White paper, June 2015

Advanced process control for grinding circuits

Unlock the unknown potential of your mill

Recent copper price evolution pushes

modern mining industry even harder
to achieve highest production with
maximum efficiency. Advanced process
control (APC) for grinding is one
important tool to operate this most
energy intense area in an optimal way,
finding the optimal trade-offs between
highest throughput, energy efficiency
and product quality based on variable
ore properties, production targets and
management priorities.
 Abstract
Recent copper price evolution pushes modern mining industry even harder to
achieve highest production with maximum efficiency. Operations are bound by
numerous contractual and environmental conditions that bring a high degree of
complexity to both the operators controlling the process and the plant managers
assessing the circuit’s potential and setting the production goals.

Advanced process control (APC) for grinding is one important tool to operate this
most energy intense area in an optimal way, finding the best trade-offs between
highest throughput, energy efficiency and product quality based on changing ore
properties, production targets and management priorities.

ABB’s state of the art control and optimization software platform enables user to
Author:Dominik Groenveld design a controller based on model predictive control (MPC). The paper and the
E-mail: [email protected] presentation show the technology used and the results achieved through an APC
controller in pilot project Aitik (Boliden SA), Sweden. Furthermore possibilities to
ABB Switzerland Ltd. interconnect grinding with other areas as stockpile management and flotation with
Local Business Unit Process Industries various interconnected APC controllers will be shown. Finally, an example of the
Segelhofstrasse 9P return on investment and the payback times based on the increased throughput
CH 5405 Baden-Dättwil and better product quality will be presented.

Keywords
Advanced process control, model predictive control, grinding circuit, optimization,
energy efficiency

2 White paper | Advanced process control for grinding circuits

Introduction
Depending on the ore characteristics and the targeted plant capacity, the design
of the grinding circuit may vary significantly. Typically, the circuit consists of several
mills (rod, ball, SAG, AG) in series and/or parallel with a number of classifiers and
sumps at appropriate locations (see Figure 1).

The goal of the grinding section is to reduce the ore particle size to a level that
enables efficient separation between gangue and precious metals in the subsiding
processing step. The trade-off between quality and quantity has to be made. It is
also crucial to execute this process step at the lowest possible energy and grinding
media consumption.

Figure 1 — Typical grinding circuit scheme

An optimal grinding process controller must manage to maintain a given fresh

feed level while keeping the product size within a given range. To accomplish the
aforementioned, it is necessary to manipulate the ore feed rate, water feed rate and
the mill speed. These variables will have a direct impact on mill load, torque, power
draw and a slurry density. Disturbances adding another degree of complexity to this
control problem are ore size and the amount of circulating load.

Current Operation Strategy

Often, in wet grinding operation there are no proper feedback control loops, but
rather operator enters feed forward signals, such as the fresh feed rate to control
the mill power signal.

In most cases, the chosen control goal is to maintain circulating load at a given
target. This is accomplished by monitoring this magnitude and controlling it by
changes in fresh feed flow. Additionally, in some cases, there exist sensors to
measure the slurry density at the mill output and this is used to control the water
addition rates.

Advanced process control for grinding circuits | White paper 3

 In the current schemes, the resultant control actions are determined entirely on
current data and thus are “reactive” by definition. Whether the history nor the
immediate evolution of the system is taken into consideration. Consequently, it
becomes difficult to deal with the constraints of the process and its dynamics.

Current Customer Challenges

The obvious result of the aforementioned situation is high variability of the key
process variables leading to product quality outside the specified limits, increased
energy consumption, and high grinding media costs.

Methodology
ABB’s solution to handle multivariable problems with constraints is model predictive
control, a comprehensive methodology for control and optimization of industrial
equipment such as crushers, mills, kilns, furnaces, flotation columns and cells, etc.

Model predictive control

Model Predictive Control (MPC) is a simple iterative procedure: according to a
prediction of the short- to medium-term evolution of the process, a sequence of
future optimal control actions is chosen. The first element of the sequence is then
applied to the plant. As the system is evolving over time and new information
becomes available, the sequence is recalculated and replaces the previous one.
Each sequence is computed by means of optimizing the performance while
protecting the system from constraint violations.

Figure 2 — Model Predictive Control basic idea

4 White paper | Advanced process control for grinding circuits

Advantages and benefits of the proposed technology
With an accurate model of the physical processes in a plant, taking into
consideration explicit material balances and estimation of the involved time delays,
MPC enables one to predict the effect of given control decisions and disturbances
on the grinding process. Given this prediction, it becomes possible to take timely
control actions in any situation and facilitate the fulfillment of given targets.

The expected results are

– – Increased production due to tighter control to process targets,
– – Longer lifetime of mill structure and liners
– – Reduced consumption of grinding media,
– – Less upsets events downstream due to better control of the grinding process.
This will also lead to more consistent final product quality.

800xA APC: ABB Implementation of Model Predictive Control

800xA APC is a model predictive control engine fully integrated in the 800xA
system, ABB’s distributed control system (DCS). 800xA APC is available as an
800xA system extension. In addition there is a tool, the Model Builder, for modeling,
controller tuning, and what if analysis.

