MILLING CONTROL & OPTIMISATION

MillSTAR

Specialists in mineral and metallurgical technology

OVERVIEW

Due to the complex nature of milling circuits, it is often found that

conventional control does not address many of the common
problems experienced. These include:

Long process response times.

Multivariable interactions and disturbances.

Non-linear systems with varying dynamics.

Constraints on variables.

Lack of reliable process measurements.

The Millstar Advanced Control System has a comprehensive suite

of control strategies that can be applied to provide an innovative
control solution for almost any milling circuit configuration. The main
goals are:

Stabilise the mill feed.

Control product quality to the downstream processes.

Optimise throughput and grinding efficiency.

Provide a robust control solution.

User
Interface
Optimisation

Millstar Power Optimiser,

Segregated Ore Controller
Millstar Mill Feed Control
Millstar Product Quality & Discharge End Control.
Millstar MPC
Millstar Specialised Surge Tank Control
Particle Size Estimator
Mill Discharge Density Estimator
Ball Load Estimator
Mill Power Filter

MillStar Intelligent Filtering and Fault

Detection for Level Signals, Density,
Size, Flow and Pressures

Estimators
Fault Detection and Signal Conditioning

Plant

Millstar
Millstar
Millstar
Millstar

Stabilisation

Plant Control
Plant Instrumentation
OVERALL MILLSTAR BENEFITS

Maximises throughput.

More consistent feed to downstream processes, thereby

Optimal usage of mills.

Individual customisation.

Improved control of mill feed rate.

Better management of fine and coarse material, thereby

increasing recovery.

preventing mill overloads.

Prevents violation of constraints, making daily operation

easier.

Robust control solutions for effectively handling

instrumentation failures.

MillStar boasts a typical payback period of just a few months.

RESEARCH
Mintek enjoys a proud history of research in the
minerals processing industry and is continually
striving to expand its knowledge in the field
through:

Partnerships with Industry

Inter-divisional collaboration within Mintek

University interaction

(Minerals Processing, Mineralogy, etc.)

MATERIALS HANDLING

MILL FEED CONTROL

Smooth operation of a milling circuit is difficult to achieve due to:

The varying nature of the feed material

Silo

(size, ore hardness, etc.).

Silo
Feeders

Mill

The unfavorable dynamics between feeders and the

weightometer. These dynamics degrade the performance
of PID controllers, making feed optimisation more
challenging.
(40 SEC)

MillStars Mill Feed Controller will:

Compensate for the feed dynamics by modeling feeder

Figure 1: Example of typical Time Delay on Mill Feed

responses.
Adapt for any model errors.

Adjust the feeder speeds in a desired ratio.

85
80
75
70
65
60
55
50

The graph on the right (Figure 2) shows a comparison between PID

control and MillStars Feed Controller for the plant illustrated in
Figure1. It is clear that setpoint tracking is much tighter and faster
under MillStar control.
Minteks control solutions are extremely flexible and can easily be
adapted for other applications. For instance, the mill feed controller

80

has been used to stabilise and optimise the operation of a Crushing

70
65

Multiple feeder and weightometer system.

Variable time delay in process responses.

Frequent belt trips due to limits on power draw of belt

5 minutes
Solids Feed Control - Mill Feed Controller

75

Circuit by overcoming the following challenges:

Solids Feed Control - PI Controller

1 minute

60
55
50
13:42
13:44
13:46
13:48
13:50
13:52
13:54
13:56

feeders.
The Mill Feed Controller was combined with Minteks safety controllers

Time

to achieve the following: (See Figure 3)

14:04
14:06
14:08

13:58
14:00
14:02

Time Delay

Figure 2: Comparison between PID control and

MillStars Feed Controller

Effectively stabilise the mass flow through the crushing

circuit, which improved screen performance and thus
increased throughput and reduced circulation load.

The safety controllers prevented belt trips by selecting the

correct operating conditions.
Tighter control enabled the plant to be pushed closer to its
Mintek Crushing
Control OFF

limits.
An increase in throughput of 4% was obtained.

