John A. Meech University of British Columbia Department of Mining and Mineral Process Engineering 6350 Stores Road, Vancouver, B.C., V6T 1Z4, Canada Tel.

John A. MeechUniversity of British Columbia

Department of Mining and Mineral Process Engineering6350 Stores Road, Vancouver, B.C., V6T 1Z4, Canada

Tel : (604) 822-3984Fax : (604) 822-5599

Email : [email protected] .ca

Intelligent Methods in Mineral Processing - Treating the Mine-Mill Complex as a Factory

OutlineOutline • Background to Problems• Strategies to Follow• Incentives for Integration• Complexity Analysis• Intelligent Manufacturing Systems• IMS Architectures - agent-based / holonic systems• Structure of an Agent• "Swarm" Intelligence• Applications in Mining and Processing• Overview of IPMM• Conclusions and Recommendations

BackgroundBackground

• The mining industry is at a crossroads facing:

– ever-declining commodity prices– difficulties in marketing– high competition from abroad– increasingly complex ores– decreasing ore grades and reserves– a very poor image in society

Strategies to FollowStrategies to Follow

1. continue the routine of cutting costs

– labour-reduction – adoption of new technologies

2. expand the organizational horizon to – integrate activities across the mine and mill

– include value-added down-stream processing

Incentives for Option 2Incentives for Option 2• Impurities and Material Quality Issues

• New Processes

• Local Markets

• Recycling

• Value-added

- may require separate processing

- allow final product production at the mine

- can sustain production of final product

- can create new markets

- additional value ( gold jewellry, Polar diamond)

Incentives for Option 2Incentives for Option 2• Regulations

• Infrastructure

• Design impact

• Local resources

• Delivery costs

- can provide reasons for value-added

- can sustain mining in remote regions

- down-stream processing can affect design decisions

- power, rail, shipping ports, etc. may provide benefits

- savings in transportation costs

An Important Additional IncentiveAn Important Additional Incentive

• Complexity Analysis

– complex, interactive decision-making across an enterprise has not been possible in the past

» poor data-communication » poor data-collection» poor data-analysis

– such is not the case today

The Advent of "Complex" AnalysisThe Advent of "Complex" Analysis• Options can provide flexible response to

– changing commodity prices– competition from other sectors

– complex ore changes (grades and hardness)

– complex technology changes communication systems robotics advanced materials nanotechnology

composite materials vs. superalloys fibre-optics vs. coaxial cable coal vs. petroleum products

aluminum vs. copper

Attributes of an Intelligent Manufacturing SystemAttributes of an Intelligent Manufacturing System

• Collect and manage large amounts of data

• Analyse data to optimize across departments

• Develop simulation models of interactions between independent parts of an organization

• Apply intelligent robots to perform routine tasks

• Simulate assembly lines & plant processes to discover new ways to coordinate processing steps

Flexibility - the Key to Intelligent OperationFlexibility - the Key to Intelligent Operation

Create alternate plans Expand mine production Maintain production costs (or reduce) Change mill circuit layout Adjust product mix and/or quality

Flows in an IMSFlows in an IMS

• Materials and Resources

• Information (messages and/or data)

Interactions between process stages are treated as seller-customer or server-client relationships

Architectural Features of an IMSArchitectural Features of an IMS

• High-level tasks are decomposed

• Simulation conducted at different times/resolutions

• Behaviours are decomposed into sub-functions

• Functions are distributed across the system

ENTERPRISE

PLANT WIDE

SUPERVISORY CONTROL

DIRECT CONTROL

PROCESS INSTRUMENTATION

PROCESS

Traditional System HierarchyTraditional System Hierarchy

NASA/NIST STANDARD REFERENCE MODELING ENVIRONMENT

maps

objects

statevariables

objectivefunctions

programfiles servo

control

SensorProcessing

WorldModeling

TaskDecomposition

detectand

integrate

modelevaluation

planand

execute

TimeScale

UserInterfaces

path planning

operationalscheduling

taskactions

M1

M2

M3

M4

M5

S1

S2

S3

S4

S5

E1

E2

E3

E4

E5 milliseconds

seconds

minutes

hours

days to years

dynamicoperations

Multiple

Intelligent Manufacturing SystemsIntelligent Manufacturing Systems

after Monckton, 1997

Elements of an Intelligent SystemElements of an Intelligent System• rule-based modeling (expert systems)• fuzzy logic inferencing • artificial neural network modeling• genetic algorithm optimization• ability to explain and justify• ability to adapt or learn from experience• management of temporal-reasoning • agent-based architectures• "swarm" intelligence

