See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/350290488

Flotation Circuit Optimization Using Modelling and Simulation

Article · March 2021

CITATIONS READS
0 98

2 authors, including:

Ronney Rogério Rodrigues Silva

FLSmidth
11 PUBLICATIONS   57 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Iron Ore View project

Flotation View project

All content following this page was uploaded by Ronney Rogério Rodrigues Silva on 22 March 2021.

The user has requested enhancement of the downloaded file.

Flotation Circuit Optimization Using Modelling and
Simulation
Ronney Rogerio Rodrigues Silva1*
1. R&D Project Manager II, Research & Development, FLSmidth, United States

ABSTRACT
Evaluation of a flotation circuit is a complex and challenging task. The professional responsible for
this task needs to take into account several variables, that range from feed related to equipment and
process. Optimization is complicated when accounting for multiple variables such as: type of ore,
variability in terms of mass flow, volumetric flow, assays, froth depths, air flow rates, reagent type,
dosages, pH, etc. Taken into consideration that some of these variables are interdependent, finding
the optimal process conditions in this scenario is non-trivial.
Developing a robust and consistent path for evaluation is essential to generating successful models.
Using tools that help to identify possible points of issues and improvements in the circuit are one of
the important characteristic included in this work.
As the first step of this evaluation a plant historical data analysis is performed. This gives a snapshot
of the current performance and includes all of the variables measured by the process plant; mapping
its interdependence and effects on the responses of recoveries, grades, efficiency, selectivity, etc.
Another important part of this process is being able to reflect the unique characteristics of different
flotation equipment and its impact in the flotation circuit.
FLSmidth, as one of the leaders in the flotation market, has developed a unique and robust flotation
circuit evaluation tool using its own models that reflect all of the characteristics mentioned above.
Those models can be used in steady state and/or dynamic simulations to reflect the variations at a
plant.

1
INTRODUCTION
Evaluation of a flotation circuit brings several challenges to the professional responsible for that task.
This is caused by the complexity of the variables involved in the process and their interdependence,
a few of mention are: process variables, ore type, equipment variables, etc. Thus, the know-how and
experience of the professional is of paramount importance when doing these types of evaluations.
However, with the advance of technology and digital world, lots of options in software are available
to help on day to day tasks. This is also true when applied to mineral processing, currently we have
several options that allow us to better evaluate a flotation circuit. The use of the right tools allied with
the know-how and experience of the professional responsible for this task, can bring exceptional
results, resulting in a more accurate and better evaluation.
A clear path and well-defined approach steps are also fundamental for a successful evaluation.
Discussed in this paper are the step by step stages resulting in a clear and systematic approach for
flotation circuit evaluations.
In the first step it is important to explore the current data available for the circuit to be evaluated.
This step can reveal information that could lead to possible improvements/optimization of that circuit
and important clues about how the plant is operated. This stage is called the plant historical data
analysis. It gives a snapshot of the current performance, including all variables measured by the
process plant, mapping its interdependence and the effect on the responses of recoveries, grades,
efficiency, selectivity, etc.
The second step for evaluation is the equipment inspection. This stage will reference the first stage
and where to focus in on the circuit.
The third step will be a plant survey, collecting more information to unveil weak points observed in
the first and second step.
And the final step combines all of the information gathered in the previous steps to run a series of
simulations, in order to optimize the circuit. The simulation models used by FLSmidth were
developed internally in partnership with prestigious universities and reflects the unique
characteristic of each piece of FLSmidth flotation equipment. The models developed can be used in
a steady state and/or dynamic simulation to reflect the variations at the plant.
To illustrate, the process, data from a copper plant evaluation is used.

METHODOLOGY

Each step of the process outlined in the introduction above is discussed in greater detail below.

First Step – Historical Data Analysis

The historical data analysis is performed in order to collect more information about the flotation
circuit performance.

2
This type of analysis is not trivial and can bring many issues, to mention a few:
• Inconsistent Data
• Dynamic Relations
• Nonsense correlation
• Etc
It is rare that data gathered over a long period of time is consistent and comparable. Standards are
modified over time, instruments change, calibrations drift, operators come and go, raw materials
change, processes age, ambient conditions change and there may be seasonal effects (Mark Anderson,
2009).
The use of a tool that is powerful enough to overcome most of the issues described above is of
fundamental importance. The present work makes use of two different tools. First is to perform a
mass balance reconciliation, to check inconsistent data. It is based on the minimization of the sum of
the squares of the closure residuals. This methodology calculates the best-fit values of mass flow rates
using the experimental data and the data is adjusted to match mass values representing the
minimized residual (Barry A. Wills & Tim Napier-Munn, 2006).
The first analysis removes any inconsistent data. The second tool is a statistical tool, Design-Expect
11 from stat-ease. The software has a section called Historical Data that creates a blank design layout
to accept component, factor settings and responses from an existing data set. Allowing for verification
of anything that emerges as statistically significant and shows how to model the results most
precisely.

