The KPIs in the mining operations range from quality improvement and consistency, equipment reliability, energy savings and safety. The Model Predictive Control looks at different variables and control the process to desired outputs, thus meeting and exceeding your operations KPIs.

1. Optimization using model
predictive control in mining

2. Optimization using model predictive control in mining | 2
Rockwell Automation is a member
of the Association for Packaging and
ProcessingTechnologies (PMMI) –
a trade association made up of more
than 700 member companies that
manufacture packaging, processing and
packaging-related converting machinery,
commercially-available packaging
machinery components, containers and
materials in the United States, Canada
and Mexico.
PMMI members are the industry-leading
solutions providers on your processing
and packaging supply chain, and PMMI
resources help you connect with them.
Table of contents
Optimization using model predictive control in mining
n Rockwell Automation at a glance. . . . . . . . . . . 3
n Optimize your mining operation. . . . . . . . . . . . 5
n The facility: Consists of MVs, DVs, CVs. . . . . . . 6
n A simple MPC problem. . . . . . . . . . . . . . . . . . . . . . . 7
n So what makes MPC different?. . . . . . . . . . . . . . 8
n MPC versus expert systems. . . . . . . . . . . . . . . . . 10
n Pavilion MPC applications. . . . . . . . . . . . . . . . . . 11
n Fine crushing. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
n Grinding circuit. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
n Flotation cell. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
n Flotation column. . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
n Thickener. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
n Material flow management. . . . . . . . . . . . . . . . . 27
n Controller matrix. . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
n How MPC generates benefits. . . . . . . . . . . . . . . 31
n Model predictive control. . . . . . . . . . . . . . . . . . . . 32
n Value first. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
n Level 1 support. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
n Level 1 and 2 support. . . . . . . . . . . . . . . . . . . . . . . 35
n Scalable support offerings. . . . . . . . . . . . . . . . . . 36

3. Optimization using model predictive control in mining | 3
$6.62B
Fiscal 2014 sales
80+
countries
22,000
employees
Rockwell Automation at a glance
Leading global provider of industrial
power, control and information solutions
Serving customers for 111years
• Technology innovation
• Domain expertise
• Culture of integrity
and corporate
responsibility
20+industries

4. Optimization using model predictive control in mining | 4
Bringing you a world of experience
PAVILION
SOLUTIONS
Helping you exceed your business goals Innovation ProductsDomain
expertise
• The most powerful
predictive modeling
software in the
industry
• Enhancing the
profitability of
global manufacturers
• Delivering the
highest ROI in the
industry
• Combining
technology
and application
knowledge
• Targeted industries
• Best practices from
multiple industries
• Pavilion8®,
Pavilion® RTO and
Software CEM®
• Certified project
managers
• Repeatable,
measurable,
auditable
• Risk management
80 Countries I 20 Languages I 2500 + employees I Average 13+ Years’Experience I Single point of contact
Biofuels | Energy Optimization | Environmental | Food Beverage | Minerals, Mining, Cement | Natural Gas Liquids | Polymers | Water/Wastewater

5. Optimization using model predictive control in mining | 5
Optimize your
mining operation
Model predictive
control solutions
Our mining capabilities help
improve your operations
• Reduce grade variability
• Increase product recovery
• Reduce reagent consumption
• Reduce energy costs
• Increase throughput
Drive your operations to
its maximum potential
everyday with MPC
APPLICATIONS
• Crushing
• Grinding
• Flotation
• Thickening
• Material flow
management

6. Optimization using model predictive control in mining | 6
Controlled Variables (CVs)
Process variables that need
to be maintained at a target
or within a set range
Controlled Constraint
Variables (CCVs)
Process variables that should
not be allowed to violate
limits (upper, lower or within
a range)
Manipulated
Variables (MVs)
Process variables you can
adjust that affect the CVs
(typically PID set-points)
Disturbance
Variables (DVs)
Measured process variables
that affect the CVs that are
not MVs
The facility:
Consists of MVs, DVs, CVs
DV = Wind,
Road Slick,
Other Cars
MV =
Accelerator
CV = Speed
(Maximize)
CCV = Oil
Pressure
Example