Real-Time Intelligent Control System ModulesReal-Time Intelligent Control System Modules

User Interfaces Multiple

Inference Engine

InterNet Bridge

Process Bridge

Blackboard

Knowledge Base

Artificial Neural Network

Genetic Algorithm

What is an Expert System?What is an Expert System?

have been in use since the early 1970s method based on how we store memories symbolic reasoning is central to the method syntax is easy to learn and use symantics of a knowledge base is easy to

understand but difficult to create expertise is acquired incrementally from

interviews with an expert (or experts)

Who or what is an Expert?Who or what is an Expert?

…someone who everyday knows more and more about an ever-diminishing field until the scope of knowledge becomes so small that he/she knows everything about nothing.

…the man from out of town!An expert is…

Einstein

… simply someone who has acquired specific knowledge about a special area acquired over years of working with a process or piece of equipment.

Acquiring KnowledgeAcquiring Knowledge

The man from out of town is not necessarily the expert.

Rather this person is

The Knowledge Engineer

The Knowledge EngineerKnowledge Engineer must work in a collaborative way with the ExpertExpert to extract the gems of knowledge and then …

...code it into a computer program using special AI techniques such as - fuzzy logic

- neural networks - genetic algorithms

Acquiring KnowledgeAcquiring Knowledge

Sometimes multiple experts are involved

Acquiring KnowledgeAcquiring Knowledge

Sometimes special consultants are needed

????

!!!!!

!!!!!!!!!!????

????

Acquiring KnowledgeAcquiring Knowledge

Sometimes knowledge overload occurs

Acquiring KnowledgeAcquiring Knowledge

…an interrogation

Care must be taken that an interview does not become…..

Acquiring KnowledgeAcquiring Knowledge

The exercise must not be viewed as a competition

Acquiring KnowledgeAcquiring Knowledge

Rule-based ModelingRule-based ModelingRule Name: water_valve_high

IF tank level is definitely "high"AND pump speed is "maximum"THEN valve position change is "closed a lot"

DEFUZZIFY (valve position) FIND (pulp flowrate * ) WAIT ("water_valve_high", 120 )

ELSE valve position change is not "closed a lot"

Fuzzy-Logic InferencingFuzzy-Logic Inferencing

Low Medium HighDegree

ofBelief

tank level

00 6 12

100

Fuzzy Associative MapFuzzy Associative Map

tank level

low medium highmed-low med-highpumpspeed

minimum

normal

maximum

openeda lot

notchanged

notchanged

notchanged

closeda little

closeda lot

closeda lot

closeda lot

a lotopened

a littleopened

a lotopened

a littleopened

a littleopened

closeda little

closeda little

Artificial Neural NetworksArtificial Neural Networks

based on the neuronal structure of the brain applied where data exists but no model has true learning capability slow to adapt but fast to operate applications

– predictive monitoring (soft sensors)

– image analysis

– pattern recognition

Artificial Neural Network ModelingArtificial Neural Network Modeling

Artificial Neural NetworksArtificial Neural Networks

W1j

W2j

W3j

W4j

Wnj

input 1

input 2

input 3

input 4

input n

output j

Basic Neuronal Equation

inputs = 0 to 1 outputs = 0 to 1 weights = - to +

Genetic AlgorithmsGenetic Algorithms

high-speed optimization method based on "Survival of the Fittest" data are coded as chromosomes

- 01101 wherein each digit represents a

different variable and its current level

each dataset is combined with another "fit"

dataset to create a "child" solution each generation is "fitter" than the previous one

Operators in GAOperators in GA

Selection for reproduction

Cross-over operator

Mutation operator

Elite strategies (cloning)

Real-Time Intelligent Control System ModulesReal-Time Intelligent Control System Modules

