Model Predictive Control for mining
Transcript of Model Predictive Control for mining
Model Predictive Control for miningProcess optimization solution
SOURCE: World Bank Group
Minerals and metals play a central role in the global economy and will continue to provide the materials the world needs.THE WORLD IS SEEING INCREDIBLE CHANGES that require resource-intensive
goods of all sorts. Overall population is growing and shifting to urban areas,
the middle class is expanding and infrastructure development is increasing.
With this growth comes a need for resources, meaning metals like iron, copper,
aluminum and nickel will come with a much higher demand.
A new report from World Bank Group states that the production of minerals,
such as lithium, graphite and cobalt, could increase by nearly 500% by 2050, in
order to meet the growing demand for clean energy technologies. The report
estimates that more than 3 billion tons of minerals and metals will be needed to
deploy wind, solar and geothermal power, as well as energy storage, required for
achieving a below 2°C future.
Mining powers the world
2The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
Today’s process characteristics Mineral processing plants are inherently complex, facing internal and external disturbances, many recirculating loads and multiple interactions. And the declining of the ore grades and continuous variability of raw material are pushing the limits of processing plants even more.
Process objectives These process characteristics that miners are dealing with today provide opportunity to improve:
• Reduce cost per ton
• Lower plant costs
• Lower energy intensity
• Improve energy and water consumption
• Extend the life of existing ore bodies
• Continue to meet regulatory requirements
• Overcome lack of experienced and trained instrumentation technicians, operators, metallurgists
How will you achieve these process objectives in such complex operations?
Mineral processing challengesOre grades are declining and becoming more complex
33Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
3The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
Although traditional optimization strategies provide adequate control in terms of plant safety, it rarely achieves optimal control of quality and economical, efficient operation.
Why? Mineral processing plants face inherent complexities that can’t be addressed through traditional technologies:
Multiple constrains
Time delay processes
Multiple variables
Key variables difficult to measure
Other potential technologies, such expert systems, can address some of these challenges but with price tag:
of mineral processing plants still use basic
optimization strategies
MOR
E THAN
75%
HIGH MAINTENANCE
HARD TO COMPREHEND
MODEL OPERATORS (INSTEAD OF PROCESSES)
4Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
4The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
MPC solution benefits for mining
INCREASED throughput
LOWER reagents consumption
BETTER recovery
OPTIMUM use of water and energy
IMPROVED process stability
REDUCES variability
ACHIEVES “plant obedience”
MANAGES the process within constraints
ACHIEVES UPLIFT - operate closer to specifications and performance limits while maintaining safety margins
Handling constraints, maximizing economics, and maintaining stable processPavilion8® Model Predictive Control (MPC) from Rockwell Automation reduces process variability and enhances stability over and above what is currently possible with more traditional control schemes. This is accomplished through our multi-variable, nonlinear, model predictive control capabilities.
Maximum profitability is realized by operating a process as close to its constraints as possible while maintaining an appropriate margin of safety.
Pavilion8 Model Predictive Control
KEY
PROD
UCT
PROP
ERTY
PER
CEN
T OF
LIM
IT
Process/specification limit
TIME
MPC ONBEFORE
Variability under operator control
Variability reduction Push process toward limits
100
98
96
94
92
90
5Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
5The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
Controller / optimizer: Pavilion8 Model Predictive Control (MPC) optimizes the process making predictions about future plant outputs responding to changes in the process input variables and disturbances.
The model: In order to operate properly, we need an accurate model of the process. Pavilion8 MPC leverages best in class machine learning technologies to build robust models of the process incorporating and combining all available knowledge of the process, such as historical data, process equations, data from plant tests and operator knowledge.
Virtual online analyzer: The efficiency of the optimization tool is related to the richness and quality of the field data. However, in most of the process some measurements are unprecise, not present, or delayed (for example data from laboratory). In this instance, Pavilion8 MPC uses an innovative virtual online analyzer that’s able to estimate such variables closing the gap of this information gap.
The console: A powerful visualization tool allowing operators to visualize the process and model KPIs for better situational awareness of the system.