600

Throughput Mass Flow

Throughput mass flow

average (t/h)

MillStar
OFF

MillStar
ON

%
Improvement

389

403

4%

Throughput mass flow

standard deviation (t/h)

94

Circulating mass flow

average (t/h)

969

916

5%

Circulating mass flow

standard deviation (t/h)

141

128

9%

73

Solids flow

Mintek Crushing
Control ON
50
40

400

30
20

200

22%
Feeder Speeds
0

10

0
500 1000 1500 2000 2500 3000 3500 4000 4500

Time (seconds)
Figure 3: Mintek Crushing Control

Feeder Speed (%)

SEGREGATED ORE FEED CONTROL (Feed Ratio Optimisation)

On milling plants fed by a segregated feed supply, such as a
stockpile, the varying size and hardness of the mill feed material
affects the residence time in the mill and the power drawn.
When the load becomes critically high, the feed needs to be cut
in order to grind the mill out. This dramatic change in mass flow

Calculate Ore
Coarseness

and particle size is passed to the downstream processes, causing

recovery to be compromised.
MillStars Segregated Ore Feed Controller is designed to:

Prevent mill overloads.

Adjust the coarse and fine feeder ratios to make optimum

Coarse

use of the ore supply and limit the need to grind the mill out.

Fine

Coarse

Feeder Speeds

Interact with operators rating of the coarseness of each

Adjust total Ore

Coarseness

Load or
Power

feeder to determine the size distribution of feed material.

MillStar Segregated
Feed Controller

Use advanced model-based techniques to reduce variation

in the feed and thus enable increase in average throughput
by operating closer to limits.

Figure 4: MillStar Segregated Feed Controller

Integrate with visual or other rock-classification systems.

THROUGHPUT OPTIMISATION

POWER OPTIMISATION

close to the maximum mill power draw. The power-load relationship

is highly non-linear and shifts around as the ore and steel load/liner
changes. Traditional control and modeling techniques can therefore
not be used. Mintek has developed a Power Optimiser that:

Power (kW)

For the most productive milling operation it is often best to operate

1000

800

600
500

20

30

50

60

Figure 5: Typical Power/Load curve

Uses the changes in the mill load and power to

Uses safety controllers to change the solids feed rate and

40

Load (%)

automatically detect whether the mill is overloaded or

10

Continuously seeks for the optimum mill operation by

underloaded.

Over-loaded so
decrease
Solids Feed

Under-loaded so
increase Solids Feed
Increase Solids Feed

700

changing the solids feed or load setpoint.

Increase Solids Feed

900

Power

Segregated
Power
Load
Ore
Optimiser
Controller

Solids
Feed
Rate

Feed Ratio
Feeder
Speeds

feed water to prevent mill overloads.

Estimates the power load curve and optimum load.

Power H/L

Load H/L

Ratio
Plant

By optimising the mill power usage and preventing mill overloads,

significantly increases in throughput from 6 to 16% have been
achieved.
RECIRCULATING LOAD OPTIMISATION
By optimising the recirculating load it is possible to maximise the
possible throughput of the milling circuit. The Recirculating Load
Optimiser dynamically adjusts the recirculating load to ensure
optimum efficiency of the circuit based on operator-specified
constraints. This strategy is particularly beneficial for overflow mills.

Safety
Controllers

Mill Feed Water

Figure 6: Mill Optimisation Control Strategy

70

Case Study 1:

MillStar Power Optimiser On/Off Data

(with a consistent ore source)

Mill
Throughput

MillStar
OFF

MillStar
ON

%
Improvement

Average
t/h miled

160.42

175.95

10

Std Dev

32.06

17.25

46

Minimum

1000

150

800
600

100
Mill Feed Cuts

50

Maximum
Mill Load
Mass Flow to
Flotation (t/h)

Std Dev
Std Dev

185
7.21
0.627

185
3.91
0.363

42
42

DAY 3

DAY 2

Mill Feed PV

Mill Feed SV

DAY 4
Mill Load

The standard deviation of the mill feed control is greatly

reduced.
The cyclone feed is more stable, allowing for consistent size
separation and feed to downstream processes.

