Developments in Process Control

With particular reference to comminution and flotation

(McGill Professional Development Seminar Mineral Processing Systems) by P. Thwaites, May 15th, 2009

Xstrata Organizational Structure

Xstrata Executive Committee

X Ni

X Cu

X Coal

XTS

X Alloys

X Zn

XPS

XT

XTS : Xstrata Technology Services Thras Moraitis - London XPS : Xstrata Process Support Frank McGlynn - Sudbury XT : Xstrata Technology Joe Pease - Brisbane

Objectives of McGill Short Course

The objectives of the Mineral Processing Systems, Professional Development Seminar are: To cover new and important developments in mineral processing in the areas of comminution, flotation, process control, environment and optimization. Developments in Process Control, with particular reference to comminution and flotation:
New sensor technologies, control strategies, optimization etc.

Opportunities to consider
Operational Performance Excellence requires a solid performance of the regulatory layer AND process optimisation. For Plant Operators: Is your feed stable? Are your instruments calibrated and performing? Are you aware of wireless instruments (including vibration)? Is your control system up to date and stable? Are you in manual or auto control? Are your Operators acting on alarms or are they nuisance? Do you understand and accept your process variability? Are you operating within the design targets and process constraints (pumps, cyclones, supplies, roasters, furnaces etc.)? Are you using your surge capacity, . or running tight level control? Are you at optimum and are the controls robust? Are you benefiting from asset management systems? Are failure / fault detection systems implemented? Can you make the same product for less energy consumption? (P. Thwaites, AUTOMINING2008, Chile)
4

Downtime Reporter (Matrikon)

Automatic downtime recognition with direct data from Process Control System

Plant Equipment

PLC/DCS

(Matrikon) Downtime Reporter: How do we improve OEE*?

Improve Availability
Downtime Pareto report clearly identifies the big-ticket items causing the most downtime in the plant.

Strategies can then be formulated that eliminate these problems and thus improve availability.
* OEE Overall Equipment Effectiveness = Availability x Performance x Quality i.e. 100% = 100% x 100% x 100% OEE defines the expected performance of a machine, measures it and provides a loss structure for analysis, which leads to improvement.

Definition of Process Control

(McKee, AMIRA P9L) Process control is a broad term which often means different things to different people. Process control is considered as the technology required to obtain information in real time on process behaviour and then use that information to manipulate process variables with the objective of improving the metallurgical performance of the plant. Control for the purpose of process improvement.

Importance of Control Performance (Emerson)

Audits

Audits are the first step to ensure control system investments meet expectations
(Feb. 1999 CONTROL ENGINEERING Journal, by Dave Harrold) Shown are the defects per loop e.g. Filter Time Constant Integral Period Derivative Period Sampling Period Proportional Band Current Operating Mode
45% 40% 35% 30% Frequency 25% 20%
16% 39% 34%

15% 10% 5%
0%

7% 4% 0%

0% 0 1 2 3 4 5 6 Number of Defects per Loop

XPS PC Grp Capabilities & Services

XPS PC Group
1. www.myxps.ca

Process Design & Commissioning

2.
3. 4. 5. 6. 7.

Controls Auditing
Control Loop Optimisation Advanced Controls Slow Process Response Off-Gas System Controls Grinding Controls

Process Control Will Not Correct Inherent Design / Flowsheet Problems

(McKee, AMIRA P9L) There is a need to determine, and if necessary correct, the condition of the plant as a pre-requisite to control development. A good example is the importance of classifier operation and its effect on comminution circuit performance. Techniques exist (plant sampling, modelling and simulation) to audit the actual plant operation. Correcting plant limitations should be seen as a first step in the control approach.

11

XPS : Process Support Groups

Process Mineralogy - Design, implement and optimize mineral processing flowsheets by matching the flowsheet to the mineralogy. Process Control - Identify and deliver robust process control technology and engineering solutions to achieve Operational Performance Excellence. Extractive Metallurgy Provide specialized extractive metallurgy services (hydro-and pyro-metallurgical). Flowsheet/project development using modeling and piloting, new process development and plant optimization. Materials Technology - Improve the reliability of critical equipment through appropriate implementation of well proven materials engineering practices at essential stages of design, procurement and operation.

