You are on page 1of 6

16th IFAC Symposium on Automation in Mining, Mineral

and Metal Processing


August 25-28, 2013. San Diego, California, USA

Using a Virtual Flotation Process to Track a Real Flotation Circuit


J. Kaartinen**, J. Pietilä*, A. Remes**, S. Torttila*

*Aalto University, Automation and Systems Technology, Espoo, Finland


(e-mail: janne.pietila@aalto.fi, sampo.torttila@aalto.fi)
**Outotec (Finland) Oy, Espoo, Finland
(e-mail: jani.kaartinen@outotec.com, antti.remes@outotec.com)

Abstract: This paper presents a novel way of using a dynamic flotation process simulation environment
for concurrent simulation alongside a real flotation process. The simulation environment is actively
adapted to a real flotation process. This yields new possibilities on the usability of the simulation model.
For example, virtual measurements can be introduced that provide completely new variables or increased
fault tolerance through redundancy. Another option is to provide feedback for the process operators in the
form of textual reports. These reports contain e.g. shift-specific performance indices that can be
compared against reference data. Furthermore, by speeding up the simulation, it is possible to make
prognosis on the effects of different control actions before actually making them.
Keywords: process simulators, operators, process models, adaptation, process automation.

(2006) propose an application of on-line tracking simulator to


1. INTRODUCTION
detect membrane degradation in Polymer Electrolyte
Froth flotation is a complex process and some parts of it are Membrane Fuel Cells (PEMFC).
still not fully understood. This is mainly due to complex
This kind of technique can be applied to flotation process as
nature of the physical details of the flotation phenomenon,
well so that non-measurable parameters, such as flotation
and also due to internal circulation loops that are needed for
factors, can be estimated. However, if the simulator can be
efficient operation of the process.
run faster than the actual process, new intriguing
One way to increase the knowledge on the flotation opportunities arise. An accurate simulator could be used to
phenomenon, and on the behavior of complex flotation make a prognosis of the future state of the process. When
circuits, is the use of simulation techniques. There are implemented to a real plant, operators could use the
commercial simulators available, such as Outotec HSC (see simulation environment to see the effects of different control
e.g. Roine 2009, Outotec 2013) and JKSimFloat (see e.g. actions in advance and choose the one that is most
JKSimFloat 2013), but they are often mainly used as an aid in appropriate in the given situation.
designing flotation circuits and process controls. Recently
The research group has previously developed a general
Bergh and Yianatos (Bergh and Yianatos, 2013) have
structure of a virtual environment intended for training
described a simulation system that can be utilized in control
purposes in off-line simulation scenarios (Roine et al., 2011).
design to take into account the changes in feed properties.
This work has been continued in order to be able to facilitate
However, due to the complexity of the mineral flotation
on-line simulations, where the simulation environment is
process it is extremely difficult to develop an off-line
actively adapted to a real flotation process.
simulator that is highly accurate and could describe a real
flotation process precisely. Recent advances in computing An overview of the simulator setup is given in section 2 and
have made it possible to design dynamic on-line simulators the application of the simulation system to Pyhäsalmi Mine
that can be run in parallel with a real process and adapt the copper circuit is described in section 3. Section 4 shows
simulation parameters so that the accuracy of the simulators another use case for the simulator; how the system is utilized
can be significantly improved. as a stand-alone simulation.
There are many applications to such simulators and a
common use case would be to track a variable that is not
2. PROCESS SIMULATION SETUP
directly measurable from the process. One such study has
been published by Jaklic et al. in 2007. They propose a The simulation environment contains all the necessary
simulator that can give accurate estimates of steel slab components needed for simulating a full flotation circuit (see
temperature fields that cannot be directly measured in a Fig. 1). The whole system is further packaged inside a
computer controlled hot-rolling process. The simulator is simulated operating system, i.e. inside a virtual machine. The
connected to the process database and current process data is structure is similar to the actual flotation process and to the
used to update the simulator. Similarly, Kawaguchi et al. automation system connected to it. The physical flotation

