10
0
0,0 5,0 10,0 15,0 20,0
Cumulative Mass Recovery (Wt %)

Figure 2 Lead distribution for dense media separation and heavy liquid separation on the Cape Ray deposit
samples

Fluorspar (CaF2) Ore – Turkey

A fluorspar ore sample was supplied by from a deposit in Turkey to determine the response of the
material to HLS and dense media separation.
The results, presented in Tables 3 and 4, are presented first without fines and then with the fines
included as part of the final tailings. Dense media separation was able to recover 95.3% CaF2 while
rejecting 41.6% of the mass fed to the DMS. The overall recovery of the system, including the fines,
dense media separation achieved a recovery of 78.5% CaF2 with 17.7% of the CaF2 remaining in the
fines. Thus if the fines were rejected the overall plant recovery would be 78.5% while rejecting
55.7% of the overall mass. A high grade product of 90.6% CaF2 was achieved at the highest specific
gravity (SG=3.04) cut point.
Table 3 Condor Pilot Plant Test Results – Fluorspar Sample Without fines

Weight *CaF2 Grade CaF2 Distribution

3-stage DMS Cumulat. Cumulat. Cumulat.
(%) (%) (%) (%) (%) (%)
Sinks 1 (D50 = 3.04) 37.5 37.5 90.6 90.6 65.1 65.1
Sinks 2 (D50 = 2.93) 16.3 53.8 79.5 87.2 24.8 89.9
Sinks 3 (D50 = 2.84) 4.66 58.4 61.1 85.1 5.45 95.3
Total Sinks 58.4 85.1 95.3
Floats 41.6 5.84 4.65
Calc. Head 100.0 52.2 100.0

Table 4 Condor Pilot Plant Test Results – Fluorspar Sample Including the fines

Weight *CaF2 Grade CaF2 Distribution

3-stage DMS Cumulat. Cumulat. Cumulat.
(%) (%) (%) (%) (%) (%)
Sinks 1 (D50 = 3.04) 28.4 28.4 90.6 90.6 53.6 53.6
Sinks 2 (D50 = 2.93) 12.3 40.7 79.5 87.2 20.4 74.0
Sinks 3 (D50 = 2.84) 3.53 44.3 61.1 85.1 4.49 78.5
Total Sinks 44.3 85.1 78.5
Floats 31.5 5.84 3.83
Fines (-20 Mesh) 24.2 68.5 35.0 67.42 17.7 96.2
Total 100.0 48.0 100.0

7
Lithium Bearing (Spodumene) Ore
A lithium bearing sample was provided by Nemaska Lithium Inc. to examine whether it would be
economically viable to use DMS as a preconcentration method to upgrade the contained
spodumene. The material treated was sized on site at 9.5mm x 0.5mm.
Heavy liquid separation test work was done with specific gravities of the heavy liquid ranging
from 2.65 – 3.00. The results presented in Table 5 show that 90.6% of the lithium oxide can be
recovered in 37.2% of the mass at a specific gravity of 2.72.

Table 5 Nemaska HLS Mass Balance

Specific Sink-Float Sink-Float Cumulative

Grade
Gravity Fraction Fraction Distribution
of Fraction (Wt.%) (% Li2 O) (% Li2 O) (% Li2 O)
> 3.00 9.4 7.07 35.9 35.9
2.93/3.00 5.5 5.54 16.3 52.2
2.86/2.93 5.9 4.61 14.7 66.9
2.79/2.86 7.8 3.36 14.1 81.0
2.72/2.79 8.6 2.07 9.6 90.6
2.65/2.72 15.6 0.74 6.2 96.8
< 2.65 47.1 0.12 3.2 100.0
Total 100.0 1.86 100.0

For the pilot test work, the material was processed through the Sepro Condor separator in stages.
The test was done in two stages with a separation SG of 2.97 in the first stage followed by
processing the floats from stage one at a separation SG of 2.69.
In Table 6 the results show that 90% of the lithium oxide was recovered in 44.8% of the mass with
an average sinks product grade of 3.4% Li 2O (combined sink 1 and sink 2 products). The objective
was to generate a product grade of at least a 5.8% Li 2O in the first sink product. The final floats
fraction contained only 10% of the Li 2O in 55.2% of the mass at a relatively low grade of 0.3% Li 2O.
The 5.8% Li2O sink product can be further upgraded to 6.1% Li 2O by magnetic separation in order
to obtain the high quality concentrate. This reduced mass and higher grade product allows for a
much smaller downstream processing plant used produce high quality lithium compound such as
LiOH*H2O and Li2CO3.

Table 6 Nemaska 2-Stage DMS Mass Balance

Weight Grade Distribution

2-Stage DMS
(%) (% Li2 O) (% Li2 O)
Sinks 1 (SG = 2.97) 10.9 5.9 37.6
Sinks 2 (SG = 2.69) 33.9 2.6 52.4
Total Sinks 44.8 3.4 90.0
Floats 55.2 0.3 10.0
Head 100.0 1.7 100.0

8
 2.65 <2.65
100,0 2.72

Cumulative Li2O Distribution (%)

90,0 2.79
80,0
2.86
HLS
70,0
60,0 2.93 3-Stage Rougher & Scav. DMS
(SG = 3.00, 3.00, 2.87)
50,0 2-Stage DMS (SG = 2.97, 2.69)
3.00
40,0
3-Stage DMS (SG = 2.94, 2.86,
30,0 2.70)
20,0
10,0
0,0
0,0 10,0 20,0 30,0 40,0 50,0 60,0 70,0 80,0 90,0 100,0
Cumulative Mass Recovery (Wt %)

Figure 3 Lithium distribution for dense media separation and heavy liquid separation on Nemaska samples

Based on these results, Nemaska has determined that it can process more than two thirds of its ore
by DMS only to obtain high quality concentrate with grades exceeding 6% Li 2O at relatively low
cost. Additionally, because ~8-10% of the Nemaska concentrate is magnetic gangue, it can be
upgraded and part of sinks 2 can be split and added to the concentrate making recovery as 6%
concentrate containing ~39-40% of the Li. The middlings, containing ~50% of the Li, will have a
grade of 2.7% Li2O and can be used as an enriched feed for flotation in their downstream processing
plant.
100,0
Cumulative Li2O Distribution (%)

90,0
80,0
70,0
60,0 HLS

50,0 3-Stage Rougher & Scav. DMS (SG =

40,0 3.00, 3.00, 2.87)
2-Stage DMS (SG = 2.97, 2.69)
30,0
3-Stage DMS (SG = 2.94, 2.86, 2.70)
20,0
10,0
0,0
3,05 3,00 2,95 2,90 2,85 2,80 2,75 2,70 2,65 2,60
Specific Gravity

Figure 4 Lithium distribution for dense media separation and heavy liquid separation on Nemaska samples

CONCLUSION
Preconcentration using dense medium separation has been used extensively in the coal and
diamond industries however its use for upgrading other mineral systems has been less common.
Examples from historical work on dense medium separation for various mineral systems were
reviewed and there was a precedent found for application in base metals and industrial minerals.