12

XPS Process Mineralogy

Sampling and Statistics

Mi n

er al

Sc ie

nc e

PROCESS MINERALOGY

al er in M si es oc Pr ng

13

XPS Process Mineralogy Crushing and Blending Plant

Staged crushing and screening prevents overproduction of fines Drill core or ROM 150mm rock at 150 kg/hr Blending technology produces RSD of < 5% in test charges

Any product size down to 1.7 mm split into any unit mass Semi-continuous operation
14

XPS Process Mineralogy Mineral Science

Developing Flowsheets that Work the first time

Modern Quantitative Mineralogy

Strategic Virtual Flowsheeting/Flowsheet Implications

15

XPS Process Mineralogy Mineral Processing

Developing Flowsheets that Work the first time

High-Confidence Flotation Testing*

Established 1995 95% confidence level Tried and tested Reproducible results Reliable scale-up * Lotter, N.O., 1995, A Quality Control Model for the Development of High-Confidence Flotation Test Data, 16 M.Sc. Thesis, University of Cape Town, June 1995

XPS Process Mineralogy Mineral Processing

Mini-Pilot Plant Demonstrating the Optimised Flowsheet

Commissioned 2005 11 campaigns to date Reproducible results Proven ability to mimic operations

17

XPS Process Mineralogy Montcalm Project

Flowsheeting from Drill-Core to Pilot Plant
Comparison of Montcalm Start-up Curve with McNulty Curves 120

Finds optimum flowsheet, or finds best performance attainable with known flowsheet Montcalm Project - Type 1 Startup*

Montcalm start-up

Type 1 Type 2 Type 3

Ni output in Ni and Cu conc,% of design

100

80 Type 4 60

40

20

0 0 2 4 6 8 10 12 14
Montcalm start-up - October 2004

Quarter after start-up

*Type 1 reaches design capacity ~4 quarters

18

Good Process Control Implementation

(McKee, AMIRA P9L) A good implementation requires a well defined operating strategy and an associated control strategy.

19

Effective Process Control Cycle (for New Processes / Plants)

Deliverable (Stage) Process Opt. (Production) Process Flowsheets & Control Philosophy (Development)

Overall Process Control Objective

Measure Control Optimise
constraint

units

As-builts (Commissioning)

higher production
setpoint

lower costs better control best control

P&IDs (Basic Eng)

poor control

time
Falconbridge Limited - Process Control

Control Config. (Construction-EPCM)

Logic Diagrams (Detailed Eng-EPCM)

20

Overall Process Control Objective

Measure

Control

Optimize
Process constraint

higher production

units

setpoint

lower costs better control

poor control

best control

time
21

Poor to Optimized Control

Best Practise: Necessary for: Operational Performance Excellence

22

5 Essential Control Loop Elements & Consider the Whole Loop (ABB)

Five Essential* Control Loop Elements:

1. 2. 3. 4. 5. Sensing element Transmitter Controller Final Control Element (e.g. Actuator or VSD) Process

Consider the whole loop

PC-program in

Advant OCS
Ramp rate Filter Deadband D/A
4-20 mA

A/D I/P
4-20 mA

3-15 psi

Filter
mV

Positioner
0-6 bar

Filter FT

Actuator

Only when all five elements are performing their best will the control system meet expectations!
* February 1999 CONTROL ENGINEERING by Dave Harrold.

ABB Corporate Research

K. Forsman, 1998, No. 10

ABB

23

Regulatory Control improvements (AG Mill Bearing Pressure measurement) Careful: 1) Appropriate filtering; 2) PI Data compaction.

24

Mill Feed Manual Setpoint Control

Opportunity in optimising feed tonnage has been estimated at $8.8m/yr!

25

General Process Control Hierarchy

Objective
Site Plant Optimization
Plant

Function
Cash Optimization Economics

Optimization
Advanced
Process

Optimize

Optimizing Control
Loop Control

Stabilize

Regulatory Manual
26

Field / Panel / DCS / PLC Instrumentation - Inputs / Outputs

Measure

Economic Return

Processes

Era of Quantity

Objectives
Tonnage of Concentrate

Control Technology
PLC or DCS
The evolution of control can be summarized in this slide. The first shift was from quantity only to quantity and quality, both applied in a DCS or PLC.