7. CV
CCV
MV
DV
Optimization using model predictive control in mining | 7
A simple
MPC problem
GOALS
Maximizing ore flow
(Operate to maximum constraints)
CONTROLLED VARIABLES
Production rate
CONSTRAINTS
Belt maximum current
Belt maximum eight
MANIPULATED VARIABLES
Feed gate
Belt speed
DISTURBANCE VARIABLES
DV material density
CLOGGING
OVERLOAD
OVERLOAD

8. Optimization using model predictive control in mining | 8
So what makes
MPC different?
PID MPC
Single input – single output controller
Multiple input – multiple output controller
Control strategy based on a centralized approach
All variables are simultaneously considered
Control based on current error
Predictive control
Controller action based on current and anticipated
future PV deviations from target
Poor ability to handle process delays,
disturbances, interactions and non-linearities
Compensates for process delays, disturbances,
interactions and non-linearities
Poor ability to handle different types of
disturbances and set-point signal forms
Optimal control for all types of disturbances
and set-point signal forms
Poor ability to handle constraints Predictive handling of constraints
Poor ability to optimize (maximize or minimize) Capability to optimize process (maximize or minimize)

9. ✔
✔
✔
Optimization using model predictive control in mining | 9
EQUAL
DISTRIBUTION
NO
OVERLOAD
So what makes
MPC different?
PREDICTIVE
CONTROL
DIRECT PROPERTY
VARIABLE CONTROL
EXPLICIT DYNAMIC
MODELS

10. Optimization using model predictive control in mining | 10
MPC v Expert Systems
EXPERT SYSTEMS MPC
Utilizes a model of operator reactions Utilizes a model of process behavior
Rule-based/fuzzy logic control solver Predictive control solver
Inefficient (operator comprehension) Optimal (mathematical sense)
High maintenance required Low maintenance

11. Optimization using model predictive control in mining | 11
Pavilion8 MPC applications drive your existing
infrastructure to Maximum Performance
CRUSHING GRINDING FLOTATION THICKENING MATERIAL FLOW
Reduced
Product
Variability
Increased
Product
Recovery
Reduce
Reagent
Consumption
Reduce
Energy
Costs
Mining
Applications

12. Optimization using model predictive control in mining | 12
CUSTOMER CHALLENGES
• Poor level control in the secondary and tertiary crushers
• Inability to keep secondary and tertiary crushers choked
in order to maximize efficiency and minimize wear
• Loss of throughput
• Frequent line shutdowns
• Poor balancing of secondary and tertiary crushers
MPC BENEFITS
• Increase throughput by 2%
• Increase efficiency by 5%
• Decrease equipment wear
by 5%
• Typical payback less than
1 year
Fine crushing
MPC | Fine crushing:
Increasing throughput

13. OBJECTIVES
• Maximize crushing throughputs
• Keep feed bin levels within
a range
• Control crusher levels
• Maintain crusher constraints
• Maintain belt constraints
Optimization using model predictive control in mining | 13
Fine crushing
MPC
Manipulated
Variable
Disturbance
Variable
Controlled
Variable
Constraint
Variable
Maximize
Upper
Constraint
Lower
Constraint
Upper Lower
Constraint
Target
MV DV CV CCV
From 1st
Crusher
Screen 1
2nd
Cone
Crusher
Belt 2
Belt 5
3rd
Cone
Crusher
FEED BIN 1
FEED BIN 2
SpeedMV
SpeedMV
SpeedMV
SpeedMV
SpeedMV
Speed MV
SpeedMV
Level
Level
CV
CV
LevelCV
Level
CV
Weight, Level, Speed,
Temperature, Position, Current
CCV
Weight, Level, Speed,
Position, Current
CCV
Weight, Level, Speed,
Position, Current
CCV
Weight, Level, Speed,
Position, Current
CCV
T OilCCV
T OilCCV
Belt 1
Belt 4
Screen 2

