You are on page 1of 5

HIGH CONFIDENCE-LEVEL FLOTATION TESTS BY USING AN AUTOMATED

LABORATORY MACHINE

Jaime Álvarez* and Rodrigo Ormeño*


 
* Departamento de Ingeniería Metalúrgica, Universidad de Concepción, Concepción, CHILE, jalvarez@udec.cl,
rormeno@udec.cl

ABSTRACT
The acquisition of reliable flotation data demands the collection and processing of a large number of results with the
least sources of errors. Main source of errors are in the addition of precise volumes of reagents, in the frequency and
velocity of concentrate removal, in the stability of gas flow delivered, and in the control of pH, impeller rotation speed,
pulp level, and conditioning and flotation times. The use of an automated flotation machine is a logical approach to
decrease errors in laboratory flotation tests, which was demonstrated in this work by recovering zinc from a
galena/sphalerite Chilean ore. An experimental program was designed using Design-Expert 8.0 from Stat Easy. The
tests were performed in four different flotation machines, three manually operated and one automated, searching for the
one with the minor variance in the zinc and lead recoveries and grades. The results demonstrated that the most reliable
and reproducible results were obtained in the automated machine (EDEMET).
Keywords: Flotation, Laboratory, Automatic, EDEMET, Cell.
Relevant Congress topic: Sulphide flotation

INTRODUCTION the application of an adequate tool for experimental


design and data processing, and the establishment of a
The need for constant grade, recovery and yield protocol for running the tests, will minimize errors and
improvements in the flotation stage at every metallurgical produce results with a higher reliability and
operation starts with preliminary laboratory tests, which reproducibility.
provide valuable information for controlling daily
operation, and also for making decisions during the EXPERIMENTAL ASPECTS
expansions or the design of new projects. These tests
are normally tedious and involve a large amount of work, Lead-zinc ore sample characteristics
as a consequence of the considerable number of flotation The selective separation of lead-zinc ores by flotation has
tests required, which must be conducted by or under the been largely studied (Sutherland, 1977; Vidal, 2012;
supervision of a highly experienced technician who is Bulatovic, 2007; Gutierrez, 1973). A combination of
often under pressure to produce results quickly. The use reagents and operating conditions recommended in the
of commercial software for organizing, processing and literature was selected for the tests. The ore was
interpreting the data also reduces errors, which although characterized preparing a representative sample for
not associated with the collection of the data, may Qemscan mineral determination. The results summarized
significantly affect the results of the whole exercise. in Fig. 1 and Table 1 showed that the major components
Several attempts for automating laboratory flotation were sphalerite and galena, but there are also some
machines, such as the works by Roberts et al. (1982), mineral associations of interest and also some contact
Jämsä-Jounela et al. (2002), Luttrell and Yoon (1982), area textural information for different species.
Hamilton and Guy (2001), Apling and Ersayin (1896),
Dell and Bunyard (1967), are an indication of the
difficulties in doing it; the most complicated aspect not
resolved yet is froth depth control. Two cells from
different suppliers, when operated at high rotor speed,
produced large differences in gas dispersion
characteristics, as demonstrated in the results of Kracht
et al. (2005). On the other hand, the applications of
experimental design and the use of adequate software
for their analysis are much common (Hasu et al., 2000;
Craig and Koch, 2003; Anderson and Fayram, 2005;
Anderson, 2006; Sheridan et al., 2002; Montgomery,
2003; Rao, 1985; Kelsall, 1961). Another aspect which
must be considered is the necessity of a using a
laboratory standard procedure as described in Brown and
Hall (1999), Greet et al. (2012), and Runge; (2010).
The main objective of this work was to demonstrate that Figure 1- Mineral associations
the use of an automated flotation machine, together with
These associations also provide a preliminary indication p-value is less than 0.05 the differences between
of liberation levels of the individual species. Chemical machine performances are statistically significant.
analysis for the composite head was 2.65 % Pb and 7.74 Results of these analyses are presented in Tables 3 and
% Zn. 4.