Quality

Tonnage + Quality Client Satisfaction Optimizing Control Systems OCS

Peak Performance

Tonnage + Quality + Peak Economic Performance

However, to shift from maintaining quality to peak performance requires something more than a DCS or PLC: an optimizing control system.

Metso minerals - Jan. 2004 Technology Presentation

27

Enabling Technologies - ABB (Overall Process Unit Control)

Production cost
Lower Knowledge Based Expert Control

MPC (Model Based) Control: Several tools are available:

Mintek (FloatStar) Emerson (DeltaV MPC) ABB (Linkman, Expert Optimizer) Invensys (Connoisseur) Honeywell (Profit Suite) Gensym (G2) Prediktor Metso (Adaptive Predictive Model)
Higher Feedback Lower Feedforward & Cascade

Multi Variable Techniques

Cross-Coupled

Higher

28

Connoisseur (Invensys) Overview

Neural Net Adaptive Cont
LP Optimizer Connoisseur Environment

Non-linear

Fuzzy Logic Inferential P Optimal Setpoint Director Calc M & MV Target

Model Predictive j Controller
Constraint Management D304 349 * L10 (LR or QP methods)

454 + S1 x + d

CVs, MVs & DVs

Regulatory Control System
29

Elements Necessary for Successful Process Control in Mineral Plants

(P. Thwaites, IFAC MMM07, Plenary Address, Quebec City)

Tools: Instruments; Systems etc.

Process Control -> Operational Performance Excellence

People: Control / Process Knowledge

Actions: Support Management; Technology Transfer

Successes: Results & Examples

30

Enabling Technologies

(McKee, 1999). In grinding, control of AG/SAG mill circuits is the dominant area. While some systems have emerged which provide a reasonable level of control, there is still much not understood about the dynamic behaviour of these mills, and there is considerable scope for further development.

1. Multivariable Controller in a PLC Function Block (Bartsch): SAG feed control, is generally done using an Expert system. These systems often deliver improved control and 4 to 5% increased throughput (e.g. Collahuasi and Raglan Mills); XPS has recently implemented a complex controller in a (Concept) control block, negating the costs of an auxiliary Expert System, and training / support of this system.

31

Entire Grinding Controls Focus: (Liberation efficiency and throughput)

Wipfrag (Size analysis) 21-CV-04 ASRi Sandvik Cone Crusher Hydrocone H4800 (250 kW) 21-CV-06 150 mt Surge Bin 11-CV-03 Underground Storage Bins 21-CV-01
Cyclone Feed Density Control

Flotation

21-CV-05

SAG Mill 24 diameter (2240 kW)

Cyc CycO/F O/F Density Density by by P/Boxwater water addition P/Box addition
DI

PI PIC
RSP

DIC

OUT

CONSTRAIN circulating load by trimming Cyc Feed Density SP

LI LIC
OUT Select Lo g ic

SY DI
H/L L im

FI

MIC

OUT

H/ L Lim

ADJUST FEED SP OR CYC FEED DEN SITY SP

DIC
OUT

Cyc Feed Density by trimming O/F density SP

Ref: FTC Report Raglan: Cyclone De nsity Control, E. Bartsch, 11 Novemb er 2005

FFE Impact Meter

Metso Double Deck Vibrating Screen (8 x 16) 1- 8 x 40 mm 2- 5 x 4 mm, 3 x 8 mm

21-CV-02

6 x 15 Krebs GMAX Cyclones Vort: 4.5, Apx: 3

Ball Mill 14 dia. x 21 (2240 kW)

32

SAG Feed Rate Control (PID)

SAG Control - December 27, 2005
180
Se tpoint = 4450 k Pa

4500

170 160 4450

Tonnage - mtph

140 130 120 110 100

04:48:00 09:36:00 14:24:00 19:12:00 00:00:00 04:48:00

4400

4350 Tonnage kPa 4300

09:36:00

Time

In Cascade control (constant kPa), The SAG tonnage varies tremendously to maintain the requested setpoint.