14. Optimization using model predictive control in mining | 14
Fine crushing matrix
OBJECTIVES
• Maximize crushing throughputs
• Keep feed bin levels within a range
• Control crusher levels
• Maintain crusher constraints
• Maintain belt constraints
Maximize
Upper Constraint
Lower Constraint
Upper Lower Constraint
Target
CV CV CV CCV CCV CCV CV CCV CV CCV CCV CCV
2arycrushingthroughput
2aryfeedbinlevel
2arycrusherlevel
2arycrusherconstraints
2arybelt1constraints
2arybelt2constraints
3arycrushingthroughput
3aryfeedbinlevel
3arycrusherlevel
3arycrusherconstraints
3arybelt1constraints
3arybelt2constraints
MV 2ary feed belt 1 speeds X X X X X X
MV 2ary feed belt 2 speeds X X X
MV 2ary crusher speeds X X X
MV 3ary feed belt 1 speeds X X X X X X
MV 3ary feed belt 2 speeds X X X
MV 3ary crusher speeds X X X
2% increase in throughput
5% increase in efficiency
5% decrease in equipment wear

15. Optimization using model predictive control in mining | 15
CUSTOMER CHALLENGES
• Oscillatory behavior and deviation from set-point
• Poor control of product quality (size)
• Reduced throughput due to constraints
• Unnecessary mill wear (steel-on-steel)
• Mill not running at maximum energy efficiency
MPC BENEFITS
• Maintain product quality (size)
• Increase throughput by 2%
• Decrease energy usage by 2%
• Decrease maintenance cost
by 2%
• Typical payback less than
1 year
Grinding circuit
MPC | Grinding:
Increasing throughput

16. Optimization using model predictive control in mining | 16
OBJECTIVES
• Maximize SAG mill fill
• Control SAG mill feed density
• Control hydrocyclone feed density
• Control P80 (product size)
• Maintain sump level within range
• Maintain hydrocyclone inlet
pressure below limits
• Maintain circulating pump
current below limits
• Maintain SAG mill bearing
pressure below limits
• Maintain ball mill power draw
below limits
• Maintain pebble recycle flow
below limits
• Maintain SAG mill power draw
below limits
Grinding circuit
MPC
Manipulated
Variable
Disturbance
Variable
Controlled
Variable
Constraint
Variable
Maximize
Upper
Constraint
Lower
Constraint
Upper Lower
Constraint
Target
MV DV CV CCV
SpeedMV
Fresh Feed MV
SpeedMV
FlowMV
FlowMV
Flow
F80
Number of
Hydrocyclone
Pressure
Power
Ball Mill
Bag Mill
Sump Tank
Cyclones
Nest
CurrentDensity
Power
CCV
DV
DV
CCV
CCV
CCVCCV
CCV
Water
Pebble
Crusher
Density
Sump Level
Level
F80
Flotation Feed
Water
CV
CV
CV
CV

17. Optimization using model predictive control in mining | 17
Grinding circuit matrix
Maximize
Upper Constraint
Lower Constraint
Upper Lower Constraint
Target
OBJECTIVES
• Maximize SAG mill fill
• Control SAG mill feed density
• Control hydrocyclone feed density
• Control P80 (product size)
• Maintain sump level within range
• Maintain hydrocyclone inlet pressure below limits
• Maintain circulating pump current below limits
• Maintain SAG mill bearing pressure below limits
• Maintain ball mill power draw below limits
• Maintain pebble recycle flow below limits
• Maintain SAG mill power draw below limits
CV CV CV CCV CCV CCV CV CCV CV CCV CCV
SAGmillfill
SAGmillfeeddensity
Hydrocyclonefeeddensity
P80
Sumplevel
HydrocycloneInletpressure
Circulatingpumpcurrent
SAGmillbearingpressure
Ballmillpowerdraw
Pebblerecycleflow
SAGmillpowerdraw
MV SAG mill ore feed rate X X X X X X X X X X X
MV SAG mill water feed rate X X X X X X X X X X X
MV Sump water feed rate X X X X X X X X
MV SAG mill speed X X X X X X X X X X
MV Circulation pump speed X X X X X X X
DV F80 X X X X X X X X X X
DV Number of Hydrocyclone X X X X X
Maintain product quality (size)
2% increased throughput
2% decreased energy usage
2% decreased maintenance cost

