minerals

Article
Application of Full Factorial Experimental Design
and Response Surface Methodology for Chromite
Beneficiation by Knelson Concentrator
Gul Akar Sen
Received: 17 November 2015; Accepted: 13 January 2016; Published: 19 January 2016
Academic Editor: Massimiliano Zanin

Department of Mining Engineering, Dokuz Eylul University, Buca 35370, Izmir, Turkey; gul.akar@deu.edu.tr;
Tel.: +90-232-301-7537

Abstract: The present work is undertaken to determine the effect of operational variables, namely:
feed rate, centrifugal force and fluidization water flow rate on the efficiency of Knelson concentrator
for chromite ore beneficiation. A full factorial design with three factors at three levels and response
surface methodology (RSM) were applied for this purpose. The quadratic models were developed to
predict the concentrate Cr2 O3 grade and recovery as the process responses. The results suggest that
all the variables affect the grade and recovery of the Cr2 O3 concentrate to some degree. However, the
fluidization water rate was found as the most effective parameter.

Keywords: Knelson concentrator; chromite ores; gravity processing; full factorial design

1. Introduction
Chromite ore (Mg,Fe)(Cr,Al,Fe)2 O4 is an oxide mineral which is the most important source of
chromium. It is used for various purposes in metallurgical, refractory and chemical industries. In
2013, South Africa was the world's dominant chromite producer, followed by Kazakhstan, India and
Turkey [1].
It is estimated that Turkey has about 31 million tons mineable reserves of high grade (30%–48%
Cr2 O3 ) chromite ore [2]. The type, amount and liberation conditions of the gangue minerals determine
the mineral processing option. The redistribution of Cr2 O3 to gangue minerals can lower the efficiency
of the physical separation plants [3]. Different beneficiation techniques such as gravity separation [4–7],
flotation [8–12] or magnetic separation [13–17] can be applied to remove gangue minerals from the
chromite concentrates.
New generation gravity processing devices such as Mozley Multi Gravity Separator (MGS), Falcon
Separator and Knelson Concentrator are promising technologies for the recovery of chromite at fine
sizes. These gravity concentrators utilize centrifugal force to enhance the relative settling velocity
between particles with different size and density [18–23].
The efficiency of centrifugal separators can be improved by adjusting the levels of operational
parameters to provide the best conditions for separation [24,25]. For this reason, the effects of
different operational parameters are crucial to understanding and controlling the separation process.
However, it is very time consuming to test the whole set of parameters at all levels. For this reason,
a well-designed experimental test program is needed to determine the response of the separation
process to each factor. As of today, factorial designs have been used successfully in many different
areas for planning of experiments to develop empirical models [26–29]. In this study, the effects of
the operating variables such as feed rate, centrifugal force and fluidization water flow rate on the
separation efficiency of Knelson concentrator were investigated using response surface methodology
(RSM) and three-level and three-factor full factorial experimental design. Full factorial design contains

Minerals 2016, 6, 5; doi:10.3390/min6010005 www.mdpi.com/journal/minerals

Minerals 2016, 6, 5 2 of 11

all possible combinations of a set of factors. The influences of all factors and interaction effects on the
responses are investigated systematically.

2. Materials and Methods

2.1. Materials
The chromite ore used in the study was supplied from Adana, Turkey. Microscopic studies
indicated harzburgite and dunite which are serpentinized to different degrees as the host rocks. The
specific gravity of the ore sample was determined as 3.2 g/cm3 using Micrometrics, Accupyc II 1340
Gas Pycnometer (Micromeritics Instrument Corporation, Norcross, GA, USA). The specific gravity of
the chromite mineral is about 4.6 g/cm3 , whereas associated gangue minerals have a specific gravity
of about 2.6 g/cm3 . These specific gravity values indicate that separation of these minerals from each
other using a gravity concentration technique is possible.
The ore sample was stage-crushed to 100% below 3.35 mm and then ground to 100% below
500 µm using a laboratory rod mill. Wet sieve analysis was carried out to determine the particle size
distribution of the ground material using a vibratory sieve shaker and micro precision sieves. The
results of size-wise chemical analysis are presented in Table 1.

Table 1. Size-wise chemical analysis results of chromite.