Second Step – Equipment Inspection

Throughout a flotation equipment inspection, a plant operator can gather valuable information about
the “equipment health” and its impact on the operation, for instance: wear parts, missing parts,
sanding caused by plant upsets, incorrect equipment configuration missed by numerous
maintenance or alternative parts. All of these can affect the equipment performance enormously, or
have a weak impact, depending on the number of issues and combinations of them.
Some plants with a large number of flotation devices make it almost impossible to inspect all of the
equipment during a shutdown. A possible solution is to build an inspection schedule and/or use the
first evaluation step (Historical Data Analysis) to guide you in this inspection, identifying stages with
low performance and focus the inspection around those stages.

Third Step – Plant Survey

At this point there is already enough information from the previous steps to guide the professional
in performing the plant survey. For instance, from the historical data analysis, data can be collected
from streams that were not part of the original plant routine, but give you important information in
order to better mass balance a cell/stage/circuit that is not performing well. Also, based on the

3
inspection, the problematic spots are known in the circuit equipment allowing focus on those areas
to better understand the impact of the equipment in the current operation. At this stage it is also
important to analyse the samples in terms of mineralogical components, assay by fraction (recovery
by fraction), etc; all information that generally is not part of a routine analysis, but gives pieces to put
this puzzle together.

Final Step – Simulations and Optimization

With the information from previous steps, it’s time to start some simulations to investigate several
possible scenarios and optimize the flotation circuit.
FLSmidth has different types of flotation machines, each one with its own characteristics. Using a
model that can reflect those fundamental differences, results in a better and more accurate response.
In this evaluation FLSmidth uses two different models, the first one is based on process characteristics
and the other is based on CFD parameters. Both models can be used in steady state and/or dynamic
simulations to reflect the variations at the plant.

RESULTS AND DISCUSSION

An evaluation done for a copper plant is used to illustrate each step of the process. The historical data
analysis and equipment inspection was performed for the rougher stage as it is the stage with the
most impact on the circuit. The plant survey was completed for the entire circuit including Rougher,
Regrinding, Cleaner, Scavenger and in the Final Cleaner stage. Final evaluation simulations were
calculated for specific stages with consideration for the overall circuit.

Historical Data Analysis - Results

The Rougher stage (2 rows) has an immediate impact on the circuit, being the tails from this stage are
the final tailings and unrecoverable, it was naturally chosen as the stage to do the historical data
analysis in order to understand the operation and performance in depth.
The first action for this evaluation is to look at the raw data and perform a mass balance reconciliation.
Figures 1 and 2 show an example of the reconciliation done for one case using data from a copper
plant. Figure 1 shows the adjusted assays, residuals and adjusted vs raw data. It is interesting to
mention that the residuals for elements 1(Cu), 2(Fe), 4(S) and 5(Si) are different than zero. Figure 2
shows a sensitivity analysis regarding the yield (Y) and recovery using the copper assays, in this case
it is verified that the Feed copper assay has the major impact on the yield and recovery. Another
possible evaluation is weighting the adjustments, as shown in Figure 3. Weighting the feed, conc and
tails as relative errors of 10%, 5% and 5%, respectively, decreased the residual for the elements 2(Fe),
4(S) and 5(Si) with less adjustments in the raw data elements, except for Mo.

4
Doing this first evaluation with the entire set of raw data (306 data points), the inconsistent data can
be easily detected and eliminated from the original set. A consistent raw data set will be less
susceptible to corrections/adjustments even with higher relative errors.

Figure 1 - Software developed by Ronney R R Silva- Weighted Adjustments

Figure 2 - Software developed by Ronney R R Silva- Sensitivity Analysis

5
 Figure 3 - Software developed by Ronney R R Silva- Adjusted Weighted

The next step is to model the reconsolidated data using the historical data section of the Design Expect
11 software. In order to have a better understanding of the process, the variables were selected based
on the control factor and noise factor (Mark J. Anderson & Shari L. Kraber. ( 2020).
✓ Feed pH each row;
✓ Feed P80;
✓ Feed assays (CuT ; Mo ; FeT/CuT ; CuAc/CuT);
✓ Feed pH;
✓ Froth Depth;
The following responses where elected to measure the process effectiveness: Cu Recovery, %Cu
concentrate and Yield. After a series of interactions to eliminate outliers the models used 306 data
points and showed statistically significant for all responses evaluated. Figure 4 shows the Cu
Recovery and Cu concentrate responses relative to pH (Row 1 & 2), it is important to notice for this
specific feed characteristic, the best conditions in terms of pH are around 9.5 to 10.5, this is where you
get the best recovery and final grade.
Exploring the effect of particle size, while keeping a constant feed characteristic and changing the
P80 from 140 microns to 80 microns the changes resulting from this modification are shown in Figure
5. The pH range was expanded, now also possible to work on a lower limit of pH 8.8, and the overall
final concentrate has decreased slightly.
These types of evaluations are very important for the plant operator, with a model that can be trusted
and with the feed characteristics in hand, the plant operator can predict/optimize the flotation circuit
based on a specific feed characteristic. FLSmidth can transfer all of these models to a standalone
software that has an optimized interface to meet the customers main characteristics and priorities
used at a plant site.