Table 1 - Quantitative mineralogy (weight %) Table 2 - Pb and Zn recoveries and grades for three
replicas in the different flotation machines.
Machine Pb Rec.% Pb %  Zn Rec.% Zn %

EA 93.96  11.08  59.63  49.92 

EA 93.79  10.42  59.96  49.72 


EA 95.87  11.86  58.69  49.64 

ED 79.89  4.20  59.70  43.60 


ED 71.61  3.83  58.79  44.50 

ED 78.70  4.44  60.46  46.30 


D 51.88  2.21  69.12  48.60 

D 76.69  4.47  66.18  47.30 

D 70.81  3.04  60.67  41.90 


W 74.31  3.98  54.53  48.60 

W 74.81  4.35  66.15  51.70 


W 71.23  4.40  62.23  47.30 

The flow-sheet selected for this first-step study, with feed EA: Automatic Edemet laboratory flotation cell
weight and lead and zinc grades, is shown in Fig. 2. ED: Semi-automatic Edemet laboratory flotation cell (same
machine than EA with Denver recipient and
mechanism adaptation, without automatic froth
remover)
D: Denver D1 laboratory flotation cell
W: Wemco laboratory flotation cell

Table 3 - Calculated means and p-values


Machine Pb rec. % Pb % Zn rec. % Zn %
a a
EA 94.5 11.1 59.4 49.8
ED 76.7 b 4.2 b 59.7 44.8
b b
D 66.5 3.2 65.3 45.9
W 73.5 b 4.2 b 61.0 49.2
Fig 2 - Flow-sheet for recovering Pb and Zn minerals
p-values 0.0063 <0,0001 0.2530 0.0651
Experimental work
Table 4 - Calculated variances and p-values
The first objective of this work was to generate the
necessary experimental data to find out, using statistics Machine Pb rec. % Pb % Zn rec. % Zn %
principles, which cell gives the most reliable and 1.3 0.5 0.4 0.0
reproducible results, when run under the same operating EA
conditions. The same experimental conditions were used ED 20.0 0.1 0.7 1.9
to run four tests in each cell. The results of these tests, D 168.1 1.3 18.4 12.6
which were completed following a randomized design W 3.8 0.1 34.9 5.1
(Table 2), are associated to each one of the flotation
machines EA, ED, D, and W (Fig. 3), and were the p-values 0.0227 0.1916 0.0357 0.0293
populations used for the statistical analysis characterizing
cell performance. The results in Table 3 clearly showed differences
The statistical analysis included the comparison of between the automated flotation cell (EA) and the
averages using variance analysis (ANOVA), which also manually operated ones, when lead recoveries and
included the Bartlett evaluation test to determine the grades are compared. For Zn recoveries and grades, the
variance homogeneity (Neter et al.; 1996). A confidence differences were significant (confidence less than 95%).
level of 0.05 was selected, which means that when the However, a comparison of the variance results
summarized in Table 4, demonstrated that for the four The plot reported by Stat-Ease (Fig. 4) allowed the
machines working at same operation conditions, the determination of the significant effects, which were B
automated cell produced improved different results which (pH), and the interactions BD and ABC. These results
were very significant (confidence over 95%). were confirmed through the analysis of variance, as
shown by the p-values of Table 6. The model generated
by the software (Eq. 1) predicts values with good
correlation with the experimental results, as shown in Fig.
5 for the Zn recoveries. Figures 6, 7, 8 and 9 provides
some selected surface responses that demonstrated the
effect of the significatives variables established in this
study.

Figure 3 - Laboratory flotation machines used in this


work.
Figure 4 - Determination of significant variables
Test Conditions
After carrying out a preliminary fractional program Table 6 - ANOVA summary with p-values for the
considering 6 variables but a test number of 26-2, to model and main effects
establish main variables (effects), a detailed program
with 24 tests was designed; four additional replicas were
included to obtain experimental error of the procedure.
The standard flotation procedure considered the Edemet
automated machine EA, with a cell volume of 2.7 L. The
four operating variables, for Pb flotation, were: collector
dose, pH, ZnSO4 dose, and MIBC dose. Operating
variables maintained constant were the collector, a
mixture of Sodium Isopropyl Xanthate and
Dithiophosphate 1:1, the grinding time of 10 min (P80 =
155 µ), the air flow rate of 9 L/min, and the impeller
speed of 1000 rpm. For the Zn flotation, selected values
for the operating variables were: addition of 30 g/ton of
xanthate, 15g/ton of Dowfroth 250, 600 g/ton of CuSO4,
an air flow of 9 L/min, a pH of 10, and impeller speed of
1000 rpm.