Power

Figure 7: Mill Power Optimisation results for Case Study 1

MillStar OFF

3500
3000

400

2500

300

2000

200

1500
1000

100
0

500
0

30

60

90

Time (min)
MillStar ON

3500
2500

300

2000
1500

200

1000

100
0

0
120

3000

400

Figures 8 and 9 on the right show results from a gold plants SAG
mill achieved with MillStars Segregated Ore Feed Controller
combined with the Power Optimiser:

DAY 1

500

Case Study 2:

400
200

500

1400
1200

200

Power(kW)

250

Power(kW)

Mill feed cuts were prevented, resulting in a stable mill

loading.
No huge power dips occurred, since any sign of the mill
overloading was detected and rectified timeously.
The standard deviation of the mill load, flotation feed flow
and density was considerably less in MillStar mode, as can
be seen in the table below.
A significant increase in throughput of 15.5t/h on average
(about 10% increase).

1800
1600

Mill Overloads

The Millstar Power Optimiser gave the following benefits:

MillStar ON

Power(kW)

Also, the mill load varied in the range of 125 to 165 tons. These
disturbances propagated throughout the milling circuit and even
to the flotation circuit.

MillStar OFF
300

Mill Feed(t/h) & Load(t)

Figure 7 shows data collected from a platinum plant, where the

ore treated was very difficult to mill. In the first two days, the mill
experienced numerous power dips (overloads), and on at least
seven occasions the feed to the mill had to be completely stopped
to grind the mill out.

500
30

Mill Feed PV (t/h)

Cyc Pressure (kPa)

60

90

Time (min)
Load (t)
Power (kW)

Mill Feed SV (t/h)

Cyc Flow (m3/h)

0
120
Load SV

Figure 8: Mill Power Optimisation results for Case Study 2

Due to the tighter control the feed setpoint can now be set closer
to its maximum operating limits with confidence.

Feed Rate (t/h)

Cyclone Feed Pressure
(kPa)
Cyclone Feed Flow
(m3/h)
Cyclone Feed Density
(SG)

56.57

19.41

65.7

9.72

5.11

47.4

64.16

52.01

18.9

0.09

0.01

84.3

% of Total Data Points

Standard Deviation
%
MillStar MillStar
Improvement
OFF
ON

Comparison of Plant vs MillStar Mill Feed Control

45
40
35
30
25
20
15
10
5
0
100

318t/h
299t/h

200

300

400

500

Feed Rate (t/h)

MillStar OFF

MillStar ON

Figure 9: Feed Control Distribution for Case Study 2

MILL DISCHARGE PRODUCT CONTROL

THE MILLSTAR SUMP/PRODUCT STABILISATION FOCUSES ON:

Screen
Mill

Taking into account multivariable interactions between

input and output mill discharge variables.
Controlling the sump level and cyclone overflow product
size and/or density.
Minimising flow variation to the downstream processes.
Optimum usage or surge capacity of sumps, hoppers and
conditioning tanks.
Handling constraints of the sump level (to prevent pump
surging and spillage) and cyclone density (to prevent
pipeline chokes and pressure variations).

Mintek has developed a Model Predictive Controller (MPC)

specifically for controlling milling circuits, with the following advanced
features:

Water

Minimise Mass
Flow Variation

D
F

Sump

To Flotation

Cond Tank

MPC

MPC

Figure 10: Example of Mill Discharge Circuit showing MPC

Controller
MillStar Stabiliser ON/OFF Data using MPC

Standard Deviation

Flotation Feed Flow

(m3/h)

MillStar
OFF

MillStar
ON

%
Improvement

15.77

2.77

82

Flotation Feed (SG)

0.0107

0.0048

55

Mass FLow (t/h)

0.169

0.013

92

350
300
250
200
150
100
1.6

400

1.5

350

1.4

300
250

1.3
1.2

200
150

90

450

75

400

60

350

45

300

30
1.45

250
1.5

1.42

Mill Feed
(t/h)

240
220
200
180
160
140
120
100
80
75
70
65
60
55
50
45
40
450

Sump
Level (%)

240
220
200
180
160
140
120
100
400

Sump Disch
Flow (m2/h)

Mill Feed
(t/h)

MillStar
Stab ON

Float Feed
Flow (m2/h)

Data collected from a platinum plant is shown on the right (Figures

10 to 12). The feed to flotation in terms of volume flow and density
(mass flow) is much more stable in MillStar mode.

MillStar
Stab OFF

1.39
1.36

0.5

1.33
1.3
10.16

11.57

13.37

15.18

16.59

18.40

20.21

22.01

MillStar
Auth

Case Study 3

Sump Feed
(Water m3/h)

Sump
Disch SG

MillStar
Stab ON

Surge
Level (%)

Fully fledged multivariable controller that can efficiently

eliminate interaction between variables.
Explicitly handles limits on input and output variables to
ensure all variables are kept within their allowed operating
range.
Very efficient in handling long time delays and
slow-reacting processes.
Special features to handle noise, integrators and model
errors.