9
The test work completed on a silver/lead polymetallic sulfide ore, a fluorspar ore and a spodumene
ore demonstrated the efficiency of DMS as a preconcentration process. Its versatility was also noted
through the ability to achieve targeted grade and recovery response based on the profile generated
through bench scale heavy liquid test work.

ACKNOWLEDGMENTS
The authors of this paper would like to acknowledge the contributions and support through both
expertise and presence of Gary H. K. Pearse, MSc. P.Eng, (Equapolar Research) for his
contributions. The authors would also like to acknowledge Benton Resources Inc. and Nemaska
Lithium Inc. for the samples provided for testing as well permission to publish the data.

REFERENCES
Altun, N., Weatherwax, T., Klein, B. (2013). Upgrading valuable mineralization and rejecting
magnesium silicates by pre-concentration of mafic ores. Physicochemical Problems of Mineral
Processing, 50(1) 203–215. Mugla University, Mining Engi6neering Department. Mugla, Turkey.
DOI: 2084 –4735.
Burton, M., Ferrara, G., Machiavelli, G., Porter, M., & Ruff, H. (1991). The economic impact of
modern dense medium systems. Minerals Engineering, 4(3/4) 225–243. Inpromin Ltd, Rochester and
Southampton England, University of Trieste, Trieste, Italy. DOI: 0892–6875.
Denysschen, D., & Wagner, B. (2009). Pre –concentration of low grade lateritic sulphide nickel ore.
The South African Institute of Mining and Metallurgy Base Metals Conference (pp. 291–306). Retrieved
from http://www.saimm.co.za/publications/conference–papers.
Ferrara, G., Machiavelli, G., Bevilacqua, P., & Meloy, T. P. (1994). Tri –Flo: A Multistage High –
Sharpness DMS Process With New Applications. Minerals and Metallurgical Processing, 296, 63–73.
Society for Mining Metallurgy and Exploration. DOI: 0747–9182.
Meyer, E. J., & Craig, I. K. (2010). The development of dynamic models for a dense medium
separation circuit in coal beneficiation. Minerals Engineering, 23(10) 791–805. Gauteng, South Africa:
University of Pretoria. DOI: 0892–6875.
Pillay, K., Becker, M., Chetty, D., & Thiele, H. (2011). The effect of gangue mineralogy on the
density separation of low grade nickel ore. The Southern African Institute of Mining and Metallurgy, 6th
Southern African Base Metals Conference 2011 (pp. 493–510). Retrieved from
http://www.saimm.co.za/publications/conference–papers.
Wills, B. A., & Napier-Munn, T. (2006). Mineral processing technology: An introduction to the practical
aspects of ore treatment and mineral recovery. Amsterdam: Elsevier Science & Technology.

10
Beneficiation of Industrial Minerals Using a Tribo-
Electric Belt Separator
Frank Hrach, Kyle Flynn and Paul Miranda
ST Equipment & Technology, USA

ABSTRACT
Triboelectric charging has been around for thousands of years. In ancient Greece, the triboelectric
effect was initially by observation of small fibers of clothing adhering to amber jewelry. By rubbing
the fibers, the situation became worse. Even during this age, people wondered about the effect.
Next, in the 18th century, one of our nations cofounders, Benjamin Franklin, performed several
experiments utilizing the phenomenon. Triboelectric charging is based on materials becoming
electrostatically charged based on contact or friction from other particles. Examples of triboelectric
charging include rubbing fur with glass or a comb through hair. During this process, electrons
from materials will jump from one material to another and therefore become charged due to
differences in surface electron affinity (or work function).
Recently, ST Equipment & Technology, LLC (STET) has developed a processing system based on
tribo-electrostatic separation. This dry technology provides the mineral processing industry a
means to beneficiate fine materials. In contrast to other electrostatic separation processes that are
typically limited to particles greater than 75µm in size, the triboelectric belt separator is ideally
suited for separation of very fine (<1µm) to moderately coarse (500µm) particles with very high
throughput. The high efficiency multi-stage separation through internal charging/recharging and
recycle results in far superior separations that can be achieved with a conventional single-stage
free-fall tribo-electrostatic separator. The triboelectric belt separator technology has been used to
separate a wide range of materials including mixtures of glassy aluminosilicates/carbon,
calcite/quartz, talc/magnesite, and barite/quartz. Separation results are presented and economic
comparison of using the tribo-electrostatic belt separation versus conventional flotation for
barite/quartz separation is evaluated.

1
INTRODUCTION
The triboelectric effect, better known as triboelectric charging, is caused when certain different
materials become frictionally charged by contacting each other. During this process, electrons from
one material will be transferred to another material. When the materials are then separated, excess
electrons will be left on one material while the other material will be electron deficient, therefore,
both materials will behave in a charged state based on losing or gaining electrons. The sign and
magnitude of the charge difference depends partly on the difference in electron affinity (or work
function) between the particle types. For example, rubbing your hair with a balloon will build up
negative charge on the balloon, leaving your hair positively charged. Even though the triboelectric
effect is well known, it can be unpredictable, however, scientific advancements have been
completed for separation of certain materials utilizing the triboelectric effect. Separation can then be
achieved using an externally applied electric field.
The lack of access to fresh water is becoming a major factor affecting the feasibility of mining
projects around the world. According to Hubert Fleming, former global director for Hatch Water,
“Of all the mining projects in the world that have either been stopped or slowed down over the past
year, it has been, in almost 100% of the cases, a result of water, either directly or indirectly” Blin
(2013). Dry mineral processing methods offer a solution to this looming problem. Wet separation
methods such as froth flotation require the addition of chemical reagents that must be handled
safely and disposed of in an environmentally responsible manner. Inevitably, it is not possible to
operate with 100% water recycle, requiring disposal of at least of portion of the process water, likely
containing trace amounts of chemical reagents.
Dry methods such as electrostatic separation will eliminate the need for fresh water and offer the
potential to reduce costs. Electric separation methods that utilize contact, or tribo-electric, charging
are particularity interesting because of their potential to separate a wide variety of mixtures
containing conductive, insulating, and semi-conductive particles.
ST Equipment and Technology (STET), a subsidiary of Titan America, has developed a full scale
triboelectric mineral separator. The technology can successfully separate fly ash and minerals at
approximately 40 tons per hour. For fly ash, separation of the carbon component (Ecotherm) and
the aluminosilicate based component (Proash) has been proven at a lab scale demonstration, pilot
plant demonstration, and full scale implementation of the technology with in the United States and
worldwide. Ecotherm products produced using the technology can be placed back into the coal
fired power plant or transported offsite for further caloric use, such as kiln furnace fuel. Secondly,
the Proash product is an excellent replacement for portlandite cement based on its Pozzolanic
properties. By successfully using the triboelectrostatic separators, utility plants using coal fired
technology are able to successfully use a waste product and transform it to a product of use in the
cement industry along with recovering the unused carbon from the fly ash.
Besides being utilized in the fly ash industry, this technology has been used to successfully separate
minerals from each other. For example, calcite purification along with quartz rejection has been
successfully tested using this type of technology. Secondly, removing magnesite from talc was also
successful. Lastly, increasing barite purification to the required clients level as also been completed.
Each of these accomplishments are discussed below.