Table 5 - Initial conditions for Pb flotation

Table 5 includes the starting conditions used in the 20


flotation tests completed. Ranges used for each variable
are easily observed. The responses (Pb and Zn
recoveries and grades) are also included in the design. Figure 5 - Predicted vs. measured Zn recovery
Predicted model
%Rec. Zn = 482.6 – 2.7 (Collector mix) – 50.1 (pH) – 0.5
(ZnSO4) – 1.7 (MIBC) + 0.3 (Collector mix ·
pH) + 3.5E-003 (Collector mix · ZnSO4) +
0.1(pH · ZnSO4) + 0.2(pH · MIBC) – 4.2E-
004(Collector mix · pH · ZnSO4) (1)

Figure 9 - Zn recovery response surface as a function


of collector and ZnSO4 dosages

The cube (Fig. 10) illustrates operating conditions


predicted by the model (Eq. 1) that maximize Zn recovery
within the ranges for the different variables initially
selected. These conditions, whose values are indicated
in Table 7, are the following: collector dosage at
minimum level, 100 g/t; Zn SO4 at the higher level, 1200
g/t; pH at the higher level, 9; and MIBC dosage at the
higher level, 80 g/t.

Figure 6 - Zn recovery response surface as a function


of pH and MIBC
Figure 10 - Model-predicted operating conditions for
maximizing Zn recovery