18. Optimization using model predictive control in mining | 18
CUSTOMER CHALLENGES
• Poor control of concentrate grade
• Lack of metal recovery control
• Excessive use of reagents
MPC BENEFITS
• Maintain concentrate grade
to reduce product give-away
or off-spec by 3%
• Increase metal recovery by 2%
• Reduce reagents by 3%
• Typical payback less than
1 year
Flotation
MPC | Flotation cell: Improving yield
and reducing reagent cost

19. Optimization using model predictive control in mining | 19
OBJECTIVES
• Control concentrate grade (purity)
• Maximize metal recovery
• Control pH
• Control froth depth
• Control bubble speed or
air hold-up
• Control bubble size distribution
(BSD) or bubble surface area
• Maintain liberation within a range
Flotation cell
MPC
Manipulated
Variable
Disturbance
Variable
Controlled
Variable
Constraint
Variable
Maximize
Upper
Constraint
Lower
Constraint
Upper Lower
Constraint
Target
MV DV CV CCV
P80DV
One GradesDV
Feed FlowsDV
Bubble SpeedCV
pHCV
Cell LevelsCV
Concentrate GradeCV
LiberationCV
Mass/Metal
Recovery
CV
Bubble SizeCV
Air Injection ValvesMV
Reagents FlowsMV
FlowMV
ModifierMV
DepressorMV
CollectorMV
Frother
P80
Grinding

20. Optimization using model predictive control in mining | 20
Flotation cell matrix
Maximize
Upper Constraint
Lower Constraint
Upper Lower Constraint
Target
OBJECTIVES
• Control concentrate grade (purity)
• Maximize metal recovery
• Control pH and froth depth
• Control bubble speed or air hold-up
• Control bubble size distribution (BSD) or
bubble surface area
• Maintain liberation within a range
Maintain concentrate grade
3% reduced product Give-away
3% reduced off-spec material
2% increased metal recovery
3% reduced reagentsCV CV CV CCV CCV CCV CV
ConcentrateGrade
MetalRecovery
pH
FrothDepth
Bubblespeedorair
holdup
Bubblesizedistribution
orsurfacearea
Liberation
MV Collector reagent flow ratio X X X X X X
MV Depressor reagent flow ratio X X X X X X
MV Modifier reagent flow ratio X X X X X X
MV Tail flow X X X X X X
MV Air flow X X X X X X
MV Frother reagent flow ratio X X X X X X
MV P80 grinding X X X X X X
DV Feed flow-grinding X X X X X X
DV Feed density-grinding X X X X X X

21. Optimization using model predictive control in mining | 21
CUSTOMER CHALLENGES
• Poor control of concentrate grade
• Lack of metal recovery control
• Excessive use of reagents
MPC BENEFITS
• Maintain concentrate grade
to reduce product give-away
or off-spec by 3%
• Increase metal recovery by 2%
• Reduce reagents by 3%
• Typical payback less than
1 year
Flotation column
MPC | Flotation column: Improving
yield and reducing reagent cost

22. Optimization using model predictive control in mining | 22
Flotation column
MPC
Manipulated
Variable
Disturbance
Variable
Controlled
Variable
Constraint
Variable
Maximize
Upper
Constraint
Lower
Constraint
Upper Lower
Constraint
Target
MV DV CV CCV
OBJECTIVES
• Control concentrate grade (purity)
• Maximize metal recovery
• Control pH
• Control froth depth
• Control bubble speed or
air hold-up
• Control bubble size distribution
(BSD) or bubble surface area
• Maintain liberation in a range
• Control bias flow (water flowing
down the column)