2
TRIBO-ELECTROSTATIC BELT SEPARATION
In the tribo-electrostatic belt separator (Figures 1 and 2), material is fed into the thin gap 0.9 –
1.5 cm between two parallel planar electrodes. The particles are triboelectrically charged by
particle to particle interaction. For example, in the case of coal combustion fly ash, a mixture of
carbon particles and silicate based mineral particles, the positively charged carbon and the
negatively charged minerals are attracted to opposite electrodes. The particles are then swept
up by a continuous moving open-mesh belt and conveyed in opposite directions. The belt
moves the particles adjacent to each electrode toward opposite ends of the separator. The
electric field need only move the particles a tiny fraction of a centimeter to move a particle from
a left-moving to a right-moving stream. The counter current flow of the separating particles and
continual triboelectric charging by carbon-mineral collisions provides for a multistage
separation and results in excellent purity and recovery in a single-pass unit. The high belt speed
also enables very high throughputs, up to 40 tons per hour on a single separator. By controlling
various process parameters, such as belt speed, feed point, electrode gap and feed rate, the
device produces low carbon fly ash at carbon contents of 2 % ± 0.5% from feed fly ashes ranging
in carbon from 4% to over 30%.

Figure 1 Schematic of triboelectric belt separator

The separator design is relatively simple. The belt and associated rollers are the only moving parts.
The electrodes are stationary and composed of an appropriately durable material. The belt is made
of plastic based material. The separator electrode length is approximately 6 meters (20 ft.) and the
width 1.25 meters (4 ft.) for full size commercial units. The power consumption is less than 2
kilowatt-hour per ton of material processed with most of the power consumed by two motors
driving the belt.

3
 Figure 2 Detail of separation zone

The process is entirely dry, requires no additional materials, and produces no waste water or air
emissions. In the case of carbon from fly ash separations, the recovered materials consist of fly ash
reduced in carbon content to levels suitable for use as a pozzolanic admixture in concrete and a
high carbon fraction which can be burned at the electricity generating plant. Utilization of both
product streams provides a 100% solution to fly ash disposal problems.
The tribo-electrostatic belt separator is relatively compact. A machine designed to process 40 tons
per hour is approximately 9.1 meters (30 ft.) long, 1.7 meters (5.5 ft.) wide and 3.2 meters (10.5 ft.)
high. The required balance of plant consists of systems to convey dry material to and from the
separator. The compactness of the system allows for flexibility in installation designs.

Figure 3 Commercial tribo-electrostatic belt separator

4
APPLICATIONS OF TRIBO-ELECTROSTATIC BELT SEPARATION

Coal Combustion Fly Ash

The tribo-electrostatic belt separation technology was first applied industrially to the processing of
coal combustion fly ash in 1995. The technology is effective in separating carbon particles from the
incomplete combustion of coal, from the glassy aluminosilicate mineral particles in the fly ash. The
technology has been instrumental in enabling recycle of the mineral-rich fly ash as a cement
replacement in concrete production due to its pozzolanic properties. Since 1995, over 20,000,000
tonnes of fly ash has been processed by the 19 tribo-electrostatic belt separators installed in the
USA, Canada, UK, Poland, and South Korea. The industrial history of fly ash separation is listed in
Table 1.
Table 1 Industrial Application of Tribo-electrostatic belt separation of fly ash

Utility/Power Station Location Start Facility Details

Duke Energy – Roxboro Station North Carolina USA 1997 2 Separators

Talen Energy – Brandon Shores Maryland USA 1999 2 Separators
Scottish Power – Longannet Station Scotland UK 2002 1 Separator
Jacksonville Electric – St. Johns Florida USA 2003 2 Separators
South Mississippi Electric Power Mississippi 2005 1 Separator
New Brunswick Power – Belledune New Brunswick CA 2005 1 Separator
RWE Power – Didcot Station England UK 2005 1 Separator
Talen Energy – Brunner Island Pennsylvania USA 2006 2 Separators
Tampa Bay Electric – Big Bend Florida USA 2008 3 Separators
RWE Power - Aberthaw Station Wales UK 2008 1 Separator
EDF Energy- West Burton Station England UK 2008 1 Separator
ZGP (Lafarge Cement) Poland 2010 1 Separator
Korea Southeast Power South Korea 2014 1 Separator
PGNiG Termika – Sierkirki Poland 2016 1 Separator

Mineral Applications-Calcium Carbonate

When applying the technology to industrial minerals, success was achieved in the removal of
quartz and other acid insoluble contaminants from finely ground calcium carbonate. Acid
insoluble (AI) testing is a standard method of measuring the amount of undesirable contaminants
in calcium carbonate. Silicates, such as quartz, mica, and talc, tend to tribo-charge strongly negative
relative to carbonates. Because of this technology, successful separations have been achieved on
over 11 sources of calcium carbonate tested at both the pilot-scale and demonstration plant.
Improvement in product brightness is also achieved for many sources of calcium carbonate as the
tribo-electrostatic belt separation technology is also effective in removing trace amounts of dark
contaminates such as graphite and metal sulfides. Monthly average results for a pilot-scale
continuously operating separator processing calcium carbonate is shown in Table 2. This

5
demonstration plant used a continuously operating pilot-scale separator with average feed rate 10
tonne per hour.

Table 2 Demonstration plant separation results for calcium carbonate

Month Avg. Avg. Product Calcium AI rejection

Feed Product Mass carbonate to by
grade grade Yield recovery product
(%AI) (%AI) (wt.%) (%) (%)
Feb 3.3% 0.6% 86% 89% 84%
March 3.7% 0.6% 89% 92% 87%
April 4.1% 0.6% 89% 92% 88%
May 4.0% 0.7% 89% 92% 84%
June 4.7% 0.6% 89% 93% 89%

Mineral Applications-Talc
The tribo-electrostatic belt separation technology is also effective in removal of magnesite, quartz,
and other dark colored contaminants from finely milled talc. Batch pilot-scale single-pass
separation test results for a representative talc ore is shown in Figure 4. Talc content is estimated by
loss-on-ignition (LOI) measurements assuming a simple mixture of talc and magnesite. The talc
feed was first milled to 95% passing 325 mesh (45 microns) then processed in a pilot-scale tribo-
electrostatic belt separator in batch mode. Feed moisture level (as measured by relatively humidity)
was adjusted by drying and/or by the addition of water to the feed batch prior to tribo-electrostatic
separation testing. For this talc ore, a talc concentrate with 88% grade was produced with 80%
recovery with this technology in a single pass.