User Interfaces Multiple

Inference Engine

InterNet Bridge

Process Bridge

Blackboard

Knowledge Base

Artificial Neural Network

Genetic Algorithm

Intelligent User InterfacesIntelligent User Interfaces

• Process mimic diagrams

• Trend diagrams of data vs. time

• Windows to view and log messages

• Explanation and Justification Abilities

• Message filtration into classes for each user type

Agent-based IMS Structure• Holonic manufacturing systems

- A holon is an individual element of a whole

- Holons can be made up of other holons- resource holons- product holons- order holons- control holons

• Modeling methodology can be applied to a hierarchy to create a heterarchical system in both time and space

Holonic Manufacturing SystemHolonic Manufacturing System

Object B

machine

Holonifier

Customer

" I want Object C ”

Information

Holon

Transport

Holon

Material flowInformation flow

AGVSystem

Holonifier

AssemblyRobot

Holonifier

after Monostori and Kadar, 1999

Object C

Holonic system

Object A Object A'

Structure of a resource agentStructure of a resource agent

Registrationmechanism

LocalDatabase

Material flow

Communicationagent

Message processing

Inputbox

Output box

KnowledgeBase

Incoming message Sent message

MaterialProcessing

Resource supervisoragent

after Monostori and Kadar, 1999

Data Models and Communication

• Product Data Management systems

- STEP system under ISO 10303

• CORBA Communication Protocol

- Common Objects Request Broker Architecture

- developed by the Object Management Group

Architecture of a CORBA Architecture of a CORBA Communication Protocol SystemCommunication Protocol System

after Nicoletti, 1999

DynamicInvocation

IDLStubs

ObjectAdapter

ORBInterface

DynamicSkeleton

IDLSkeleton

ORB Core

ORB CoreClientObject

Implementation

Agent Types and System DesignAgent Types and System Design• There are 4 agent types:

- problem-solving agents

- information agents

- service agents for other agents

- translation agents

• Aspects of designing an agent system are:

- number of agents required

- number of types of agents

- number of actions performed (complexity)

System Design IssuesSystem Design Issues Structure - level of self-containment of an agent

Communication - protocols & interchange language

Group formation - persuading machines to participate in

a group -- reward/penalty systems

Configurability - addition/deletion of machines/groups

Scalability - scale-up to the extended enterprise level

Global vs. local optima - dealing with 'selfish' agents

1. Virtual Rapid Prototyping on the Web - interactive automation tools to simulate conceptual design

2. Enterprise Information Integration Agent System - a collaborative infrastructure for large-scale integration

3. Multi-Agent Framework for "Lean" Manufacturing - customer-driven with globally synchronized-scheduling

4. Internet agent-based Infrastructure for Mass Customization - Internet supports global communication between

customers and manufacturers

Intelligent Manufacturing and the WebIntelligent Manufacturing and the Web

after G. Nicoletti, IPMM-2001

Web-based Collaborative Engineering Design

- Adaptive Modeling Language (AML) demo by TechnoSoft Inc.

- developed from a single-user, single-computer environment used to model complex engineering problems

- a Dual Use Science & Technology (DUS&T) agreement Air Force Research Laboratory,

Lockheed-Martin Electronics & Missiles, and TechnoSoft Inc.

- multiple users interact simultaneously with a unified parts model over a network of geographically-distributed machines

Chemaly, IPMM-2001

Matrix of Launch Vehicle Design DisciplinesMatrix of Launch Vehicle Design Disciplines- Zweber et al., IPMM-2001- Zweber et al., IPMM-2001

Lockheed-Martin's Missile Design NetworkLockheed-Martin's Missile Design Network- Zarda et al., IPMM-2001

Optimization-based Design:Optimization-based Design:The Multi-Process Design ExecutiveThe Multi-Process Design Executive

- software package to design multi-stage materials processes

- based on the Adaptive Modeling Language (AML)

- integrates models of materials, geometry, processes, equipment, and cost with optimization algorithms

- a tool for preliminary selection of manufacturing processes

- to evaluate alternate processing sequences and parameters at early design stages, when decisions have the greatest influence on cost