Table 7 - Operation conditions given by the model for


maximizing Zn recovery

Figure 7 - Zn recovery response surface as a function


of collector dosage and pH
The model was also interrogated to predict Zn recoveries
outside the initial operating range. The application of an
optimizing procedure (point prediction node) available in
the Stat-ease program (Stat-ease manual, 2010), which
allows to include factors or levels to each of the model
parameter, resulted in a new set of conditions to increase
even further the Zn recovery. Duplicate flotation tests run
at these new conditions (Zn SO4: 1200 g/t; pH: 9;
collector: 100 g/t; MIBC: 80 g/t) improved Zn recovery to
values around 80%, an increase of about 10% over the
maximum value obtained at the conditions within the
initial range, as shown in Table 8. The Zn grade obtained
went down, as expected, but a cleaning stage with tail
recirculation is considered to increase grade while
Figure 8 - Zn recovery response surface as a function maintaining the improved recovery.
of pH and ZnS04 dosage
Hamilton J, Guy P(2001). Pulp level control for flotation
Table 8 - Zn recovery and grade results obtained at — options and a CSIRO laboratory perspective.
the proposed new operating conditions CSIRO, Division of Minerals, Box 312, Australia.
Hasu V, Hätönen J, Hyötyniemi H. (2000). Analysis of
flotation froth appearance by design of experiments.
Control Engineering Laboratory, Helsinki, Report
University of Technology, IFAC.
Jämsä-Jounela S, Dietrich M, Halmevaara K, Tiili O
CONCLUSIONS (2002). Control of pulp levels in flotation cells. Control
Engineering Practice 11 (2003) 73–81.
The results obtained in this work demonstrated that the Kelsall D. (1961).“Application of probability in the
use of an automated flotation cell (EDEMET EA) were assessment of flotation systems”, Institution of Mining
significantly more reliable and reproducible than those and Metallurgy, 1961, Transactions, American Society
obtained with manually operated units. The comparison of Mining and Metallurgical Engineers, 70, pp. 191-
of the metallurgical results (recovery and grade) 204.
averages obtained in the four machines working at the Kracht W., Vallebuona G., Casali A. (2005). Rate
same operating conditions, were significantly different constant modeling for batch flotation, as a function of
and higher in the automated unit, and the comparison of gas dispersion properties. Minerals Engineering 18
the variances were very significantly lower in the (11): 1067-1076.
automated cell. Luttrell G., Yoon R. (1982).Automation of a laboratory
The results also demonstrated that the most significant flotation machine for improved performance.
variable was the pH, but the interactions pH-frother International Journal of Mineral Processing, 10 (1983)
dosage and collector dosage-pH-CuSO4 dosage were 165--172 165.Elsevier Scientific Publishing Company,
important. Amsterdam.
The model generated by the software predicted Zn Montgomery D.(2003) “Diseño y análisis de
recovery in close agreement with the values obtained experimentos”, Segunda edición, John Wiley and
from the experimental tests, within the range of the Sons, Cap. 3, pp. 60 – 116, 218 – 276, 287 – 301.
selected initial operating conditions. Model predictions Neter, J., Kutner, M.H., Nachtsheim, C.J. and
with no range restrictions for the variables involved to Wasserman, W. (1996). Applied linear statistical
optimize performance resulted in a set of conditions (Zn models. McGraw-Hill, Boston
SO4: 1200 g/t; pH: 9; collector: 100 g/t; MIBC: 80 g/t) that  Rao V. (1985). “Factorial design and statistical analysis
improved Zn recovery to values around 80%, an increase of flotation experiments”, Proceedings of National
of about 10% over the maximum value obtained at the Seminar on Mineral Processing and IX Annual
conditions within the initial range. Technical Convention of Indian Institute of Mineral
The use of an appropriate statistical package is highly Engineers, 11th & 12th March, National Metallurgical
recommended in this type of work. Laboratory (CSIR), pp. 196 – 202.
Roberts T., Firth B. and Nicol S.(1982). Note: A Modified
REFERENCES Laboratory Cell for the Flotation of Coal. International
Journal of Mineral Processing, V9 I2,191-200.Elsevier
Anderson, C. and Fayram T. (2005).The Use of Design of Scientific Publishing Company, Amsterdam.
Experimentation Software in Applied Flotation Testing. Runge K. (2010). Laboratory Flotation Testing – An
Centenary of Flotation Symposium Brisbane, QLD.63- Essential Tool for Ore Characterization. Flotation
71. Plant Optimization, C. J. Greet Editor, Spectrum 16.
Anderson C. (2006). “Applied metallurgical process AUSIMM
testing and plant optimization with design of Stat-ease manual (2010). Software Multifactor RSM
experimentation software”, TMS. Tutorial (Part 2 – Optimization).  
Apling, A. and Ersayin, S. (1986). Reproducibility of semi-   http://statease.info/dx8files/manual/DX8-04D-
batch flotation test work with the Leeds Open-Top Cell MultifactorRSM-P2.pdf
and of derived kinetic parameters. Trans. Ins. Min. Sheridan M, Nagaraj D, Fornasiero D, Ralston J.
Metall., 95, C83-C88. (2002).The use of a factorial experimental design to
Brown D, Hall S. (1999). Froth flotation: preparation of a study collector properties of N-allyl-O-alkyl
laboratory standard. Fuel 78 (1999) 1621–1629. Thionocarbamate collector in the flotation of a copper
Bulatovic S. (2007) “Handbook of flotation reagents”, ore. Minerals Engineering 15 (2002) 333–340.
Elsevier, Chapter 14, pp.323 – 366. Sutherland D. (1977).“An appreciation of galena
Craig I, Koch I. (2003).Experimental design for the concentration using a steady-state flotation model”,
economic performance evaluation of industrial International Journal of Mineral Processing, vol.4, pp.
controllers. Control Engineering Practice 11, 57-66. 149-162.
Dell C., Bunyard M. (1967).Development of an automatic Vidal S. (2012). “Selección de colector para la flotación
flotation cell for laboratory. Trans. Inst. Min. Metall., de cinc a través de pruebas en planta piloto de Minera
81: C246-C248. San Cristóbal”, V congreso de flotación, Flotamin,
Greet C, Kinal J, Small G. (2012). Laboratory flotation Viña del Mar, pp. 33 – 48.
tests… who cares?.9th International Mineral
Processing Conference, Procemin 2012.
Gutierrez C. (1973) “The mechanism of flotation of
galena by xanthates”, Miner. Sci. Engng., Vol. 5, N°2.

You might also like