23. Optimization using model predictive control in mining | 23
Flotation column matrix
Maximize
Upper Constraint
Lower Constraint
Upper Lower Constraint
Target
OBJECTIVES
• Control concentrate grade (purity)
• Maximize metal recovery
• Control pH
• Control froth depth
• Control bubble speed or air hold-up
• Control bubble size distribution (BSD)
or bubble surface area
• Maintain liberation in a range
• Control bias flow (water flowing down the column)
Maintain concentrate grade
3% reduced product give-away
3% reduced off-spec material
2% increased metal recovery
3% reduced reagentsCV CV CV CV CV CV CCV
Concentrategrade
Metalrecovery
pH
Frothdepth
Bubblespeedorair
holdup
Bubblesizedistribution
orsurfacearea
Liberation
Bias
MV Collector reagent flow ratio X X X X X X X
MV Depressor reagent flow ratio X X X X X X X
MV Modifier reagent flow ratio X X X X X X X
MV Tail flow X X X X X X X
MV Air flow X X X X X X X
MV Frother reagent flow ratio X X X X X X X
MV P80 grinding X X X X X X X
MV Wash Water flow ratio X X X X X X X
DV Feed flow-grinding X X X X X X X
DV Feed density-grinding X X X X X X X

24. Optimization using model predictive control in mining | 24
CUSTOMER CHALLENGES
• Fresh water availability and cost
• Reagent cost
• Existing conventional control is inefficient
(long residence time, large disturbances and
non-linear behavior)
• Thickener shutdown may stop all the production line
MPC BENEFITS
• Increase water recovery by
3% (typically 9,000 m3
/day)
• Reduce flocculant by 2%
• Reduce possibility of
emergency feed
shutdown
• Typical payback less than
1 year
Thickener
Model Predictive Control (MPC)

25. Optimization using model predictive control in mining | 25
Thickening
MPC
OBJECTIVES
• Maximize (or control) % solids
(or density) in the mud
• Keep rake arm torque
below a limit
• Keep mud pump amps
bellow a limit
• Keep bed level in a range
• Keep water quality (turbicity)
in a range
Manipulated
Variable
Disturbance
Variable
Controlled
Variable
Constraint
Variable
Maximize
Upper
Constraint
Lower
Constraint
Upper Lower
Constraint
Target
MV DV CV CCV
Mad Pump AmpsCCV
Turbicity CCV
Feed FlowDV
Feed pHDV
Feed DensityDV
Mud DensityCV
Flocculant FlowMV
Rake Arm TorqueCCV
Flocculant FlowMV
Water Flow

26. Optimization using model predictive control in mining | 26
Thickener matrix
Maximize
Upper Constraint
Lower Constraint
Upper Lower Constraint
Target
OBJECTIVES
• Maximize (or control) % solids (or density)
in the mud
• Keep rake arm torque below a limit
• Keep mud pump amps below a limit
• Keep bed level in a range
• Keep water quality (turbicity) in a range
3% Increase in Water Recovery
2% Reduction in Flocculant
Avoid Feed Shut-down
CV CV CV CV CV CV
Bedpressure
Muddensity
Rakearmtorque
Mudpump
Bedlevel
Waterturbicity
MV Mud bottom flow X X X X X X
MV Flocculant flow X X X X X X
DV Feed pH X X X X X X
DV Feed density X X X X X X
DV Feed flow X X X X X X

27. Optimization using model predictive control in mining | 27
CUSTOMER CHALLENGES
• Inconsistent production rates
• Loss of production at shift change
• Conveyor system trips
MPC BENEFITS
• Maximizing ore flow
• Operate to maximum
constraints
• Minimize equipment trips
• Extending equipment life
• Minimize interruptions
during shift changes
Material flow management
MPC | Route optimization:
Maximizing material throughput

28. Optimization using model predictive control in mining | 28
Material flow management
MPC
OBJECTIVES
• Maximizing ore flow
• Operate to maximum constraints
• Minimize equipment trips
• Extending equipment life
• Minimize interruptions during
shift changes