Float
Feed SG

Time

CYCLONE SWITCHING CONTROL

Clusters of cyclones are often used for mill discharge classification.
By controlling the number of open cyclones the pressure within the
cluster can be controlled. Stabilising the pressure improves the
stability of the cyclone overflow particle size and density, which
improves the consistency of the feed to the flotation. The module
can also balance cyclone usage and, if required, estimate the
cyclone's state (open/closed).

Float Feed SG - Error from Setpoint

45

% of Total Data Points

Stabilising the product quality from the milling circuit leads to

improved recovery downstream. In the case of flotation processes,
improvements of between 0.5 and 1.5% in recovery have been
shown. On gold leaching circuits, MillStar will minimise grind size
while still maintaining throughput targets. This leads to considerable
reduction in residue grinds.

Figure 11: MPC results for Case Study 3

0.003t/h

40
35
30

0.030t/h

25
20
15
10
5
0
-0.06

-0.04

-0.02

0.02

0.04

Error from Setpoint

MillStar OFF

MillStar ON

Figure 12: Flotation Feed SG for Case Study 3

0.06

MILLSTAR ESTIMATORS

MILL DISCHARGE DENSITY ESTIMATOR (MDDE)

Mill

The MDDE estimates the mill discharge density and flow using
dynamic volume and mass balances. An energy balance can
optionally be performed to enhance the accuracy of the estimate.
Grinding efficiency is highly dependent on the viscosity of the
material within the mill, which correlates closely with the density,
The calculated density can therefore be controlled by changing
the amount of water added to the mill, to ensure optimum grinding.

MDDE

Water

Sump

PARTICLE SIZE ESTIMATOR (PSE)

The PSE is a soft sensor capable of accurately predicting the particle
size of the hydrocyclone overflow product. It has the following
features and uses:

MILL POWER FILTER

An accurate mill power measurement is essential in ensuring reliable
control of a milling circuit. It can also give useful information about
the load in the mill. The power measurement is, however, cyclical
in nature and quite noisy. While conventional filters fail to separate
actual process variations and process noise effectively, the Mill
Power Filter succeeds by using high-frequency data and a
sophisticated filtering algorithm. The result (as shown in Fig. 15) is
a smooth power signal suitable for control and immune to noise
and cyclical variations.
FAULT DETECTION
MillStar completes and complements its control suite with a range
of fault detection and signal conditioning tools that can be used
in conjunction with expert rules to timeously detect instrumentation
failure and take necessary action to ensure smooth operation.
STEEL BALL ADDITION MODULE

Estimates the steel ball load in the mill.

Reminds operators of steel ball additions at pre-defined
intervals.
Displays steel ball addition history.
Suggests the amount of steel balls that should be added,
based on ore characteristics and changes.
Can be linked to an automated ball addition system to
control ball addition.

60

Particle Size (% - 75 )

The PSE uses an empirical model, identified from extensive

plant testwork.
It can be used as a back-up sensor for a physical size
measurement device.
It can provide a continuous estimate of the particle size
in-between slow updating or multiplexed measurements.
When a size measurement is available, this information is
automatically used by the PSE to continuously adapt the
model.
The PSE can also serve as an intelligent filter that eliminates
the delay associated with physical measurements by
inferring the particle size from measurements that have
much shorter delay times.

Figure 13: Mill Discharge Density Estimator

55

50

45

40

35
18:04

18:18

18:33

18:47

19:01

19:16

19:30

Time
Size Measurement

PSE

Figure14: The graph above shows how well the PSE

tracks the real particle size measurement. Note how it
can be used to eliminate spikes and noise in
the measurement.
900
800
700

Power (kW)

600
500
400
300
200
100
0

10

15

20

25

30

Time (seconds)
Raw Power

First Order Filter

Mill Power Filter

Figure15: MillStar Power Filter result

Contact Details:
Website: www.mintek.co.za/millstar.html or www.starcs.co.za email: millstar@mintek.co.za
Tel: + 27 (0)11 709 4379

Specialists in mineral and metallurgical technology