6
Figure 4 – Rougher Copper Recovery and %Cu concentrate responses related with pH Row 1 & 2 –
P80 (140 microns);

Figure 5 - Rougher Copper Recovery and %Cu concentrate responses related with pH Row 1 & 2 –
P80 (80 microns);

Equipment Inspection - Results

Equipment inspection is one of the most important steps in this process, it will at once tell you if the
flotation machine is operating according to expectations. Using the data from the previous historical
data evaluation, it can give you indications of where to focus the inspection by looking at points
where the flotation performance is operating below expectation.

7
FLSmidth has the state of art in terms of flotation machines and each one has its own characteristics
and configurations. During the equipment inspection it is important to look for the main parameters
that makes each design unique. For instance the unique components for the Wemco designs are the
Draft Tube, false floor, submergence, engagement, etc and for the nextSTEP design it is clearance,
profile eccentricity, etc. Figure 6 shows the Wemco (self-induced air) and nextSTEP (forced air)
designs with the main parameters.

Figure 6 – Wemco (Self-induced air) and nextSTEP (Forced air) FLSmidth Flotation Cells.

Plant Survey - Results

Based on previous steps, a plant survey was conducted in order to collect more information around
the critical points identified. Table I below shows the main rougher stage feed characteristics during
the survey and Table II shows a comparison with the actual results from the survey and the results
from the simulation using the historical data models. As we can see the simulation and plant survey
agree well and the models can be used for the plant operators as guidance for operation.

Table I – Main Plant Characteristics during the survey

Slurry Level P80 CuT Mo FeT/ CuAc/
Description pH Row1 pH Row2
mm Microns (%) ppm CuT CuT
Plant main Characteristcs
9.87 9.76 416 180 0.641 66.33 3.49 0.047
(survey period)

Table II – Rougher performance during the survey vs Historical data model simulation comparison
Description Y(%) Cu Rec. (%) Conc. Cu(%)
Plant Rougher Results
7.62 92 7.74
(Survery period)
Historical Data Simulation 8.25 92.32 7.16
Standard Deviation -- 10.4 1.8

Simulations - Results

8
The models used by FLSmidth seek to reflect the unique characteristics of its design in the results,
thus promoting better and more accurate simulations. The models are based on compartment
models, the schematics of both models are shown in Figure 7 below.

Figure 7 – FLSmidth flotation process and CFD models.

One of the main evaluations requested by plant personnel are to evaluate an increase in throughput,
and how that would affect the flotation circuit performance. In order to answer that properly, the
entire circuit was simulated varying the throughput from 4,698 tph to 5,450 tph. The feed
characteristic was set the same as the survey. The survey summary is shown in Table III. For each
throughput simulation an attempt was made to keep the same concentrate quality found for each
stage as shown in Table III.
Table III – Plant survey results by stages
Feed Conc Tails
Solids Solids Solids SE RT
Cu% Fe% Cu% Fe% Cu% Fe% S%
Description tph tph tph
Rougher 0.64 2.23 4698.00 7.74 8.59 357.76 0.06 1.71 1.13 4340.24 85.36 28.8
Cleaner 14.66 14.19 627.29 30.95 16.52 238.07 4.69 12.77 14.20 389.21 56.2 29.1
Scavenger 4.69 12.77 389.21 17.73 23.86 98.45 0.27 9.01 9.19 290.76 76.4 152.2
Cleaner II 30.95 16.52 238.07 40.12 6.72 67.00 27.36 20.36 26.90 171.08 17.6 21.3
Hydrocyclone 10.32 19.56 3795.42 10.38 20.61 3339.21 9.89 11.88 15.79 456.21 --

Figure 8 below shows the summary of simulation results. Figure 8-A shows an important decrease
in the overall recovery with an increase in the feed throughput, the same concentrate quality was
kept for each stage as shown in Figure 8-B. In terms of Separation efficiency (using Equation 1) of
each process, the biggest impact was in the rougher stage as seen in Figure 8-C.