100%

90%

80%

70%
Talc Recovery (%)

60%

50%

40%

30%

20%
57% talc in feed
10%

0%
50 60 70 80 90 100
Talc Product Grade (%)

Figure 4 Pilot-scale separation results for a representative talc sample

6
Mineral Applications-Barite
The tribo-electrostatic belt separation technology is also effective in removing quartz and other low
specific gravity (SG) components from barite used as a weighting agent in drilling applications. The
American Petroleum Institute (API) specification requires that the barite be milled to a particle size
with less than 3% retained on a 200 mesh (0.075 mm) sieve. Low grade barite from several sources
milled to this specification has been successfully tested to produce a barite product that meets the
API specification 13A-7 of 4.2 SG minimum in a single-pass (rougher separation only) for barites
with a feed specific gravity of 3.78 or higher.
Batch pilot-scale single-pass separation test results for upgrading a representative low-grade barite
ore sample is shown in Figure 5. The results are also presented showing the increase in barite
product SG in Figure 6. For this low grade barite feed, with an average 3.75 SG, the tribo-
electrostatic belt separator was able to produce API 13A-20 grade (4.1 SG barite) in a single pass,
and produce API 13A-7 grade (4.2 SG barite) in a second pass (rougher and cleaning separation
only, without scavenging). Scavenging of the tailings would significantly improve the overall barite
recovery.

Figure 5 Separation Results for representative low-grade barite

7
 Figure 6 Specific gravity improvement for low-grade barite with cleaning

In addition to these examples, the tribo-electrostatic belt separation technology has been
demonstrated to effectively beneficiate potash (sylvite/halite/kieserite), mineral sands
(ilmenite/rutile/zircon/quartz), phosphate (apatite/dolomite/quartz), and graphite/quartz. Since the
separator can process materials with particle sizes from about 0.5 mm to less than 0.001 mm, and is
effective for both insulating and conductive materials, the technology greatly extends the range of
applicable materials over conventional electrostatic separators. Since the tribo-electrostatic process
is entirely dry, its use will eliminate the need for final product drying and liquid waste handling
from flotation processes.

COST OF TRIBO-ELECTROSTATIC BELT SEPARATION

Comparison to Conventional Flotation for Barite

A comparative cost study was commissioned by STET and conducted by Soutex Inc., a Quebec
Canada based engineering company with extensive experience in both wet flotation and
electrostatic separation process evaluation and design. The study compared the capital and
operating costs of tribo-electrostatic belt separation process to conventional froth flotation for the
beneficiation of a low-grade barite ore. Both technologies upgrade the barite by removal of low
density solids, mainly quartz, to produce an API drilling grade barite with greater than 4.2 SG.
Flotation results were based on pilot plant studies conducted by the Indian National Metallurgical
Laboratory, NML (2004). Tribo-electrostatic belt separation results were based on pilot plant studies
using similar feed ores. The comparative economic study included flowsheet development, material
and energy balances, major equipment sizing and quotation for both flotation and tribo-electrostatic
belt separation processes. The basis for both flowsheets is the same, processing 200,000 t/y of barite
feed with SG 3.78 to produce 148,000 t/y of drilling grade barite product with SG 4.21 g/ml. The
flotation process estimate did not include any costs for process water, or water treatment.

8
Flowsheets were generated by Soutex for the barite flotation process (Figure 7), and tribo-
electrostatic belt separation process (Figure 8).

Figure 7 Barite flotation process flowsheet

Figure 8 Barite tribo-electrostatic belt separation process flowsheet

9
Theses flowsheets do not include a raw ore crushing system, which is common to both
technologies. Feed grinding for the flotation case is accomplished using a wet pulp ball mill with
cyclone classifier. Feed grinding for the tribo-electrostatic belt separation case is accomplished
using a dry, vertical roller mill with integral dynamic classifier.
The tribo-electrostatic belt separation flowsheet is simpler than flotation. Triboelectostatic belt
separation is achieved in a single stage without the addition of any chemical reagents, compared to
three-stage flotation with oleic acid used as a collector for barite and sodium silicate as a depressant
for the silica gangue. A flocculant is also added as a reagent for thickening in the barite flotation
case. No dewatering and drying equipment is required for tribo-electrostatic belt separation,
compared to thickeners, filter presses, and rotary dryers required for the barite flotation process.

Capital and Operating Costs

A detailed capital and operating cost estimate was performed by Soutex for both technologies using
equipment quotations and the factored cost method. The operating costs were estimated to include
operating labor, maintenance, energy (electrical and fuel), and consumables (e.g., chemical reagent
costs for flotation). The input costs were based on typical values for a hypothetical plant located
near Battle Mountain, Nevada USA. The total cost of ownership over ten years was calculated from
the capital and operating cost by assuming an 8% discount rate. The results of cost comparison are
present as relative percentages in Table 3.

Table 3 Cost Comparison for Barite Processing

Wet Beneficiation Dry Beneficiation Technology

Froth Flotation Tribo-electrostatic belt separation
Purchased Major Equipment 100% 94.5%
Total CAPEX 100% 63.2%
Annual OPEX 100% 75.8%
Unitary OPEX ($/ton conc.) 100% 75.8%
Total Cost of Ownership 100% 70.0%

The total purchase cost of capital equipment for the tribo-electrostatic belt separation process is
slightly less than for flotation. However, when the total capital expenditure is calculated to include
equipment installation, piping and electrical costs, and process building costs, the difference is
large. The total capital cost for the tribo-electrostatic belt separation process is 63.2% of the cost of
the flotation process. The significantly lower cost for the dry process results from the simpler
flowsheet. The operating costs for the tribo-electrostatic belt separation process is 75.5% of the
flotation process due to mainly lower operating staff requirements and lower energy consumption.
The total cost of ownership of the tribo-electrostatic belt separation process is significantly less than
flotation. The study author, Soutex Inc., concluded that the tribo-electrostatic belt separation
process offers obvious advantages in CAPEX, OPEX, and operational simplicity.