- demo-ed processes to manufacture Ti-alloy turbine engine disks

E. Medina and W. G. Frazier, IPMM-2001

Processing Sequence Design Processing Sequence Design ProblemProblem

Process Sequence Object

Initial Workpiece

Final Workpiece

P1 P2 Pi Pn

Optimization Objects

D1 F1

Design System Main Control Object

Ai

Di Fi

Bi

Virtual Manufacturing Environment

Web-based interface to integrate material process design and analysis modules

models of various manufacturing processes

module to view the output as a 3D model in a web browser

interface for headgear and Data-Glove to provide an interactive, immersive environment

B. Mehta, IPMM-2001

http://webme.ent.ohiou.edu//vm/

VRML model of the strip rolling process

Swarm IntelligenceSwarm IntelligenceAnt Colonies exhibit

"collective" intelligence

The Civil or Mining Engineers

of the Insect World

Ants can Fold a LeafAnts can Fold a Leaf

Ants can Build a BridgeAnts can Build a Bridge

Ants can FarmAnts can Farm

• Harvesting food• Storing food• Feeding their young• Serving their Queen

WHAT IS SWARM INTELLIGENCE?WHAT IS SWARM INTELLIGENCE?

Refers to a higher-level "intelligence"

autonomous agents acting in their natural environment

each with local low level behaviour

collective action results in an "apparent" intelligence

Swarm Intelligence and ModelingSwarm Intelligence and Modeling

• Can help solve complex problems by providing- a distributed model- an adaptable model- a flexible and robust model- an extremely fast optimization algorithm

• Fits in well with agent-based models- a centralized program is replaced by an emergent

and distributed set of autonomous functional entities

ANT COLONY OPTIMIZATIONANT COLONY OPTIMIZATION

Applications

– Travelling Salesman Problem– Telecommunication Channel Assignment– Vehicle Routing (shovel/truck scheduling)

ANT COLONY OPTIMIZATIONANT COLONY OPTIMIZATION

Benefits

– Solve extremely large-scale problems– Faster than Genetic Algorithms– Highly adaptable to changing conditions

Adaptability of AntsAdaptability of Ants

Adaptability of AntsAdaptability of Ants

Adaptability derives from cooperation of individuals (not intelligence) because of 2 factors:

1. Pheremone signals between ants

2. Stimulus-response of each ant

The collective response guarantees survival of the colony

Block 1

Block 2

Block n

Maintenance/ Emergency repairsMaintenance/ Emergency repairs

TargetTarget

Analogy Between Ants and Shovel-Analogy Between Ants and Shovel-TrucksTrucks

Feeding and Herding the African AntsFeeding and Herding the African Ants

Applications in MiningApplications in Mining

• Systems Design and Simulation

• Orebody Modeling

• Long-Range Planning of Production Options

• Mine Planning and Scheduling

• Optimization Studies on Mine/Mill Interface

Applications in MiningApplications in Mining

• Improvements in Environmental Control

• Vertical Integration Opportunities

• Strategic Planning of Investment/Expansion

• Intelligent Stockpiles

• Enhanced Comminution Systems

Applications in MiningApplications in Mining

• Coordinated Real-Time Maintenance

• Tele-remote Operations

• Enhanced Data Communication Protocols

• Discovery of New Ideas

• Value-Added Production at the Mine

Example 1: Highland Valley CopperExample 1: Highland Valley Copper

• Optimized comminution requirements- blasting (fragmentation)- primary crushing (10" > 4")- semiautogenous grinding (SAG)

• Benefit - increased throughput by up to 30%

• Discovery- SAG milling is not a legitimate unit process

Example 2: Mount Isa Mines Example 2: Mount Isa Mines

• Examined orebody to provide stable mill feed- ore hardness (variance reduced by 10%)- head grades (variance reduced by 25%)- ore reserves reduced by 25%

• Benefits - increased throughput by 15%- improved recovery by 5%

• Discovery- new methods to treat lost reserves

Example 3: Harmony Gold MineExample 3: Harmony Gold Mine

• Installed new process to produce 99.99 % Au- refining stage after bullion production- new process for gold bars (powder metallurgy)- jewelry production at mine site

• Benefits - new opportunities for local labor force- increased marketing opportunities