(𝑅 − 𝑀)
𝐸=
(1 − 𝐺𝑓)

9
 (1)
Although you have a decrease in the overall % recovery the final concentrate production (tph)
increases due to the higher feed throughput. An economic evaluation accounting for maintenance
costs, plant costs, etc, was performed to determine the breaking point where the increase in the
throughput wasn’t worth it anymore. Figure 8-D shows the revenue vs throughput curve, that point
is at 120,000 tpd plant feed, where the maximum is reached in terms of revenue.

Figure 8 – Rougher steady-state simulation results summary.

As shown in the previous evaluation, the main impact in the circuit was observed in the rougher
stage. In order to get more detail out of this stage a simulation was performed in a dynamic
environment varying throughput and %solids feeding the rougher and the simulation was allowed
to run freely. The simulation ran 200 hours of operation and collected a sample every 0.1 hour,
generating more than 2002 data points that were also mass balanced, Figure 9 shows the results from
this specific simulation.
Looking at a broader throughput range and not restricting the final concentrate, a grade recovery
curve is shown in Figure 9 and the variation of each component evaluated. This type of simulation
allows for better understanding of the performance of that stage, giving more confidence in the
results.

10
Another important evaluation using this simulation was to evaluate each cell in the rougher stage
and compare the ideal performance vs the current results found during the survey. The rougher
stage has 8 cells in a row and the results of this comparison are shown in Figure 10.

Figure 9 – Dynamic simulation rougher stage varying throughput and %solids feeding the circuit.

Figure 10 – Rougher simulation by cell - copper recovery and %Cu concentrate.

Comparing the ideal curve vs the current plant data, there is a possible opportunity for improvement.
Specifically in cell 04 to cell 05, there is a strong jump in the curve, that could be caused by reagent
addition, cell performance, etc. This type of evaluation helps to identify possible points of

11
optimization. Going further with the evaluation, another tool that can be used in this case is the CFD
model simulation. Figure 11 show the CFD model calibrated using the parameters collected in the
survey (Black dotted line). Once it is calibrated, we can start to change some parameters in order to
optimize the area of cell operation. For this specific case cell 04 was selected, the froth height was 384
mm and that was changed to 300 mm. The result of this modification is shown in Figure 12. The
recovery increased, shifting the curve closer to the ideal curve. Other parameters can be changed to
improve the optimization (Jg, air hold up, cell parameters (tip speed), etc).

Figure 11 – Rougher stage by cell – CFD model calibration

Figure 12 – Rougher stage by cell – CFD model simulation (Cell 04 froth depth changes).

CONCLUSION

Historical data analysis models have a good agreement with the survey results. The models can be
used to guide plant operators in order to optimize the process parameters.
Flotation simulation tools using Process and CFD based models are a good tool to help achieve a
more in depth analysis of the flotation circuit, for instance identifying ideal operation parameters,
possible points of issues (cell/process related), etc.

12
 Dynamic simulations also give a more realistic view of the results, once real life changes are
incorporated, a few of mention are: feed variation, %solids, tph, bubble size, etc.

ACKNOWLEDGEMENTS
I would like to thank the following people that contributed and made this work possible.
1. Rebecca Siwale - Program Manager, Productivity and Audit, Mining Industry. FLSmidth Salt
Lake City. The present paper is a section of a project developed by FLSmidth which Rebecca
is managing. The project called Total Cost of Ownership (TCO) Tool, allows operations to
identify maintenance activities that are economically viable and generate optimal and
reliable degrees of up-time and reliability for the equipment.
2. Dariusz Lelinski - Global Product Manager, Flotation – FLSmidth Salt Lake City.

NOMENCLATURE
E reduced recovery efficiency
R recovery of valuable mineral in concentrate
M mass percentage reporting to the concentrate
Gf feed grade

REFERENCES
Barry A. Wills & Tim Napier-Munn. (October 2006) Mineral Processing Technology, 7th edition,
Elsevier Science & Technology Books.

Mark J. Anderson & Shari L. Kraber. (January 2020) ‘Cost-Effective and Information-Efficient
Robust Design for Optimizing Processes and Accomplishing Six Sigma Objectives’, Society of
Manufacturing Engineers, (https://www.statease.com/casestudies).

Silva, R. R. Ronney. (2020) ‘Flotation Circuit Evaluation Using Simulation Approach, MineXchange
2020 SME Annual Conference & Expo, Phoenix Convention Center, Phoenix AZ.

Amini, Hassan & Silva, R. R. Ronney. (2020) ‘Development and Validation of a Multi-Compartment
Model for the Performance Prediction of Flotation Circuits, MineXchange 2020 SME Annual Conference
& Expo, Phoenix Convention Center, Phoenix AZ.

Mark Anderson. (September 2009) Problems Analyzing Historical Data,

https://www.statease.com/webinars.

13

View publication stats