Average Std Dev. Minimum Maximum

SAG Feed SAG Load mpth kPa 131.3 4 454 9.7 24 100.5 4 392 152.7 4 503

SAG KPA

150

33

Why Online Feed Size Distribution?

Mill Load & Power INCREASING Mill Load & Power DECREASING Bearing Pressure
PRI-21WIC0204.SP 130.00000 Mtph PRI-21ML01.AV 2163.43750 kW PRI-21PIT0457C.AV 4363.49463 kPa PRI-21VT28.D75.AVG 0.04566 m

Plot-0 200 2500 4500 0.085 180

160

140

120

Mill Feed Set-point

Mill Power

100

80

60

40

20

Large size fraction (D75) INCREASING

0 1600 0 0.035 9/29/2002 9:30:00 AM TONNAGE ALIM.BROYEUR AUTOG PUISSANCE BROYEUR AUTOGENE PRESS. D'HUILE PALIER DECH Alimentation Moulin AG D75

Large size fraction (D75) DECREASING 12.00 Hour(s)

D75 Size Fraction

9/29/2002 9:30:00 PM

34

SAG Charge Multivariable Fuzzy Controller (Bartsch, CMP 2007)

Inputs
(Measurements)

Outputs
(Set-points)

Power Charge Impactmeter Granulometry (prediction !) Assays

Fuzzyfication

Feed Rate Density (water addn)

Crusher Gap

Fuzzy Rules

De-fuzzification
Programmed in existing plant PLCs

Mill Speed

35

ASRi (Automatic Setpoint Regulation) Sandvik Technology - Crushing Plants

Crusher Gap Control: Kidd Mill -> Strathcona Mill -> Raglan Mill

Field Device

DCS Display (via OPC)

36

Cyclone Feed Density Control

Mill Objectives *:
1. 2.

Increase the fineness of the grind to P80 = 75 Reduce Cyclone Feed Density to improve classification efficiency.

* Ref: Report Raglan Optimization Project : Raglan Site Visit L Urbanowski, Jan 14 2006

37

Cyclone Feed Density Control

Typical Control Strategy:

Water addition monitored / controlled to the pumpbox?:

LI

LIC

SY

DI DIC

38

Cyclone Feed Density Control ..or Cyclone Overflow Density Control ?

Deficiencies in existing Strategy:

Controlling cyclone feed density directly by water addition not considered best practice.
1. 2.

The feed density is slow to respond which means the loop cannot be tuned for load disturbance The influence of adding water changes depending on your mill capacity: A. spare grinding capacity it will trim the feed density as expected. B. at or over capacity -; it will increase circulating load and density, resulting in more water addition .

39

Cyclone Feed Density Control

Requirements of a Density Control Strategy

1. 2. 3. 4.

Control water addition by using the more responsive process variable (overflow density). Consider the circulation load to avoid overloading the ball mill. Maximize circulating load in order to maximise mill efficiency. Filter measured variables appropriately (i.e. match the process response times)

Ref: FTC Report Raglan: Cyclone Density Control, E. Bartsch, 11 November 2005

40

Cyclone Feed Density Control (Best Practice)

1
DI PI

1.

Control water by using the more responsive process variable (Cyclone Overflow Density). Use the output of the Cyclone Feed Density controller as the remote (ext.) set-point for Cyclone Overflow Density controller.
(This cascade is more suited to the slow dynamics of the loop.)

2. PIC
RSP OUT

DIC

LI

LIC

OUT

Select Logic

3
SY DI
H/L Lim

3.

FI

MIC

OUT

H/L Lim

ADJUST FEED SP OR CYC FEED DENSITY SP

OUT

DIC

Use a Circulating Load controller to either trim Feed Rate or Cyclone Feed Density setpoint. N.B. feed will
need adjusting if circulating load does not reach steady state.