Particle Size (µm) Weight (%) Grade (Cr2 O3 %) Distribution (Cr2 O3 %)

500–300 5.0 17.0 3.4
300–212 5.7 20.1 4.6
212–150 8.1 20.3 6.6
150–106 7.7 24.8 7.7
106–75 10.6 32.0 13.6
75–53 11.0 32.2 14.2
53–38 11.1 31.9 14.2
´38 40.8 21.7 35.6
Total 100.0 24.9 100.0

2.2. Methods
A laboratory scale Knelson separator “KC-MD3” (FLSmidth USA Inc., Midvale, UT, USA) was
used for the experiments. The concentrator has 45 kg/h dry and 8 L/min volumetric throughput. This
manual discharge-laboratory scale separator has a 3 inches diameter truncated-cone-shaped separation
bowl. The concentrate capacity of the separation bowl varies between 80 to 150 g depending on the
fluidization conditions and heavy mineral properties. The amount of the feed material to be used in
the study was determined as 220 g considering the chromite mineral content of the ore and results of
the trial experiments. At the end of the study, the standard deviation of the weight of the concentrates
was calculated as 1.8 g. The solid ratio of the feed slurry was kept constant 23% (w/w) during each
test. The volumetric pulp flow rates corresponding to 12, 24 and 36 kg/h feed rates were 0.7, 1.4 and
2.2 L/min, respectively. The required fluidization water (FWFR) was supplied using a centrifugal
pump with 4, 8 and 12 L/min flow rates.
The concentrate and tailing of the separation tests were filtered, dried and weighed prior to
chemical analysis. Titration method (ASTM-E342-04, 1999) [30] was employed for the determination of
the Cr2 O3 content of different products.
The effect of different operating parameters on the separation efficiency of Knelson separator was
investigated using a three-level three-factor full factorial experimental design approach. The list of the
independent variables (A, B, C) with their coded and actual levels are presented in Table 2.
Minerals 2016, 6, 5 3 of 11

Table 2. Variables and levels for the three-level and three-factor full factorial design.

Real Values of Coded Levels

Variables Symbol
´1 0 +1
Fluidization water flow rate ((FWFR), L/min) A 4 8 12
G force B 60 90 120
Feed rate (kg/h) C 12 24 36
´1: factor at low level; 0: factor at medium level; +1: factor at high level.

A second order polynomial equation was chosen to fit the experimental results. This model
represents the effects of process variables (A, B, C) and their interactions on the response variables
(Cr2 O3 grade and Cr2 O3 recovery). The general form of the model chosen is represented as following:

Y “ b0 ` b1 A ` b2 B ` b3 C ` b12 AB ` b13 AC ` b23 BC ` b11 A2 ` b22 B2 ` b33 C2 (1)

where, Y is the predicted response, b0 is model constant; b1 , b2 and b3 are linear coefficients; b12 , b13
and b23 are cross product coefficients and b11 , b22 and b33 are the quadratic coefficients. Statistical
Stat-Ease Design Expert 9.0.3.1 software (Stat-Ease Inc., Minneapolis, MN, USA) was used to establish
the validity of the models on the basis of analysis of variance (ANOVA).

3. Results and Discussion

3.1. The Model Equations and Statistical Evaluation

A total of 32 tests including five control experiments (center points) were conducted during the
experimental study. The actual data collected from the tests were used to construct the empirical
models representing concentrate grade and recovery as process responds to the variables. The results
obtained from the tests are presented in Table 3.

Table 3. Design matrix and the results of the three-level and three-factor factor full factorial design.

Conditions Grade, Cr2 O3 % Recovery, Cr2 O3 %

Run Number
A B C Actual Predicted Actual Predicted
1 12 60 24 42.3 42.8 68.1 68.7
2 12 90 24 41.3 40.1 70.8 69.4
3 8 90 24 37.9 37.6 66.1 66.8
4 8 90 24 36.5 37.6 64.7 66.8
5 12 90 12 40.3 39.9 71.7 71.9
6 4 60 12 33.3 33.6 61.0 61.8
7 4 90 36 30.3 30.1 53.2 54.6
8 12 120 12 37.0 37.4 71.9 71.8
9 12 120 24 37.8 37.8 68.5 69.2
10 4 120 36 28.2 28.3 50.3 49.4
11 4 120 12 29.1 28.8 52.9 52.4
12 4 60 36 32.2 32.2 59.7 59.0
13 4 90 24 30.0 30.4 54.4 55.3
14 8 90 24 37.1 37.6 66.4 66.8
15 8 120 36 35.5 35.8 63.2 63.3
16 8 120 24 35.7 35.5 64.2 64.2
17 8 90 12 38.8 37.9 69.9 69.2
18 8 60 36 40.5 40.0 67.2 67.9
19 4 90 12 30.6 31.0 57.2 57.5
20 12 60 12 42.2 42.8 70.9 71.1
21 8 60 24 40.9 40.1 68.6 68.7
22 8 60 12 41.2 40.6 71.5 71.0
Minerals 2016, 6, 5 4 of 11

Table 3. Cont.