• Discovery- can market gold to consumers on the InterNet

Example 4: Ekati Diamond MineExample 4: Ekati Diamond Mine

• Invested in jewelry production outside of CSO- set up new facility in Yellowknife- marketing the "Polar Diamond"- about 20% of total production

• Benefits - new job opportunities for local labor force- increased marketing opportunities

• Discovery- can market diamonds directly to consumers

The Polar Diamond BrandThe Polar Diamond Brand

The Polar Diamond CertificateThe Polar Diamond Certificate

Example 4: Globalcoal.comExample 4: Globalcoal.com

• Joint Venture by 4 of the largest mining companies- Anglo American- Billiton- Glencore International - Rio Tinto

• Created a single online marketplace for thermal coal- set to begin February 2001- will be expanded to iron ore and base metals- threatens conventional markets such as LME- provides opportunity to market to many customers

Example 5: Internet CommerceExample 5: Internet Commerce

• Australian mining companies have set up a B2B market web site to provide auction opportunities for multiple suppliers and consumers of raw materials

• BHP is planning to sell "rough" uncut diamonds over the Internet to wholesalers wishing to take their stones to a jeweller to have them cut and designed the way they want, bypassing numerous intermediaries. GST payments are reduced as well.

ConclusionConclusion

• Alternate strategies to cost-cutting are required

• Opportunities exist to apply Intelligent Manufacturing Systems based on Agent or Holonic principles

• IMS can provide data collection and data analysis at various time and resolutions to conduct simulation modeling

• Value-Added production at the mine site can be examined using an IMS system

• High-tech Internet applications can lead to significant improvements in the industry's image and competitiveness

IPMM-2001IPMM-2001

Conference Theme"Cross-Disciplinary Research in IPMM

- an Essential Ingredient for Innovation!”

A Brief History of IPMMA Brief History of IPMM

Founded in 1997 in Gold Coast, Australia. In 1999, the 2nd International Conference

was held in Honolulu, Hawaii. Now we have completed the traverse of the

Pacific arriving at the home of IPMM -

Vancouver, British Columbia

What is IPMM ?What is IPMM ? An eclectic group of scientists, engineers, and researchers with a wide variety of backgrounds

• materials science & engineering, mechanical & electrical engineering• mining, processing & metallurgical engineering• computer science & engineering and biological computing• manufacturing & industrial engineering• chemical engineering and civil engineering (structures & transportation)• environmental sciences & engineering• astronomy and space exploration• HMIs & ergonomics / psychology (emotions in decision-making)• image analysis & vision analysis / measurement & instrumentation

Bill Reid’s Jade Canoe - The Spirit of Haida

Gwaii

"...The Spirit Canoe…is an exploratory vessel, sailing an unknown course through unknown seas. Beings looking for other beings to speak to, feast with, trade with..."

Bill Reid - 1992

Background to IPMMBackground to IPMMsimilar to the creatures in the Jade Canoe, IPMM Members are also travelling:

• an uncertain odyssey to an unknown destination

• looking for – new ways to understand materials– new processes to fabricate products

• we focus on applications but there is always room in the boat for new theories and ideas

• we gather every two years to share in our new knowledge and experience

SomeSome people may ask:

“Why should a mining or processing engineer participate in a conference with an astronomer?”

“How can a scientist studying manufacturing techniques possibly gain anything from listening to a psychologist?”

“What can a ‘soft’ scientist learn from a mining engineer?”

“I'm a materials researcher. Why should I care about these so-called "intelligent" methods?”

Legitimate Questions - here are some answers

the world has become a much more complex place in which to work and study.

no single person or group can adequately hope to find the "right" answer any more.

there may no longer even be a "right" solution. "intelligent" methods derive from single minds operating in

a collaborative environment. issues must be addressed using a multi-dimensional

approach – one which lends itself to input from cross-disciplinary

teams.

Collaboration Collaboration – the Key to – the Key to InnovationInnovation

Ideas spring from a single mind. Even the best minds freely admit that they

performed at the top of their abilities when they were "collaborating".

The question is - "how can we create environments whichcapture the best of truly ‘great’ collaboration?"