ABB Model Builder

ABB’s Model Builder allows the user to break down the process, for example a
grinding circuit, into well-defined building blocks (see - Figure 3). Every block may
contain further blocks or elements such as filters and mathematical operators. In
the control design section of APC model builder, objective functions with the proper
weights can be defined and the controller responses can be tested.

Figure 3 — ABB’s Model Builder displaying a typical grinding circuit

Advanced process control for grinding circuits | White paper 5

 800xA APC real time components
In 800xA APC there is a dedicated control module for a model predictive control.
Using this control module, the MPC algorithm is easily connected to measured
signals and to the PIDs downstream. Once the basic configuration is finished, a
number of tailored made faceplates are generated. These faceplates contain the
information that the operators and the APC engineer need to work (see Figure 4).

Figure 4 — Operator faceplate

Implementation of mpc in pilot grinding circuit

The grinding circuit consists of two lines with one Autogenous mill (AG) and one
pebble mill (PM) each. The primary objective of the grinding circuit is to grind
as much material within a certain particle size distribution. According to Bond’s
equation, the more energy is put into the system, the finer the product one obtains.
Generally speaking, higher fill level and speed of the mill lead to more energy
transmitted to the ore and hence finer particles.

The control structure consists of two levels of controllers (see Figure 5). The
supervisory controller takes two inputs from the operator: the setpoint of the
particle size and the absolute maximal throughput. The SC tries to minimize
Expression 1, which leads to a trade-off between the deviation from a given
particle size setpoint and maximal throughput (note the opposed signs, leading to
minimization or maximisation respectively).

min{∆particle_size – feed} (SC)

min{∆AG_load + ∆PM_load – feed – PM_torque} (LC)

6 White paper | Advanced process control for grinding circuits

Figure 5 — Control structure of pilot grinding circuit

Based on the measured particle size of the product, the setpoint for the loads and
the high limit of the throughput for the mills are calculated and passed to the low-
level controller.

The LC’s will then attempt to run at these setpoints (see Expression 2) by
manipulating speed and the actual feed without violating constraints such as torque
and power draw.

ABB process performance services

ABB offer services from the estimation of the potential of an APC controller to
the implementation and maintenance of it. It has to be mentioned that a drift in
the process or change of equipment have to be counteracted by fine-tuning or
readapting the model.

Advanced process control for grinding circuits | White paper 7

 Results and discussion
The evaluation phase of 800xA APC at the pilot grinding circuit will end in
November 2014. Final results will be presented at the conference. Nevertheless
preliminary results indicating the correct behavior of the controller can be shown.

Figure 6 shows how the setpoints of AG load given by the supervisory control are
followed by the low-level controller. If a limit such as a torque constraint shown in
Figure 7 is reached, the low-level controller approaches the setpoint as close as
possible without violating any constraints. Solid lines in the figures are measured
values, the dotted lines are predictions of the MPC controller.

Figure 6 — AG mill load during APC ON and OFF phases

Figure 7 — AG mill load and torque predictions

8 White paper | Advanced process control for grinding circuits

Figure 8 shows the particle size (P80) during 4 days with on and off intervals. During
the offline phase, particle size is not specifically controlled. When APC is online, the
particle size setpoint is smoothly followed. Having a dynamic setpoint for particle
allows the operator to react to changing ore properties and daily production targets,
ensuring the optimal trade-off between production quality and quantity.

Figure 8 — Particle size during APC ON and OFF phases

Conclusion
Final results will be presented at the conference.

Further applications of model predictive control

MPC is generic and can be used in any area of the plant. For instance, the processes
crushing and flotation preceding and succeeding grinding can be controlled and
interconnected with APC, such that e.g. a disturbance in ore hardness will already be
registered in the crushing area and according actions will be taken before the material
enters the grinding and flotation stage. Merging individually treated processes into
one optimization problem leads to better and easier control of the plant. Further con-
trol and optimization packages are under development.

Acknowledgements
ABB Switzerland, CPM group. http://www.abb.com/CPM
ABB Switzerland, Mining & Minerals processing industries. http://www.abb.com/mining

Nomenclature
APC Advanced process control
AG Autogenous mill
MPC model predictive control
PM Pebble mill
PSA Particle size analyser signal
SP Setpoint
LAG Load of the AG mill
TPM Torque of the pebble mill
wx Weights

Advanced process control for grinding circuits | White paper 9

 Co-authors

Maria Rouzou
[email protected]

Florian Kolb
[email protected]

Eduardo Gallestey
[email protected]

10 White paper | Advanced process control for grinding circuits

Advanced process control for grinding circuits | White paper 11
Contact us

ABB Switzerland Ltd.

3BHS 820 145 ZAB E01

Local Business Unit Process Industries
Segelhof 9P
CH-5405 Baden 5 Dättwil
Switzerland
Phone: +41 58 586 84 44
E-mail: [email protected]

ABB’s Mining business is represented in the following countries:

Australia, Brazil, Canada, Chile, China, Germany, Kazakhstan,
Peru, South Africa, Sweden and Switzerland.

For contact details, please visit our website:

www.abb.com/mining

We reserve all rights in this document and in the subject matter and illustrations
contained therein.

Any reproduction, disclosure to third parties or utilization of its contents — in

whole or in parts — is forbidden without prior written consent of ABB
Switzerland Ltd.

© Copyright 2015 ABB. All rights reserved