2
Ref: FTC Report Raglan: Cyclone Density Control, E. Bartsch, 11 November 2005
4.

Filter appropriately

41

Raglan Cyclone Feed Density Control

Ref: FTC (XPS) Report Raglan: Cyclone Density Control, E. Bartsch, 11 November 2005

Controller placed in 700 service

600

130.0

120.0

% Circ
500

110.0

Load
400

100.0

90.0

P80 90
Circ Load Circ Load SP D80

microns

300

80.0

P80 80
microns

70.0 200 60.0 100 50.0

P80 68
microns

0 1 3 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 35 37 39 41 43 45 47 49 51 53 55 57 59 61 63 65 67 69 71 73 75 77 February March

40.0

Reduced* p80 by 10 microns = 0.6% Ni rec increase worth approx. $2,000,000 pa.
(* ref feasibility study benchmarking 2004)

42

Oscillating on/off Cyclone Switching

=> Oscillations in Cyclone AND Process Feed rates

- Cyclones (often) operating below their design pressure (40-60 instead of 100 kPa): Increased short-circuiting of water & undersized particles; IsaMill feed density decreases below design value; Low IsaMill feed density may create pumping issues (& trip on low flow/pressure); - Cyclone pressure should be controlled using the cyclone feed-rate. 43 - Implement Surge Tank Control (using 150 m3 of tank) NOT Tight Level Control.

Strathcona Mill Regrind Cyclone Overflow Density Results - 3 days operation (including startup)
Plot-0 350 50 CGCYCRBmv~C_PIC_2057:B_PID2057.MEAS 192.74 kPa CGCYCRBmv~C_DIC_2060:B_AIN2060_1.PNT 29.0277 % solids

300

250

200

Cyclone Feed Pressure Cyclone Overflow density

1/21/02 7:16:18 PM

150

100

Overflow Density:
50 0 20 1/18/02 7:16:18 PM B REG CYC PRESSURE B REG CYC O/F DENSITY 3.00 Day(s)

Plot-0 350 50

9% density improvement 65% reduction in standard deviation

300

Improved startup

CGCYCRBmv~C_PIC_2057:B_PID2057.MEAS 148.07 kPa CGCYCRBmv~C_DIC_2060:B_AIN2060_1.PNT 38.3157 % solids

250

200

150

100

50

20 0 3/21/02 7:16:18 PM B REG CYC PRESSURE B REG CYC O/F DENSITY

3.00 Day(s)

3/24/02 7:16:18 PM

44

Enabling Technologies
2. On-Demand Sampling Automation:

Excellent practise is an OPPORTUNITY in our Canadian Mills!

45

Enabling Technologies
3. Camera Imaging, flotation level & reagent controls

Feed size analysis; Froth camera imaging; Optical System for cathode quality:
CSQA: Cathode Surface Quality Analyzer (Aplik)
Froth Camera Imaging Technology

39

46

Float to the Top Energise your Flotation Performance

Authors: A Okely, A Rinne, A Peltola (Outotec)
The relevant cost factors for a flotation plant are investment, energy, reagent consumption, and maintenance. The chart below shows the breakdown of these factors, based on typical ownership costs of a large mechanical flotation machine (100-200 m3) over a 25 year lifespan.

If we look at the energy efficiency we find that three aspects are critical: 1. Air dispersion 2. Rotational speed of the mechanism 3. Component wear
Optimal air dispersion is one of the basic requirements for good metallurgical performance. Plants operating with forced air cells have often noticed that the best results are achieved using individual and varying air feed rate in each cell.

47

Canty Vision Froth Camera (Strathcona Mill)

48

Flotation Level Controls

Level Control:
Absolute basics for Flotation. The movement of the pulp level setpoint provides opportunities for flotation optimization if there is tight control in each cell. (Ref., PT, 1983)

49

Rougher Level Disturbance

Level Setpoint

Froth Velocity

1. Velocity (mass pull) changes by over 1000%! 2. Baseline pull is only 0.5 1 cm/s! : Level SP too low?
50

Escondida Froth Velocity is Cascaded to Control Flotation Cell

(metso minerals & Ramon A. Brito Minera Escondida)

SP Aire

SP Velocidad

PV Velocidad

Escondida: Instalacin en la Flotacin Primaria Lneas 1,2,3,4,5,6 (54 Cameras)

Camara

SP Nivel

Tapon

Celda WEMCO

51

Rougher flotation
Test Control strategy step1

52

Froth velocity control

Feed grade

Froth veloc.