10
CONCLUSION
The tribo-electrostatic belt separator provides the mineral processing industry a means to
beneficiate fine materials with an entirely dry technology. The environmentally friendly process can
eliminate wet processing and required drying of the final material. The process requires little, if
any, pre-treatment of the material other than grinding and operates at high capacity – up to 40
tonnes per hour by a compact machine. Energy consumption is low, less than 2 kWh/tonne of
material processed. Since the only potential emission of the process is dust, permitting is relatively
easy.
A cost study comparing the tribo-electrostatic belt separation process to conventional froth flotation
for barite was completed by Soutex Inc. The study shows that the total capital cost for the dry tribo-
electrostatic belt separation process is 63.2% of the flotation process. The total operating cost for
tribo-electrostatic belt separation is 75.8% of operating cost for flotation. The study’s author
concludes that the dry, tribo-electrostatic belt separation process offers obvious advantages in
CAPEX, OPEX, and operational simplicity

REFERENCES
Blin, P & Dion-Ortega, A (2013) High and Dry, CIM Magazine, vol. 8, no. 4, pp. 48-51.
Manouchehri, H, Hanumantha Roa, K, & Forssberg, K (2000), Review of Electrical Separation Methods,
Part 1: Fundamental aspects, Minerals & Metallurgical Processing, vol. 17, no. 1 pp 23 – 36.
Manouchehri, H, Hanumantha Roa, K, & Forssberg, K (2000), Review of Electrical Separation Methods,
Part 2: Practical Considerations, Minerals & Metallurgical Processing, vol. 17, no. 1 pp 139-166.
Ralston O. (1961) Electrostatic Separation of Mixed Granular Solids, Elsevier Publishing Company, out
of print.
NML (2004) Beneficiation of low grade barite (pilot plant results), Final Report, National Metallurgical
Laboratory, Jamshedpur India, 831 007

11
Multiple Concentrator Survey of Real-Time
Individual Cyclone Particle Size Tracking (PST)
Christian O’Keefe, Robert Maron, Dylan Cirulis and Juan Francisco Medina
CiDRA Minerals Processing, USA

ABSTRACT
Proper grind control has been limited by the inability to track the particle sizes in the overflow in
real-time and on an individual cyclone basis. Instruments based on older technology, such as those
using ultrasonics, laser diffraction, or mechanical calipers, do not perform real-time measurements,
having typical update rates of three to nine minutes. In addition, they require samplers, typically
have low uptime rates, and cannot cost effectively monitor individual cyclones. These limitations
have been overcome with the advent of a new technology, acoustic impact based particle size
tracking (PST). This new measurement has enabled higher level control of the particle size through
control of the grind circuit and individual cyclones.
The first commercial implementation of acoustic impact based particle size tracking has resulted in
a system that monitors the overflow of every cyclone in a cluster with real-time four-second
updates. This paper details the calibration and resulting accuracy of this technology under
different cyclone overflow pipe orientations, different ore types, particularly with varying specific
gravities, and different size fractions. The results from seven different concentrators will be
presented. The dry solids specific gravities ranged from 2.7 to 4.1. The overflow pipe orientations
ranged from vertical to horizontal. The size fractions varied from a P80 of 75 to 200 microns. The
total measurement accuracies obtained were within 5 percentage points for one standard deviation.
These accuracies were measured under plant operating conditions as referenced to samples
obtained from the overflow pipes and sieved using both wet and dry sieving procedures.

1
INTRODUCTION
The value of a real-time particle size measurement is that it allows for both a determination of the
impact of changing operating conditions on particle size and more importantly, the effective
implementation of a control system that has demonstrated increased economic value to a
concentrator (Cirulis et al., 2015). The value of such a measurement on each cyclone overflow pipe
is that it allows for the monitoring of the performance of each cyclone, which can vary substantially
between cyclones in the same cluster. The value of a non-sampling particle size measurement
system is that it requires minimal maintenance and has a very high availability rate or uptime.

PRINCIPLE OF OPERATION
Acoustic impact based particle size tracking is a unique method for measuring and tracking particle
sizes in cyclone overflow lines. The implementation of this technology is centered upon a probe
that is inserted into the slurry stream via a two inch (50 mm) hole in the overflow pipe as seen in
Figure 1. Particles within the slurry stream impact the surface of the probe generating traveling
stress waves within the probe. A sensor converts these traveling stress waves into an electrical
signal and proprietary signal processing techniques translate these signals into a particle size
measurement that is output every four seconds. The sensor effectively samples a few percent or
more of the slurry stream, which is orders of magnitude more than is sampled by other traditional
technologies that utilize online samplers and that do not sample individual hydrocyclones. Also,
because of the location of the sensor downstream of the hydrocyclone and the presence of an air
core at that point, the sensor produces no change in the back pressure seen by the hydrocyclone.

Figure 1 CYCLONEtrac PST - Installed on overflow pipe (left); illustration of principle of operation (right)

On-line cyclone overflow sizing methods, whether laser diffraction, ultrasonic, caliber or impact
based, require calibration by correlating their signals to reference particles or to samples that are
correspondingly analyzed with laboratory screens (Outotec, 2009). The impact based
CYCLONEtrac PST also requires calibration to compensate for influences from cyclone type, ore

2
type, and sensor installation location. To ensure a good composite calibration that can be applied
across all the cyclones in a cluster, calibration samples must be taken from the overflow of each
cyclone in a cluster. Once such calibration takes place, it does not have to be performed again even
if the probe is replaced. In addition, samples must be taken beyond the expected operating range of
the cyclones to ensure accurate measurements when the cyclone is operating outside its normal
operating range including but not limited to roping events, startups, shutdowns, and grindouts.
This avoids the measurement uncertainty that occurs when calibration models are used to
extrapolate measurements beyond their calibrated range.

TYPICAL TESTING METHODOLOGY

The tests were grouped into two categories: full cyclone cluster/plant tests or single cyclone tests.
In every case, 2 inch (50 mm) holes were drilled into the cyclone overflow pipes and a PST probe
assembly was installed. For full cyclone cluster/plant tests, the PST probes were wired to a central
processing computer; whereas, for the single cyclone tests the PST probe was connected to a
standalone AC/battery powered data logger. While recording data, samples were taken from the
cyclone overflow typically via a variety of samplers. These samples were either dry stack sieved,
single screen wet sieved, or multiple screen (stack) wet sieved using a variety of different
procedures depending on the the metallurgical laboratory protocol or if it was done by CiDRA
personnel. The time at which the sample was taken was recorded, and the sample results were
used in conjunction with the recorded data to generate calibration coefficients. During a full cyclone
cluster or plant sampling campaign, the sample results were split between calibration samples and
validation samples.

SITE LOCATION #1 (USA): COPPER, MOLYBDENUM IN PORPHYRY ORE

This concentrator processes 168 kt/d of copper porphyry ore with a typical copper head grade of 0.2
to 0.5%. In addition the ore body contains molybdenum with a grade from 0 to 0.1%, as well as
gold and silver. After crushing, the ore is sent through four parallel primary grind lines each
consisting of a SAG mill and pebble crushing circuit followed by two parallel ball mills each
operating in reverse closed circuit with a dedicated cluster of 10 to 14 cyclones. The overflow is
sent for flotation to either Wemco mechanically agitated cells or Outotec forced air cells. Particle
size measurements are performed through 12 hour composite shift samples.
PST systems were installed on each cyclone cluster for a total of 92 PST probes. Over a period of
several months, over 1000 samples were taken from the cyclone overflows and processed through
either stack dry sieves or wet sieved through 100 mesh (150 micron) screens. These samples were
used to either generate the calibration coefficients or to validate the calibrated PST measurements.
The parameter used in the control system is %+100 mesh. The operating range for the %+100 mesh
typically ranges from 20 to 35% with the desired size set to maximize economic value of the
production. The control system uses the PST measurement to set and hold the desired particle size.
During the course of collecting the calibration and validation samples, the particle sizes ranged
from 6.7 to 34.6 %+100 mesh with some samples collected at much higher ranges during roping
events. When the calibrated PST measurements over the non-roping range were compared to the
samples gathered for calibration and validation, the coefficient of determination or R 2 of 0.803 was
obtained. In terms of the difference between the PST measurements and the sample results, a
standard deviation of 4.45 percentage points or 4.45 %+100 mesh was calculated. Cyclone to

3
cyclone and cluster to cluster variations in the calibration are captured in this standard deviation
since the same calibration coefficients were applied to all cyclones in the plant. A comparison of the
calibrated PST measurements to the sample results, along with lines indicating a deviation of two
standard deviations from the ideal correlation are shown in Figure 2.