29. Optimization using model predictive control in mining | 29
Controller matrix
Controller consists of a matrix of
process model pairs that explain
IMPORTANT INTERACTIONS
in the process
PREDICT future values
of the CVs by movement of
all the MVs and DVs
PROACTIVE CONTROL
to coordinate MV setpoints
to minimize CV deviations
from targets, hence reducing
variability
CV1 CV2 CV3
MV1
MV2 NO MODEL NO MODEL
MV3
DV1
DV2 NO MODEL

30. Optimization using model predictive control in mining | 30

31. Optimization using model predictive control in mining | 31
How MPC generates
benefits REDUCES Variability
ACHIEVES‘Plant
Obedience’
MANAGES the
process within
constraints
ACHIEVES UPLIFT
– operate closer to
specifications and
performance limits
while maintaining
safety margins

32. Optimization using model predictive control in mining | 32
MODEL
PREDICTIVE CONTROL
Solution for Flotation Cell
at Latin American Iron Mine
PAVILON8 + VOA (VIRTUAL ONLINE ANALYZER)
STABILIZATION CONTROLLER QUALITY AND RECOVERY
OPTIMIZER CONTROLLER
Non-Linear Advanced Process Control
FLOTATION CONTROL
3%
REDUCED
Product Give-away
and Off-Spec
2%
INCREASED
METAL
RECOVERY
3%
REDUCED
REAGENTS
Reduced product give-away and off-spec of 3% = 18,000 tonnes

33. Optimization using model predictive control in mining | 33
ACCESS
• Propose and Plan
• Confirm Business
Value
• Set Expectations
• Determine
Benchmark Metrics
DELIVER
• Design, Develop,
Deploy
• Data Gathering and
Validation
• Model Development
• Application
Deployment
AUDIT
• Commissioning
• KPI Configuration
• Training
• Performance
Validation
SUSTAIN
Value-Based Proposal Value-Based Solution Measure Value Ongoing Value

34. Optimization using model predictive control in mining | 34
Level 1 Support
Email and live
telephone support
MPC System recovery
assistance from
server failures
Service-Pack Releases
provide updates and
system changes

35. Optimization using model predictive control in mining | 35
Level 1 and Level 2 Support
Email and live
telephone support
MPC System recovery
assistance from
server failures
Service-Pack Releases
provide updates and
system changes
Quarterly Pavilion8
MPC status reporting
and troubleshooting
APC application issues
Web-based Pavilion8
Suppost Knowledgebase
and annual onsite visits
available for APC

36. Optimization using model predictive control in mining | 36
Scalable Support Offerings Level 1 Level 2
Major and Minor
Service-Pack Releases
Included Included
Email and Live Telephone
Support
Product Support Product and Application Support
MPC Recovery from
Server Failure
Assisted System Recovery Comprehensive System Recovery
Application Support
5 Hour Reactive Support
(Remote)
Unlimited Reactive and Proactive
Support (Onsite Availability)
System Health Check
and Reporting
Remote Annual System
Health Check
Quarterly Full System Performance
and Health Report
MPC System Back-ups Pavilion8-based Applications
Refresher Training Remote or Onsite Available
Our sustained value services help protect your investment

37. Pavilion, Pavilion8, PlantPAx, Integrated Architecture, Listen. Think. Solve., are trademarks of Rockwell Automation, Inc. Trademarks not belonging to Rockwell Automation are property of their respective companies.
Rockwell Automation, Inc. (NYSE:ROK), the world’s largest company dedicated to industrial automation, makes its customers more productive and the world more sustainable.
Throughout the world, our flagship Allen-Bradley® and Rockwell Software® product brands are recognized for innovation and excellence.
Follow ROKAutomation on Facebook, Twitter and Google Plus. Subscribe to us on YouTube. Connect with us on LinkedIn.
Publication MIN-BR001A-EN-P January 2016. Copyright © 2016 Rockwell Automation, Inc. All Rights Reserved. Printed in USA.