Level SP

Air SP

Tails %Ni
53

Flotation Control: Points on Control Structure and importance of layering (Gilles)

Set Point: Operator Recovery or Grade Target Metallurgical Control Set Point: Gram/Ton Ratio Control Measurement: Tonnage or Tonnage*Head Grade

Measurement: (Tail or Conc Assay)

Set Point: cc/min Flow Control

All of the Met. Loops are PI/PID controllers for supportability. If problems occur, operators may turn off one level of control at a time, independently for each reagent.

54

Addendum: Thickener Controls

For effective Thickener control 3 measurements are important: 1. 2. 3. Inventory (Inferred from pressure measurement in the Cone) controls underflow pump speed. Bed level controls flocculent addition rate. Overflow clarity indication/warning of poor control.

None of the above measurements are trivial and require attention to equipment selection and installation.

55

Typical Bed Mass Level Measurement on a Thickener

Shown is an E&H Deltabar S smart (2 wire) differential pressure transmitter. It is ideally used to measure level, pressure or differential pressure. Uses a flush mounted ceramic diaphragm since they out last the metallic diaphragms by at least 10 years.
(Ref. M. Gribbons)

56

Enabling Technologies
4. Flow Measurement: SONAR A New Class of Meters with Distinct Application Advantages
Ref: Christian OKeefe
cokeefe@cidra.com

Volumetric Flow Meter

ENTRAINED AIR Meters

Gas Void Fraction GVF-100

Multivariable DP

Vortex

Ultrasonic

SONAR

Gas Holdup Meter GH-100

Magmeter

Coriolis

57

SONARtrac Enables Accurate Mass Balance in the Presence of Air

Application: Hydrocyclone Feed Line Challenge:
Variable entrained air levels causing errors in density reading and hydrocyclone split This affects both flotation performance as well as ball mill circulating load
18000
V F (g p m ) G a s V o lu m e F r a c t io n

5 .0 %

16000

4 .5 %

14000

4 .0 %

Volumetric Flow (gpm)

3 .5 %
12000

Flow Rate Entrained Air %

3 .0 %

10000

2 .5 %
8000

2 .0 %
6000

1 .5 %

4000

1 .0 %

2000

0 .5 %

0 1 /2 5 /2 0 0 6 4 :4 8 :0 0

1 /2 5 /2 0 0 6 6 :0 0 :0 0

1 /2 5 /2 0 0 6 7 :1 2 :0 0

1 /2 5 /2 0 0 6 8 :2 4 :0 0

0 .0 % 1 /2 5 /2 0 0 6 9 :3 6 :0 0

Gas Volumne Fraction(%)

T im e

58

Enabling Technologies
5. Rotating Equipment Machinery Health:

59

Enabling Technologies
6. Wireless Instrumentation:

HighService
INDUSTRIAL SUPPORT COMPANY

Business Case 3: Smart Wear Sensor

2.- Proposed Solution

Electronic Wear Sensor (EWS) embedded in liners and lifters fastening bolts Main components of the EWS: Transducer
8-12 levels

CPU (coder & signal processing)

statistical filtering

RF Transmitter,
Sensor Life: 18 month
GITE-322-05-PRE-15

FM FSK, 916 MHz, 1mW, anti-collision algorithm

Rev 0 Fecha 100807 Pgina 19 de 29

GITE / DES

60

Case Study Example

CBU: Xstrata Nickel (Sudbury operations) Site: Strathcona Mill Project: Primary Grinding Control Status: Completed in 2008 Comments: Maintain Consistent Grind to Flotation Adjust, & optimise Rod Mill Feed Automatically to Capacity of Grinding Circuit Eliminate Process Upsets due to Ore Variability (i.e. Mill Overloads) Reduced energy consumption by 7.1% & 7.5% in the rod & ball mills respectively Implement on 2 grinding lines Presented at the Jan. 2009 meeting of the Canadian Mineral Processors.
Rod Mill (Power/tonne):
Histogram of RM kW/t
Normal 3.5 3.0 2.5 Density 2.0 1.5 1.0 0.5 0.0 2.64 2.76 2.88 3.00 3.12 RM kW/t 3.24 3.36
Status Old Ctrol New Ctrol Mean StDev 3.123 0.1275 2.900 0.1168