- - 2=8.9%+100mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 2 CYCLONEtrac PST 150 micron measurements compared to the sieved samples for a
copper/molydenum concentrator processing copper porphyry ore

SITE LOCATION #2 (MEXICO): COPPER, GOLD AND SILVER IN PORPHYRY ORE

This concentrator processes 8,000 t/d of ore containing 0.9 – 1.1% copper, 48-53 g/t of silver, and 1.5
– 1.9 g/t of gold. It has six open circuit ball mills in its primary grinding stage, followed by cyclone
classification with one 15 inch cyclone per ball mill such that the overflow reports to flotation and
the underflow reports to secondary grinding. In secondary grinding the slurry passes through two
ball mills that are in closed circuit with two cyclone clusters of 20 inch cyclones. Particle size
measurements are currently performed via manual sampling and sieving.
A PST probe was installed on the cyclone associated with ball mill #5 in the primary grinding stage.
During the course of 5 days in a one week period, 42 samples were taken and dry stack sieved with
mesh sizes of 60, 100, 150 and 200 mesh. Calibration coefficients were generated for the percentage
by weight of solids greater than 200 mesh (%+200 mesh). This was the size of interest with a
nominal operating level of 40% +200 mesh with an operating range of 20-60% +200 mesh. By
varying the solids feed rate and water addition, a variety of particle sizes in the overflow were
created ranging from 11.3 to 40.9 %+200 mesh for non-roping conditions and 68.4 and 68.8 %+200
mesh during roping events. When the calibrated PST measurements were compared to the
sampled and sieved results, an R2 of 0.964 was obtained. In terms of the difference between the
PST measurements and the sample results, a standard deviation of 2.19 percentage points or
2.19%+200 mesh was calculated. A comparison of the calibrated PST measurements to the sample

4
results, along with lines indicating a deviation of two standard deviations from the ideal correlation
are shown in Figure 3.

- - 2=4.37 %+200mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 3 CYCLONEtrac PST 74 micron measurements compared to the sieved samples for a copper, silver and
gold plant dominated by silica and feldspar in the ore

SITE LOCATION #3 (CHILE): COPPER IN PORPHYRY ORE

This concentrator processes approximately 100,000 t/d of ore containing typical porphyry deposit
minerology consisting of chalcocite, pyrite, chalcopyrite, bornite and covelite with a copper grade
of over 0.5%. The remainder of the ore consists primarily of silica. The grind circuit consists of a
SAG mill followed by multiple ball mills operating in reverse closed circuit with a cluster of 33 inch
cyclones with more than 10 cyclones per ball mill. The overflow from the cyclones report to
flotation. The control parameter of interest is the percentage of material retained by 100 mesh (150
micron) screens. The typical range of operation is 15 to 20 %+100 mesh (%+150 micron).
An entire cluster of 10+ hydrocyclones was instrumented with a PST system, with individual PST
probes on each cyclone overflow. Due to a very short vertical downflow section, the PST probes
were installed on a long horizontal section of the overflow pipe. Over a period of several weeks,
268 samples were taken and wet sieved with CiDRA’s rapid wet sieving equipment and procedure
at 100 mesh (150 micron). Between natural variations in the process, deliberate manipulations of
the pressure via cyclone openings and closings, as well as the particle size differences between
cyclones, a range of particle sizes was captured. The range extended from 6.8 to 40.5% +100 mesh.
Calibration coefficients were generated and applied to the PST system in order to compare the PST
measurements with the sampled results. This comparison yielded an R2 of 0.86 and a standard
deviation of 2.7 percentage points for the difference between the PST measurements and the sample
results. A comparison of the calibrated PST measurements to the sample results, along with lines
indicating a deviation of two standard deviations from the ideal correlation are shown in Figure 4.

5
 - - 2=5.4%+100mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 4 CYCLONEtrac PST 150 micron measurements compared to the sieved samples for a large copper
concentrator

SITE LOCATION #4 (BRAZIL): IRON IN ITABIRITE IRON ORE

This concentrator in Brazil is expected to produce over 15 Mt of iron in 2016. The grind circuit
consists of several stages of crushing, followed by HPGR grinding and then two parallel ball mills.
Each ball mill operates in direct closed circuit with a cluster of 33 inch cyclones. The overflow of
the cyclones passes through desliming cyclones to reverse flotation. The control parameter of
interest is the percentage of material passing 150 mesh (106 micron). Currently, the particle size is
measured via manual sampling and sieving.
A single cyclone was instrumented with a PST probe, and over a single day 15 samples were taken
over a range of five operating conditions. Operating conditions were varied by opening and closing
cyclones to vary the pressure and by changing the water addition to the pump box feeding the
cyclone cluster pump. The samples were wet stack sieved by the metallurgical laboratory with
mesh sizes of 325, 200, 150 and 100 mesh. Calibration coefficients were generated for the percentage
of solids passed by the 150 mesh (106 micron) screen. For the purposes of this paper, the results
were converted from the percent of solids passed to the percent of solids retained. When the
calibrated PST measurements were compared to the sampled and sieved results, the R2 value of
0.968 was obtained. A comparison of the difference between the CYCLONEtrac PST results and the
sieved samples resulted in a standard deviation of 0.5 %+150 mesh as shown in Figure 5.

6
 Figure 5 CYCLONEtrac PST 74 micron measurements compared to the sieved samples for an iron ore
concentrator

SITE LOCATION #5 (USA): COPPER IN PORPHYRY ORE

This mill processes over 50,000 t/d of copper ore containing less than 0.4% copper and less than
0.1% molybdenum. The comminution circuit consists of three stages of crushing followed by
parallel lines of ball mills each operating in direct closed circuit with a dedicated cyclone cluster of
33 inch inclined cyclones. The particle size parameter of interest is +65 mesh (%+210 micron) with a
typical operating range of 15 to 14 %+65 mesh.
The horizontal section of the overflow from a single cyclone was instrumented with a PST probe,
and over three days 29 samples were taken over a range of operating conditions. Operating
conditions were varied by changing the solids feed rate to the ball mill and changing the water
addition to the cyclone cluster pump box. The samples were dry stack sieved by the metallurgical
laboratory with mesh sizes of 48, 65, 100, 150 and 200. Calibration coefficients were generated for
the percentage of solids passed by the 65 mesh (210 micron) screen. For the purposes of this paper,
the results were converted from the percent of solids passed to the percent of solids retained. When
the calibrated PST measurements were compared to the sampled and sieved results, the R 2 value of
0.93 was obtained. A comparison of the difference between the CYCLONEtrac PST results and the
sieved samples resulted in a standard deviation of 1.97 %+65 mesh as shown in Figure 6.