978-3-902823-42-7/2013 © IFAC 116 10.3182/20130825-4-US-2038.00061


IFAC MMM 2013
August 25-28, 2013. San Diego, USA

SURFHVV LV PRGHOHG ZLWK 2XWRWHF¶V +6& 6LP VLPXODWLRQ kinetics are obtained by model fitting them to the plant circuit
software (see section 2.1) and the process is connected to the sampling data. Calculation of mineral recoveries is based on
automation system via programmable logic controllers compartment model presented by Savassi (2005) in steady
(PLCs), just as in real life. The PLCs are also virtualized by a state form. Description of the applied calculation equations
custom made logic emulator (see Roine et al., 2011). for true flotation, entrainment and froth recovery are
summarized in Roine et al. (2011). True flotation recovery R
(%), with residence time t, for fast-, slow- and non-floating
particles with corresponding flotation kinetic rates (kF, kS and
kNon) and mass proportions of them (mF, mS and mNon) in
steady state is obtained by Equation (1). Here, the recovery
by flotation is calculated dynamically,

4 L srr @I¿
Þ· ç ÞÄ ç
E IÌ E IÇâá Û rA. (1)
5>Þ· ç 5>ÞÄ ç

The Pyhäsalmi copper flotation circuit, shown in Fig. 2,


consists of rougher ± scavenger bank, followed by four
cleaning stages and middling flotation. The average head
grade of Cu was 1.0% with 95% mill recovery in 2012, the
milling feed capacity is roughly 160 t/h (Inmet, 2013). The
dynamic flotation circuit model (Fig. 2) was set up with HSC
Chemistry® 7 Sim simulation module (Roine, 2009, Outotec
2013).

2.2 Simulated PLC and Automation System

If the only target would be to track the real process, one


would get by with a much simpler setup than the one depicted
in Fig. 1. In that setting the logic emulation and the simulated
automation system could be replaced by simply reading
Fig. 1. Components of the simulation setup. values from the plants automation system. This would
already give many benefits like completely new virtual
measurements and increased fault tolerance, for example.
Since the simulated process is not a single software package, However, this would not enable accurate prognosis
but a collection of separate pieces of software each capabilities.
performing a dedicated task, there must be some way to
synchronize the execution and to transfer the necessary data In order to carry out the prognosis as described in section 3,
between the different components. This is accomplished by the simulated system contains an installation of the
custom made communication software (marked with orange automation system, as well as, a logic emulator realized with
color in Fig. 1). The idea is similar to routing solution used in Matlab®. Both the automation system and the logic emulator
the Internet. Each node in the system is connected to the are initialized with the parameter values of their real life
central hub acting as the router. The router handles incoming counterparts. As the tracking simulation progresses, the
messages from each node by putting them into queue and simulated system is kept in check by Matlab® based
subsequently delivering each message into output queue adaptation routine, as described in the following subsection.
based on the desired target address. The messaging protocol Once the prognosis capabilities are needed, the simulated
is designed solely for this application. logic and automation system can continue independently.

2.1 Dynamic Flotation Circuit Model 2.3 Adaptation Scheme to Track the Real Circuit Response

The simulation model of the flotation circuit (i.e. the blue Adaptation algorithms are also coded with Matlab®, and are
rectangle in the lower left corner of Fig. 1) was set up as a thus integrated into logic emulation software. Several
property based simulation. It means that the feed material is adaptation approaches have been tested and the most
described with particles, each of them having their own size promising ones are currently under investigation. The results
classes, mineralogy and specific gravities. A brief description reported in this paper are obtained with a simple
of property based simulators for minerals and metals implementation of PID based adaptation. This is a good
processing can be found in Lamberg 2010. starting point for verifying correct operation.