The Return of the Generalist The Return of the Generalist

"You can lead a person to knowledge, but you can't make them understand it.”

"While the Internet may have democratized the availability and access to Knowledge, Intelligence

is a commodity that can never be distributed uniformly

- it must be shared to be useful and to be used!"

Concern for People is Key Concern for People is Key

• sharing comes from mutual respect and trust

• a collaborative system must do more than simply provide a common work space

• it must not inhibit creativity and innovation

• searching for "intelligence" must be the goal

• an individual reward system is essential

Paper-Recycling and the Paper-Recycling and the InternetInternet

The goal of the University of Newcastle

Paper Usage Action Plan is

To reduce the consumption of paper products.

with a 4-point plan:– Reduce paper consumption for University communication.

– Maintain & establish programs for recycling and reuse of paper products.

– Encourage "environmentally-friendly" stationery and business equipment.

– Encourage "environmentally-friendly" bathroom paper products.

The Paperless OfficeThe Paperless Office

The Internet was supposed to give us the Paperless Office. Instead paper use has increased steadily - Why is this? In migrating from one paradigm to another, change is resisted and we continue using paper

- even more so, as we search for the "perfect" draft! As more people use the new tools, paper use goes up. As our comfort-level with the environment increases, slowly we

stop using paper naturally and entirely! – no hard copy reports – only email communication or wireless cell phones

IPMM and Paper UseIPMM and Paper Use

InterNet--

Year

1997

1999

2001

2003

Pages

220015501750

?

hard copy--

CD-ROM--

Price ($ US)~ 20,000~ 15,000

~ 500~ 0

in 2003 the Proceedings will be entirely on the Internet

Paper-Recycling and the Paper-Recycling and the InternetInternet

» Rules should be made for the benefit of the group in total, not for a single individual or sub-group

» Rules should not stifle creativity and innovation

» The World is made up of three main groups the Sergeants (or Bosses/Decision-makers) the Anarchists (or Thinkers/Rebels) the Uppers (or Workers/Believers)

» I submit - we must find the "intelligence" in these activities -- who are these systems designed for?

Why legislate something that is a natural evolution?

Other ExamplesOther Examples

» The Vancouver "Air-Care" Program» "Blue-Box" Programs» Regulating the Internet

"There is nothing more useless than doing efficiently what shouldn’t be done at all "

- Peter Drucker

Why legislate something that is a natural evolution?

The Evolution of the InternetThe Evolution of the Internet Year Number of hosts Innovation1965 2 ARPA(DARPA)1968 4 ARPANET1971 8 Telnet, Ethernet1974 32 TCP/IP - UUCP - FTP1978 100 USENET1983 1,000 DNS1987 10,000 T11989 100,000 World Wide Web/HTML1992 1,000,000 T31995 10,000,000 first e-business1999 100,000,000 first software agent

soon there will be more host computers than people

The Trans-humanist and Post-humanist Societies

The Age of MachinesThe Age of Machines

"If you can hang on until 2016, you will never die!" - J.W. Lewis

The Age of MachinesThe Age of Machines

BenefitsDemocratization of Information and the advent of "Empire"

"a fluid, infinitely expanding and highly organized system

encompassing the world's entire population." - Michael Hardt and Antonio Negri

Computers outperform Humans in thinking and in emotionsNanotechnology will combine with Computational IntelligenceNo more Human "Wet" Diseases

Closing

» organizing biannual meetings to discuss "intelligent" methods for material production and manufacturing

» providing a collaborative environment to share in new ideas across multiple disciplines

» creating a society that understands the importance of "intelligent" approaches to processing and

manufacturing of materials» promoting the use of "intelligent" thinking in the

important technical activities of the 21st Century

IPMM is fulfilling an important function byIPMM is fulfilling an important function by

IPMM’03IPMM’03

The 4th International Conference on

Intelligent Processing and Manufacturing of Materials

Tohoku University, Sendai, Japan

May 18 - 23, 2003

Theme:

Nanotechnology for the 21st Century

– do good things really come in small packages?

Fuzzy-Woozy meets Fuzzy Logic

Fuzzy-Woozy Logic

An Illusion of a Reality that is of itself a Reality