Implementation of an Expert Control and
Online Image Based Oversize Detection System
at the Pilbara Iron Tom Price Operations
D. Morgan, J. McKay, D. La Rosa, M Wortley, O. Guyot
Tom Price High Grade Process Plant
Three Stage Crushing and Screening Plant
Designed to produce two product size fractions:
Lump Product: 31.5 mm + 6.3 mm
Fines Product: -6.3 mm
Capable of processing up to 4000 tph
2285 t / DOH 2004 average
Typical product split ranges from 45-65 % Lump
58% 2004 average
Annual Production
12.6 Mt SOP 2004
Design
15 Mt SOP per Annum
Tom Price High Grade Process Plant
Why an Expert Control and Vision System?
Increasing Demand for Iron Ore
Conservative Operating Strategies
Reprocessing Misplaced Oversize costs lots of money
Opportunity to vastly improve Process Automation
Gives Analysis of Bottlenecks to target upgrades more effectively
Myth: Crushing plants
are too simple to
benefit from Advanced
Process Control
OCS
SET POINTS / DATA / STATUS
HMI ALARMS
(VOICE, EMAIL ETC)
COMMUNICATIONS
EXPERT SYSTEM VIA
DCS OR PLC
PROCESS VARIABLES
ADAPTIVE
NEURAL
STATISTICS VISION PREDICTIVE
NETWORK
MODELS
ACOUSTICS OPTIMISER
TRENDS
&
EVALUATION
SENSORS
OCS
If then rule structure
Crisp and Fuzzy Logic
Modules used in this installation
Expert
HMI
Statistics
Vision
OCS Control Strategy - Objective
Maximise plant throughput by optimising Dynamic Constraints
Avoid Emergencies due to Exceeding Dynamic Constraints (Max
Equipment Utilization and thus Production and Minimize cost from
equipment damage)
OCS Control Strategy Process Variables
Amps - Crusher, Conveyors and Screens
Production Limits -Weightometers
Bin Levels Screen house and Tertiary Crusher
Bin Availability
Run Status -Feeders, Crushers, Screens and Conveyors
OCS Control Strategy - Manipulated
Variables
Secondary Crusher Feeder outputs
Tertiary Crusher Feeder outputs
HGSP and Tertiary Bin Availability/Feeder outputs
Implementation of Strategy
Strategy constructed of easy to read crisp and fuzzy If Then
rules combined with equations
Equations calculate relative values, ratios limits and dynamic
constraints
When in Expert Control Strategy controls
1. secondary feeder outputs,
2. tertiary bin availability/feeder outputs and
3. HGSP bin availability and feeder outputs
How do we measure its effect?
DCU - Dynamic Constraint Utilization
Formulated by Rio Tinto Asset Utilization
Measurement of how close a process is to its current constraint
True measurement of circuit performance independent of
feed/circuit changes
Constrained System - any system that functions by pushing
against process limitations such as motor amps or bin levels
Dynamic Constraint Utilization
Dynamic Constraint Units normalized constraining values for
comparative use
DCU = 100 * present value/Constraining Limit
% DCU - Calculated as the maximum of all individual DCUs
(dynamic constraining units) such as (tph, amperage, level)
The higher the %DCU the better circuit performance and thus the
best economic performance
Expert System Results 8% Increase in DCU
160
140
120
100
DCU %
80
60
40
20
0 500 1000 1500 2000 2500
Time (Hours)
Challenges of Expert Control
Time Delays between Process Variables Difficult to tune rules
All plant instrumentation must be kept in top working order and
calibration Difficult with number of Electricians available
Control board operators culture is difficult to change
Different feed types change constraints
VisioRock
VisioRock
Installation
Analogue Cameras 1/1000th or 1/2000th shutter speed
Initial Installation 6 Cameras for Oversize Detection
Fibre Optics back to control room
USB Frame Grabbers to convert video signal to digital video
stream
VisioRock
Screen Capture of VisioRock processing 13M Lump Product
VisioRock Strategy
Calculates Size distribution on each belt derives percent oversize
Generate alarm when outside normal operating limits
Operators sent to confirm by visually inspecting belt
Trends misplacements (Fines in Lump, Lump in Fines, Oversize in
Lump) on each belt allowing much better feed rate control in screen
house
VisioRock Results
Variable results to date mainly due to sampling rate for each camera
6 Cameras being processed sequentially over a one minute cycle
(10 seconds each)
Many false alarms
Missed oversize events
Accurately controlling screen house feed rates based on Fines in
Lump
Current and Future Direction
Expert System running well
Now includes low grade plant Running extremely well however no
results to date
Separating 6 VisioRock Cameras to increase sample rate
Use Count of particles strategy to detect Oversize
Cameras installed on Crusher feed conveyors to enable automatic
crusher gap setting
Integration of Plant and truck dispatch system will enable feedtype
modelling of constraints
Myth: Crushing plants
are too simple to
benefit from Advanced
Process Control