Recoveries in the flotation units are modeled based on


flotation kinetic rates for each mineral and size fraction. The

117
IFAC MMM 2013
August 25-28, 2013. San Diego, USA

PYHÄSALMI MINE - Cu FLOTATION Analyzer_Courier Analyzer_PSI500

Total solids t/h LSA Sampler

CuEV CuRV
CuEV-VK12 CuEV-VK13 CuRV-VK14 CuRV-VK15
PK4
Box 1 CuRV-S
LCY
CuEVS CuRVS PK-3-4
160 CuEV12J CuEV13J CuRV14J CuRV15J
184 152 CuJ
155 152 151 151

CuEV12R CuEV13R CuRV14R CuRV15R 154


29.2 3.34 0.34 0.33
PK-11-12 PK-9-10

CuKV1S CuVVS
6.02
33.2
CuVV
Box 3 CuPV
CuVV-VK21 CuVV-VK22 CuVV-VK23 CuVV-VK24
Box 2 CuKV-1 CuPV-VK25
CuKV1-VK20 CuVVS2
CuVV21J CuVV22J CuVV23J CuVV24J
CuKV1S2
Cu1KV20J
PK-7 4.99 4.03 3.73 3.23 CuPV25J

24.3 Cu1KVJ 2.93


CuVV21R CuVV22R CuVV23R CuVV24R
1.04 0.95 0.30 0.49
Cu1KV20R 24.3
CuPV25R
8.85
0.30

CuKV2-VK33 CuKV2-VK34
PK-5-6
CuKV-2 Cu2KV33J Cu2KV34J CuVVKV
2.82 1.48 Cu2KVJ
3.97 CuVVKV-S CuVVKV-VK35 CuVVKV-VK36
Cu2KV33R Cu2KV34R CuVVKVS
6.02 1.35 CuVVKV35J
1.99 CuVVKV36J
1.30 0.90

PK-1-2 CuVVKV35R CuCCKV36R


0.69 0.40

CuKV3S
9.42

CuKV3-VK37 CuKV3-VK38
CuKV-3
Cu3KV37J Cu3KV38J
3.85 2.50
Cu3KV37R Cu3KV38R
5.57 1.35

PK-13-14

Cu4KVS

CuKV-4 6.93 CuKV4-VK39

Cu4KV39J Lattiakaivo
0.96 Cu4KV39R
5.97
PK-8

CuR
5.97

Fig. 2. Dynamic HSC simulation flowsheet of the Pyhäsalmi copper flotation circuit.

The general idea of the adaptation process is depicted in line laser diffraction analyzer), feed tonnage belt scales, and
Fig. 3, where it is shown that some measurements are merely slurry density transmitters.
fed through the adaptation system. These include, for
In addition, the flotation model parameters are updated based
example, the necessary variables describing the changes in
on the plant measurements to track the key measurements e.g.
the ore flow that is fed to the flotation circuit, as they can be
when the ore type is changing. Each flotation stage in the
used directly in the simulation model. The flotation kinetics
model (Fig. 2) has its own flotation kinetic rates. The
rates, however, do not have a direct physical interpretation
recoveries are tracked based on rougher-scavenger response
(i.e. cannot be measured directly) and thus their values have
to the final tails recovery, by manipulating the kinetic factors
to be estimated indirectly in the Parameter Estimator. The
(kF and kS in Eq. 1) of all flotation stages through the circuit.
parameter estimation concept is covered in detail in Pietilä et
The final concentrate Cu grade (%) is tracked by
al. (2013).
manipulating the pyrite kinetics, since it is the major sulphide
mineral tending to recover to the concentrate, thus affecting
to the Cu grade.
The variable pairings for the adaptation are listed in Table 1.
Still, the calculation of the measured recoveries involves
assays of feed F (%), concentrate C (%) and tails T (%)
streams as follows (Wills and Napier-Munn, 2006),
¼:¿?Í;
4 L srr . (2)
¿:¼?Í;

The physical meaning of the fast floating components of the


minerals, listed in Table 1, is related to the particles having
Fig. 3. Principle of the adaptation of the simulator highest degree of mineral liberation in contrast to slow and
parameters for tracking the real process. non-floating particles. This physical property relationship has
been reported e.g. in Welsby et al. 2010.