7
 - - 2=3.9%+100mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 6 CYCLONEtrac PST 210 micron measurements compared to the sieved samples for a
coppery/molybdenum plant

SITE LOCATION #6 (BRAZIL): PHOSPHATE IN ORE WITH SILICA, MICA,

MAGNETITE
This is a phosphate rock plant with a complex ore mineralogy that includes phosphate, magnetite,
mica, and silicates. Comminution contains a reverse closed circuit ball mill with cyclone
classification. The cyclone cluster associated with each ball mill has 26 inch cyclones. The particle
size of interest is 100 mesh.
A single CYCLONEtrac PST probe was installed and over the course of four days, 39 samples were
gathered and wet stack sieved with screens of 48, 65, 100, 150, and 200 mesh. Calibration
coefficients were generated for the %+100 mesh (%+150 microns). When the calibrated PST
measurements were compared to the sampled and sieved results, the R2 value of 0.89 was obtained.
A comparison of the difference between the CYCLONEtrac PST results and the sieved samples
resulted in a standard deviation of 1.73 %+100 mesh as shown in Figure 7.

8
 - - 2=3.5%+100mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 7 Comparison of CYCLONEtrac PST results and sieved samples

SITE LOCATION #7 (BRAZIL): GOLD IN ORE WITH VARYING MINERALOGY

INCLUDING QUARTZ
This is a gold leach concentrator processing over 1 Mt/y of ore containing 3 g/t of gold along with a
complex minerology containing quartz, sulphides, and carbonaceous metapelite. The grind circuit
consists of crushers followed by two parallel ball mills, each operating in closed circuit with a
cyclone cluster. The overflow from the cyclones feeds a leach circuit. The particle size of interest is
150 mesh (106 microns).
A single CYCLONEtrac PST probe was installed on the one operating cyclone in a two cyclone
cluster, and over the course of three days, 54 samples were gathered and wet stack sieved with
screens of 100, 150, and 200 mesh. Operating conditions were changed by varying solids feed rate
and water addition to the cyclone cluster pump box. Calibration coefficients were generated for the
%+150 mesh (%+106 microns) which ranged from 6 to 30.7 %+150 mesh. When the calibrated PST
measurements were compared to the sampled and sieved results, the R2 value of 0.87 was obtained.
A comparison of the difference between the CYCLONEtrac PST results and the sieved samples
resulted in a standard deviation of 2.36 %+100 mesh as shown in Figure 8.

9
 - - 2=4.7%+150mesh
 ideal correlation

Conservative  2% pts
sampling/sieving error
bars shown

Figure 8 CYCLONEtrac PST 106 micron measurements compared to the sieved samples for a gold
concentrator

RESULTS AND DISCUSSION

A summary of the results from applying the acoustic impacted based particle sizing technology at
seven sites is given in Table 1. A literature search for similar in-plant testing of other particle size
measurement technologies was unsuccessful; therefore, a critical review or comparison was not
possible.

10
 Table 1 Acoustic impact based PST particle size measurement results including differences from
sampled and sieved results. Note: sampling and sieving uncertainties were not determined but
can be a significant source of the differences between the PST and the sample measurements.

Under plant operating

conditions. Includes
sampling and sieving
errors which can be
significant

Site Type and Ore Measurement of Test particle Cyclone Overflow PST vs (PST-
interest size range orientation pipe Sampled R 2 Sampled)
(excluding orientation Standard
roping events) deviation

Copper/Molybden %+100# (% +150 6.7 to 34.6 Vertical Vertical 0.803 4.45

um in Porphyry microns) %+100#

Copper, Gold, %+200# (%+74 11.3 to 40.9 Vertical Vertical 0.964 2.19
Silver in porphyry microns) %+200#

Copper in %100# (%+150 6.8 to 40.5 Inclined Horizontal 0.860 2.70

porphyry microns) %+100#

Iron ore with silica %+150# (%+106 17.7to 26.2 Vertical Vertical 0.968 0.50
microns) %+150#

Copper in %+100# (%+150 26.3 to 49.7 Inclined Horizontal 0.930 1.97

porphyry microns) %+100#

Phosphate with %+100# (%+150 5.2 to 27.4 Slightly off Vertical 0.890 1.73
silica, mica and microns) %+100# vertical
magnetite

Gold in quartz, %+150# (%+106 6.0 to 30.7 Vertical Vertical 0.870 2.36
metasedimentary, microns) %+150#
metavolcaniclastic
rocks, and
metabasalts

11
CONCLUSION
The acoustic impact based particle sizing technology implemented in the CYCLONEtrac PST has
demonstrated excellent performance when compared to the sampled and sieved data for a variety
of ore types, cyclone types and orientations, overflow pipe configurations, and particle sizes. When
properly installed and calibrated, the output of the instrument is driven by particle size and the
impact from other influences, such as from ore density, could not be detected due to the level of
uncertainties typically found in sampling and sieving. Differences between the PST measurements
and the sieved samples are quite low and are due to both measurement noise and sampling/sieving
uncertainty. These low differences between measured and reference data indicate good
measurement accuracy as evidenced by high R2 values of 0.8 and above.

ACKNOWLEDGEMENTS
This work would not have been possible without the support of the mine sites including their
maintenance staff, engineers, management and metallurgical laboratories. At CiDRA Daniel
Charles and David Winkowski were instrumental in processing data while the tests were
underway. Susan McCullough contributed to the editing of this paper.