Ball Mill (Power/tonne):

Histogram of BM kW/t
Normal 1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0.0 6.6 6.9 7.2 7.5 BM kW/t 7.8 8.1 8.4
Status Old C trol New C trol Mean StDev 7.773 0.2732 7.189 0.2862

Density

61

Strathcona Mill Grinding Circuit

McIvor and Finch, 1991: Besides achieving the desired mineral d80 size, it is clearly desirable to produce as narrow a size distribution as possible to squeeze the maximum amount of the mineral value into the highest recovery region. Inadequate mineral liberation in itself leads to higher energy consumptions, as finer grinding has to be performed for liberation.
Histogram of RMF (Tons/hr)
Normal 0.16 0.12 D e n s it y 0.08 0.04 0.00 0 35 70 105 140 RMF (Tons/hr) 175 210 245
Status 1Before 2A fter M ean StDev N 173.3 33.95 719 188.2 2.866 721

Histogram of COFD (%Solids)

Normal 0.8 0.6 0.4 0.2 0.0 39 42 45 48 51 COFD (%Solids) 54 57 60
Status 1Before 2A fter Mean StDev N 49.81 4.844 704 47.14 0.4740 721

D e n s it y

62

Process Modeling
Step R esponse 0.8 0.7

Strathcona Grinding Circuit

Grinding Control Objective: Maximize the throughput (quantitative obj.) while maintaining the cyclone OF density at target (qualitative obj.)

0.6

PBL 0.73 e = RMFsp 1 + 485 s

10 s

0.5

0.4

0.3

0.2

0.1

0 0

500

1000

1500 Tim e

2000

2500

3000

Step R esponse 0

-0.02

-0.04

COFD 0.18 e = PBWsp 1 + 125 s

15s

-0.06

-0.08

-0.1

-0.12

-0.14

-0.16

-0.18

-0.2 0

100

200

300

400

500 Tim e

600

700

800

900

1000

11

Process Modeling:
Input and output signals 80
M e asu red and sim u late dm o del outp ut

PumpPump-box level (PBL)

70

15

Two SISO, simple PI controllers: 1. Cyclone Over Flow Density (COFD) by manipulating the Pump-Box Water (PBWsp) 2. Rod Mill Feed (RMFsp) based on the Pump-Box Level (PBLsp)

y1

60
10

50
5

40 0

COF Particle Size and Pulp Density

1000

2000

3000

4000

5000

6000

7000

8000

9000

10000
0

180

140

50 00 T im e

10 000

15 000

1000

2000

3000

4000

5000 Tim e

6000

7000

8000

9000

10000

COF P.Size (%-150 mesh)

Rod mill feed (RMFsp) RMFsp)

170

-5

160
-10

Correlation between COF Particle Size and Pulp Density

86.00 84.00 82.00 80.00 78.00 76.00 74.00 72.00 35.00 40.00 45.00 COF Density (% solids) 50.00 55.00

u1

150
-15 0

Input and output signals 49 48

M e asu red and sim u late dm o del outp ut 3

Cyclone OF density (COFD)

47 y1 46 45 44 43 0 500 1000 1500 2000 2500 3000 3500

0 1 2

145

PumpPump-box water (PBWsp) PBWsp)

140 135 u1 130 125 120 115 0 500 1000 1500 Tim e 2000 2500 3000 3500

-1

-2

Data from Strathcona Mill November 2007 Grinding surveys

-3 0 500 10 00 1500 20 00 T im e 25 00 30 00 35 00 40 00

63

Old Control Strategy

64

New Control Strategy

65

Before - Old Control

Some control, but not process control Fixed tonnage and oscillating cyclone overflow density Mill feed shut down intermittently to handle mill overloads Upsetting all OLD downstream flotation processes CONTROL Increase of milling rate can take hours PBL manipulating PBW