Pavilion8 Model Predictive Control architecture overview
LABORATORY
PROCESS
PROCESS KNOWLEDGEPLANT TEST DATAHISTORICAL DATA
MODELING, IDENTIFICATION, ANALYSIS
VIRTUAL ONLINE ANALYZER CONTROLLER/OPTIMIZER
CONSOLE
PLANT CONTROL SYSTEMS
66Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
6The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
The controller matrix shows dynamic relationships between process outputs and inputs. The system predicts future values of the outputs by movement of all the inputs including manipulated variables and disturbance variables.
Controlled variables (CVs) Process variables to maintain at a target or within a range (can be considered outputs)
Constrain variable (CCVs) The state is forbidden to penetrate or may have physical limitations
Manipulated variables (MVs) A manipulated input is one that can be adjusted by the control system (or process operator)
Disturbance variables (DVs) Disturbance variables - these are also called “load” variables and represent input variables that can cause the controlled variables to deviate
The controller matrix
CV3 MV3
CV1CV2
DV1DV2DV3
MV1MV2
MV4 MV5DV4
CV4CV4
Illustrative example of a flotation
matrix model
MV1 MV2 MV3 MV4 MV5 DV1 DV2 DV3 DV4
Collector reagent
flow ratio
Depressor reagent
flow ratioTrailings
value Air flowFrother reagent
flow ratio
P80 from grinding circuit
Feed flow from grinding
Feed density
from grinding
pH
CV1 Concentrate grade
CV2 Metal recovery
CV3 Froth depth
CV4 Bubble speed of air
hold up
CV5Bubble size distribution or surface
area
77Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
7The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
PROCESS CHALLENGES
• Balance between the individual crushers and the overall circuit goal
• Maintain circuit efficiency
• Ensure that crushers are choked
• Crusher capacity vs. feed
• Equipment wearing
• Power consumption
RESULTS
• Maximized crusher circuit throughput
• Process stability
• Optimum energy efficiency
• Decrease equipment wear
• Fewer trips over current, high-level silos
Crusher circuits
MOD
EL P
REDI
CTIV
E CO
NTR
OL
Belt 2
Belt 3
2nd cone crusher
From primary crusher
Feed bin 1
Feed bin 2
3rd cone crusher
Belt 5
Belt 4
Belt 1
Screen 1
Screen 2
MV Speed
MV Speed
MV Speed
MV Speed
CV Level
CV Level
CV Level
MV DV CV CCVManipulated Variable
Upper & lower Constraint
Disturbance Variable
Upper Constraint
Controlled Variable
Lower Constraint
Constraint Variable
Maximize Target
T OilCCV
T OilCCV
Weight, Level, Speed, Position, Current
CCV
Weight, Level, Speed, Position, Current
CCV
Weight, Level, Speed, Position, Current
CCV
Weight, Level, Speed, Temperature, Position, Current
CCV
MV Speed
MV Speed
MV Speed
CV Level
8Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
8The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
POTENTIAL GAINS: Up to 10% increased throughput, 10% on reduced specific energy, with a P80 variability reduction of up to 50%
PROCESS CHALLENGES
• Complex behavior
• Process disturbances
• Relationship between variables isnon-linear
• Energy waste due to overgrinding
• Requires operators with extensiveexperience
• Balance between throughput anddownstream constraints such asoptimum particle size
RESULTS
• Maximized throughput rate
• Decrease energy/ton
• Increased stability
• Reduction in particle size variation
• Delivers grinding circuit stabilityto achieve throughput and grindsize targets
Comminution / grinding circuits
MOD
EL P
REDI
CTIV
E CO
NTR
OL Cyclones
Number of hydrocyclone
Ball mill
Flotation feed
Water
Fresh feed
Cyclone feed tank
CrusherSAG MILL
MV Flow
MV Flow
MV Flow
MV Speed
DV F80
CV Density
CV Sump level
CV P80
MV DV CV CCVManipulated Variable
Upper & lower Constraint
Disturbance Variable
Upper Constraint
Controlled Variable
Lower Constraint
Constraint Variable
Maximize Target
CCV Power
CCV Pressure
PowerCCV
DensityCCV
FlowCCV
MV
DV
9Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
9The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