As mentioned, the model is run with similar feed composition


as the real concentrator plant. The feed mineralogy and
capacity are updated continuously based on plant on-line
elemental assays (XRF analyzer), particle size analysis (on-

118
IFAC MMM 2013
August 25-28, 2013. San Diego, USA

119
IFAC MMM 2013
August 25-28, 2013. San Diego, USA

120
IFAC MMM 2013
August 25-28, 2013. San Diego, USA

the improvement of the simulation model with regard of the Savassi O.N. (2005). A Compartment Model for the Mass
effects of reagent dosage changes. Transfer Inside a Conventional Floation Cell, Int. J.
Miner. Process, 77, pp. 65-79.
Welsby S.D.D., Vianna S.M.S.M., Franzidis J.-P. (2010).
ACKNOWLEDGEMENTS Assigning physical significance to floatability
components, Int. J. Miner. Process, 97, pp. 59-67.
The work reported in this paper is done in the SOREX
Wills B.A., and Napier-Munn T.J. (2006). :LOOV¶ 0LQHUDO
project, which is a part of the Green Mining program funded
Processing Technology, Seventh Edition. Butterworth-
by TEKES ± the Finnish Funding Agency for Technology Heinemann, Oxford, United Kingdom.
and Innovation. The authors would like to thank Outotec Oyj
and Pyhäsalmi Mine Oy for the fruitful collaboration.

REFERENCES

Bergh L., Yianatos J. (2013). Control of rougher flotation


circuits aided by industrial simulator, Journal of Process
Control, 23, pp. 140-147.
Inmet (2013). Pyhäsalmi Mine.
<http://www.inmetmining.com/ouroperations/production
/Pyhasalmi/default.aspx>, referred Feb 11, 2013.
Jaklic A., Vode F., Kolenko T. (2007). Online simulation
model of the slab-reheating process in a pusher-type
furnace, Applied Thermal Engineering, 27, pp. 1105-
1114.
JKSimFloat (2013). <http://www.jktech.com.au/jksimfloat>,
referred Feb 13, 2013.
Kawaguchi K., Onoe Y., Nakaya M. (2006). An application
of On-Line Tracking Simulator to a PEMFC, In
Proceedings of the SICE-ICASE International Joint
Conference, Oct. 18 ± 21, 2006, Bexco, Busan, Korea.
Lamberg, P. (2010). Structure of a Property Based Simulator
for Minerals and Metallurgical Industry, In Proceedings
of the 51st SIMS Conference on Simulation and
Modelling, SIMS 2010 October 14 to 15, 2010, Oulu,
Finland.
Moilanen, J., and Lamberg, P. (2010). Virtual Experience for
Operator Training. In Proceedings of the 7 th International
Mineral Processing Seminar (PROCEMIN2010),
Santiago, Chile.
Nakaya M., Xinchun L. (2013). On-line tracking simulator
with a hybrid of physical and Just-In-Time models,
Journal of Process Control, 23, pp. 171-178.
Outotec (2013). <http://www.outotec.com/en/Products--
services/HSC-Chemistry/>, referred Feb 13, 2013.
Pietilä J., Kaartinen J., Reinsalo A-M. (2013). Parameter
Estimation for a Flotation Process Tracking Simulator,
15th IFAC symposium on Control, Optimization and
Automation in Mining, Mineral and Metal Processing,
25-28 August 2013, San Diego, California, USA
(submitted).
Roine, A. (2009). New simulation tool for process engineers
and research scientists, Outotec press release 19th of Oct.
2009.
Roine T., Kaartinen J., Lamberg P. (2011). Training
simulator for flotation process operators, In Proceedings
of the 18th IFAC World Congress, August 28 ±
September 2, 2011, Milano, Italy.

121

You might also like