66

After New Control

Process control Controlled cyclone overflow density (particle size) and maximised rod mill feed rate No more mill feed shutdowns controller prevents mill overloads No more grindouts NEW CONTROL Controller reacts quickly to changes in ore hardness 1. COFD manipulating PBW and based on each PBL shift Easier to operate, 2.RMF consistently for

11

67

Time Series Plot of RMWi, RMF SPT (t/h)

200 RMF SPT (t/h) 180 160 140 120
01/10/07 03/01/07 04/14/07 06/08/07 07/29/07 10/02/07 12/09/07 02/25/08 04/18/08 06/13/08 09/24/08

55 45 35 25 15 RMWi

V ariable RM Wi RM F S P T (t/h)

Key Results
(Over a Longer Period)
Period: (Jan.07-May.08 vs Jun.08-Sept.08)

Date

Histogram of RMF SPT (t/h)

Normal
S tatus 1Before 2A fter M ean S tD ev N 168.2 5.756 517 181.2 7.098 77

Increase in circuit throughput of 7.7%: (168 to 181 tph); fully realised upon treatment of Ni Rim South ore; Grinding circuit feed to flotation (COFD) maintained at target density: reduction in variability from 2.0 to 0.8; An increase in energy efficiency (kW/t) of: 7.1% for the rod mill and 7.5% for the PBM; No Mill overloads; No degradation in Ni or Cu Recoveries.

0.060 Density 0.045 0.030 0.015 0.000 156 162 168 174 180 RMF SPT (t/h) 186 192 198

Time Series Plot of COFD (%)

60

COFD (%)

50

40

30 01/10/07 03/01/07 04/14/07 06/08/07 07/29/07 10/02/07 Date 12/09/07 02/25/08 04/18/08 06/13/08 09/24/08

Histogram of COFD (%)

Normal 0.48 Density 0.36 0.24 0.12 0.00 42 44 46 48 COFD (%) 50 52
S tatus 1Before 2A fter M ean S tD ev N 46.61 2.008 517 46.22 0.7656 77

68

Feedback Comments
Operators:
This has made the operations of the mills that much easier and efficient!DD It is working great and it is helping flotation as well. RA This is better and the right way to do it. SH To do better than the controller, I have to take samples every 15 minutes. BR I thought this will not work, but it is working very well. JM

Phil Thwaites (Manager Process Control):

Xstrata operates many grinding circuits. I believe that benefits such as these (as also demonstrated at Raglan Mill & Kidd Mill) can be duplicated at several other plants with similar grinding circuits. We often see very poor cyclone and grinding circuit controls that have not been tuned or optimised. In this project we have found the best way to control these two circuits the sweet spots (after all these years of operation).

69

Perceptive Engineerings Focus: (Vision - Sandoz)

40 35

Soft Sensors

30

Classification for Quality Control

FeO

25

Intelligent & Soft Sensors Optimisation & Scheduling

Operating Constraints

20

15

10 0

20

40

60 Sample No.

80 100 120 Operating Plans

Management Information Systems

Performance Reports

Advanced Process Control

Setpoints

Conventional Regulatory Controls

Early Warning Process Condition Monitors

40 35 30 25 SPE Confidence Limit

Valve position

Control System Integration

Conventional Sensors & Instrumentation

SPE

20 15 10 5 0

20

40

60

80

100

120

140

160

180

Time

Integrated Condition Monitoring and Advanced Process Control

70

Concluding Remarks
Plant automation is often seen as the project deliverable, when what is really required is plant control. There are many approaches, instrumentation, and multiple control systems, together with numerous advanced control packages to select from. Process control is more than just tools. Successful plant implementation is reliant on these together with:
process knowledge, a solid control engineering background / experience, and the operations team willing to act / implement / support the implementations.

Together, robust solutions can be realised, minimising process variation and optimising process performance.
This will result in an easier, efficient and safer process to operate.

(P. Thwaites, AUTOMINING2008, Chile)

71

At previous (high) metal prices the results from good control, and Operation Performance Excellence are substantial! At current metal prices good control and Operational Performance Excellence is essential! Operational Performance Excellence requires a solid performance of the regulatory layer AND process optimisation. Organizational structure and human resources are important in achieving Operational Performance Excellence. Thank you.
72