REFERENCES
Cirulis, D; Dunford, S; Snyder, J; Bartsch, E; Rothman, P; Maron, R; Newton, D; O’Keefe, C;
Mercuri, J (2015) ‘Process Optimization Using Real-Time Tracking of Coarse Material in Individual
Cyclone Overflow Streams,’ MetPlant 2015 7–8 September 2015, Perth, Western Australia, The
Australasian Institute of Mining and Metallurgy, pp 93–109
Outotec Oy (2009), Outotec Literature PSI 300 EN_150109.pdf

12
Smart Wireless Solutions - Process Control and
Variables Monitoring in a Leach Pad Operation
Hugo Barrientos
Emerson Process Management, Chile

ABSTRACT
The growing need to optimize the mining process creates a need for users to have a simple, reliable,
secure and cost-effective method to monitor the values of the variables measured in the field.
The solutions described in this abstract and presentation will demonstrate how the implementation
of Smart Wireless technology helps to optimize the heap leaching and electro winning processes,
increasing process efficiency and uptime.
An effective monitoring of the variables involved in the operation of a heap leach can improve
production and optimize the use of sulfuric acid, which results in a more effective management of
resources and helps reducing consumable costs. Constant monitoring and control of key variables
along the pad ensures uniform irrigation at all levels of the modules, avoiding injuries of the
personnel during maintenance sampling perforation. Moreover, online wireless monitoring gives
you actionable information from the process to make informed decisions based on real time data
and from the safety of the control room, protecting the plant personnel of toxic environments and
risky working sites, minimizing maintenance rounds in the pad.
As a result, of the applications studied, users can monitor variables in real time getting all the
process information needed to make better and informed decisions to increase productivity and
plan better actions plans based on real time data.
The open communications platform allows an easy and quick installation reducing costs and
maintenance, also releasing operators to depend of proprietary tools and resources requirements.
The non-proprietary solution that can be easily installed without the need of electrical
infrastructure nor mechanical modification of the process.

1
INTRODUCTION
Whether you are leaching oxides or sulfides, today’s market places more importance on improving
the efficiency of your process, including optimizing solvent flow or providing a stable environment
for your bacteria. In order to do this, you need continuous, accurate measurements of your pad,
however, with the size and location of most pads, automation can be difficult and the process is
lightly controlled. And, because many things are controlled manually, your people have to spend
more time at the leaching site taking readings, adjusting valves, and being exposed to potentially
hazardous materials.
This document describes introductory concepts pertaining to heap leaching in pads. This process
requires a wide supporting base to place the first layer for waterproofing purposes, on which
minerals obtained from crushing and milling are irrigated. This material is distributed in different
sections in trapezoid shape and predetermined height to proceed with its irrigation combined with
a sulfuric acid solution. Following the solution percolation throughout the whole pad, the enriched
liquids are collected and taken to the mineral-substance recovery processing plant. The sulfuric acid
used during pad irrigation is a critical and costly component in this process, from which copper
sulfate is obtained. Efficient operation depends on constant monitoring at critical points during the
whole process.

METHOD

Challenge
Fulfilling the minerals and metals demand of the world’s growing population is a daunting task.
Add to it the array of economic, environmental, and political challenges you face and that endeavor
can seem nearly impossible sometimes. Nevertheless, your stakeholders expect profitability, which
means you need to create process efficiencies and keep your equipment operational, while also
ensuring the safety of your staff.
During heap leaching, monitoring the uniform irrigation of sulfuric acid is crucial for ensuring
accuracy and efficacy in the leaching process. Monitoring this variable consists of properly
controlling irrigation rates and pressure management. Other challenges to face in this process
pertain to monitoring of the pH, bacterium temperature, air injection and pressure control valves.
The pad in which the solution in question was applied has a length of 2 km and a width of 650
meters, and its remote location as well as harsh operational conditions made wired equipment
difficult to implement.

Solution
Emerson Process Management offers a complete solution to approach each of the challenges
described with wireless field instrumentation following protocol WirelessHART. This solution
allows the removal of wiring and external power in highly corrosive environments in large and
remote areas.
WirelessHART is a communication protocol implemented in wireless mesh networks designed to
fulfill the needs of process automation applications. Wireless networks installed among field
devices have redundant, safe and robust paths, are capable of auto resetting and offer at least 99.9%
reliability.

2
Auto resetting refers to the capability of field devices to automatically detect others in the mesh
with which they can establish communication as well as to the ability to automatically
communicate with other devices that they deem most appropriate.
Emerson’s SmartWireless solution in heap leaching consists of implementing a WirelessHart
network for variable control and monitoring. The instrumentation or devices were incorporated
into the wireless network and their information was entered in the plant information system
(SCADA) through a Gateway port.
This measuring system transfers data among field devices to the current control or data acquisition
system in a fully user-friendly manner, not interrupting the operation process.
The physical layer is compatible with ISM (Industrial, Scientific and Medical) of 2.4GHz (from 2,400
to 2,482.5 MHz) with DSSS modulation and frequency hopping. The device performs the hopping
to the bandwidth channels it requires, doing so automatically and not needing pre-configuration.
The data link layer is compatible with IEEE 802.15.4, which is the basis for the development of
WirelessHart and certified as communication standard by IEC 62591 to ensure interoperability
among different manufacturers.
Another tool used is the Time Division Multiple Access (TDMA) to arbitrarily assign and
coordinate communication among network devices.
Communications are highly safe using encrypted blocks AES-128, which include a password per
device to access the network and other codes for accessing Data-Link through a network password.
The wireless network architecture with Industrial Wi-Fi communication (IEEE 802.15.4- IEC 62591-
WirelessHART) has been designed as a device data-acquisition network.
 The flow measurement solution includes the use of a WirelessHart differential pressure
transmitter and a measurement primary element, which is defined according to process
features (measurement principle and materials used).
 The temperature measurement solution uses a temperature transmitter on a single
measurement point or a multivariable WirelessHart temperature transmitter with a 4-
input capacity and its corresponding temperature sensors.
 The pressure measurement solution uses a WirelessHart manometer transmitter, which
allows remote monitoring.
 The pH measurement solution uses a pH analyzer and Ph sensor element defined
according to the process characteristics.
 The pressure control valve solution uses discrete transmitters that integrate with the
pressure control valve performing opening and closing tasks as well as remotely flow rate
control.
 The positioning monitoring of the proportional valve utilizes a WirelessHart positioner
integrated into a proportional control valve, which allows to determine the valve position
remotely.

3
RESULTS
It is noteworthy to mention that this presentation will analyze the actual implemented solution,
which obtained the following results:
 A collapse risk-reduction of 25%
 A 3% increase in copper production
 A reduced environmental impact due to decreasing the use of sulfuric acid down to 14%,
which resulted in yearly savings of $140,000.

CONCLUSION
The WirelessHart protocol adds a new dimension to the process automation field. Based on the
proven success of the HART protocol, WirelessHart technology broadens the scope of process
instrumentation ensuring compatibility with current systems and practices.
WirelessHart is the first wireless standard available for industrial process monitoring. Smart
Wireless solutions allow users to design new applications and plant procedures, both on site and
remotely, offering process-improvement opportunities as well as extending the useful life of the
plant assets.
Emerson SmartWireless products extend digital architecture benefits, improving plant efficiency,
providing information previously not accessible, allowing asset control optimization, process
automation and simplifying both decision-making and management procedures.

4
EDITOR

Claudia Velásquez

Claudia holds a M.Sc. degree in Extractive Metallurgy Engineering from the Universidad
de Santiago de Chile. She has over 20 years of professional experience in the mining
industry, primarily focused on metallurgical laboratory testing, simulation of grinding
circuits and geometallurgical development.