DV
CV Weight
CV Density
DV Flow
PROCESS CHALLENGES
• Complex process that is notfully understood
• Still relies on operator control
• A small disturbance in one processvariable propagates to the finalprocess output
RESULTS
• Increased recovery
• Increased and stabilizedcleaner grade
• Optimized reagent consumption
Flotation
MOD
EL P
REDI
CTIV
E CO
NTR
OL
POTENTIAL GAINS: Increased recovery of up to 3%, reduction of reagents of up to 3%
MV DV CV CCVManipulated Variable
Upper & lower Constraint
Disturbance Variable
Upper Constraint
Controlled Variable
Lower Constraint
Constraint Variable
Maximize Target
MV Air
MV Collector
MV Valve position
MV Depressor
MV Modifier
MV Valve position
DV Feed flows
DV P80
DV Ore grades
CVMass/metal recovery
CV Bubble size
CV Bubble speed
CV pH CV Concentrate grade
CV Cell levels
10Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
10The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
MV Air
DV Ore Mineralogy
Tailings
Concentrate
MV Frother
PROCESS CHALLENGES
• Water conservation
• Optimum reagents consumption
• Little control and local knowledge
• Not often integrated into controlsystem
• Conventional control isinefficient (long residence time,large disturbances and non-linear behavior)
RESULTS
• Increased underflow density
• Decreased flocculantconsumption
• Improved overflow clarity
• Better water utilization
Thickeners
MOD
EL P
REDI
CTIV
E CO
NTR
OL
POTENTIAL GAINS: Increased water recovery of up to 3%, reduction of flocculant of around 2%
MV DV CV CCVManipulated Variable
Upper & lower Constraint
Disturbance Variable
Upper Constraint
Controlled Variable
Lower Constraint
Constraint Variable
Maximize Target
MV Underflow
DV Feed density
DV Feed pH
DV Feed flow
CV Mud density
Rake arm torque
CCV
CCV Turbidity
Thickener overflow
11Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
11The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
CV Underflow pump amp
DV Rake height
CV Bed pressure
MV Flocculant flow
PROCESS CHALLENGES
• Inconsistent production rates
• Loss of production at shift change
• Conveyor system trips
RESULTS
• Maximizing ore flow
• Operate to maximum constraints
• Minimize equipment trips
• Extending equipment life
• Minimize interruptions during shiftchanges
• Energy conservation andmaintaining the maximumthroughput
Material flow management
MOD
EL P
REDI
CTIV
E CO
NTR
OL
POTENTIAL GAINS: 20% increased throughput
1212Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
12The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
PROCESS CHALLENGES
• Product under specified grade
• Change in feed characteristics
• Final product variability
• Temperature control
RESULTS
• Stabilization of the operation
• Improved quality
• Product consistency
• Energy savings
• Reduce emissions
Material refining
MOD
EL P
REDI
CTIV
E CO
NTR
OL
1313Solutions to meet steel needsPG 2
The connected steel plantPG 5
Increase operational efficiencyPG 9
Knowledge driven operationsPG 14
Modernize andempower yourworkforcePG 19
Design, build and upgrade with confidencePG 22
13The global mining marketPG 2
Model predictive control solutionsPG 5
Model predictive control in mining applications PG8
Publication MIN-BR007B-EN-P January 2021
Copyright © 2021 Rockwell Automation, Inc. All Rights Reserved. Printed in USA.
Expanding human possibility, Pavillion8, and Rockwell Automation are trademarks of Rockwell Automation, Inc. All other trademarks are property of their respective companies.
Expanding human possibility is a registered trademark of Rockwell Automation. Alphabetically list all Rockwell Automation trademarks used in the publication. Positioning should be approximately 0.3” (7.62mm) from above list of locations and from below publication number.
Font: Barlow Condensed, Regular, 8 pt., 80% black, centered.
Publication ENCOMP-QR004I-EN-P - Month 20XX | Supersedes Publication ENCOMP-QR004H-EN-P - Month 20XXCopyright © 2020 Rockwell Automation, Inc. All Rights Reserved. Printed in USA.