Conditions Grade, Cr2 O3 % Recovery, Cr2 O3 %

Run Number
A B C Actual Predicted Actual Predicted
23 12 120 36 38.5 38.5 68.0 68.2
Minerals24
2016, 6, 5 8 90 24 36.9 37.6 68.0 466.8
of 11
25 12 90 36 41.1 40.7 69.0 68.4
26 8 120 Table 3. Cont.
12 34.9 35.5 66.1 66.6
27 4 120
Conditions 24 Grade,29.2Cr2O3 % 28.4
Recovery, Cr250.1
O3 % 50.1
28 Run Number
8 90 24 37.9 37.6 66.9 66.8
A B C Actual Predicted Actual Predicted
29 8 90 24 37.5 37.6 68.6 66.8
27 4 120 24 29.2 28.4 50.1 50.1
30 8 90 36 37.5 37.7 66.6 66.0
28 8 90 24 37.9 37.6 66.9 66.8
31 12 60 36 42.7 43.1 67.7 67.8
29 8 90 24 37.5 37.6 68.6 66.8
32 4 60 24 32.8 32.7 60.9 59.6
30 8 90 36 37.5 37.7 66.6 66.0
31 12 60 36 42.7 43.1 67.7 67.8
32 4 60 24 32.8 32.7 60.9 59.6
Table 3 also shows the predicted responses of the process to the variables considered. The actual
model equation for grade (Y1 ) and recovery (Y2 ) of the chromite concentrates is given in Equations (2)
Table 3 also shows the predicted responses of the process to the variables considered. The
and actual
(3) respectively.
model equation for grade (Y1) and recovery (Y2) of the chromite concentrates is given in
Equations (2) and (3) respectively.
Y1 “ 37.64 ` 4.86A ´ 2.34B ´ 0.057C ´ 0.16AB ` 0.41AC ` 0.20BC ´ 2.37A2 ` 0.16B2 ` 0.16C2 (2)
Y1 = 37.64 + 4.86A − 2.34B − 0.057C − 0.16AB + 0.41AC + 0.20BC − 2.37A2 + 0.16B2 + 0.16C2 (2)
2 2 2
Y2 “ 66.83 ` 7.05A
Y2 = 66.83 ´ 2.25B
+ 7.05A ´−1.58C
− 2.25B 1.58C ` 2.51AB−´
+ 2.51AB 0.15AC
0.15AC ´ 0.057BC
− 0.057BC ´ 4.51A
− 4.51A ´ 0.42B
2 − 0.42B 2 + 0.78C` 0.78C2(3) (3)

TheThe
models fit fit
models thethe
data
datawell
welland
andthe
the differences between
differences between thethe observed
observed values
values and and the model's
the model's
predicted values
predicted areare
values relatively
relativelysmall
smalland
and unbiased. Thepredicted
unbiased. The predicted values
values are are in consistent
in consistent with with
the the
experimental results.
experimental results. The
Thecoefficients
coefficients of
of determination
determination (R(R2) 2 ) obtained
obtained for for
the the grade
grade and and recovery
recovery are are
0.9852 andand
0.9852 0.9830 respectively,
0.9830 respectively,indicating thatthe
indicating that theregression
regression is significant
is significant as illustrated
as illustrated in Figure
in Figure 1. 1.

Figure
Figure 1. Relationshipbetween
1. Relationship between observed
observed and
andpredicted
predictedvalues ((a)((a)
values Grade; (b) Recovery).
Grade; (b) Recovery).

Figure 2 shows graphically the residuals for predicted values of grade and recovery. As can be
Figure 2 shows graphically the residuals for predicted values of grade and recovery. As can be
seen from these figures that the residual values are uniformly distributed.
seen from these
The figures that
significance themodel
test of residual values
fit for gradeare uniformly
and recovery of distributed.
the concentrates were performed
The significance test of model fit for grade and recovery of
using Design Expert software version 9.0.3.1 based on analysis of variance the concentrates were performed
(ANOVA). The results using
Design
showed that the models are significant as the F value is high and the Prob > F value is lessresults
Expert software version 9.0.3.1 based on analysis of variance (ANOVA). The showed
than 0.05
that (Table 4). Theare
the models standard deviations
significant as theof F value
the predicted
is highmodels
and thewere > F value
Probfound 0.62 and 1.0 for
is less thangrade
0.05 and
(Table 4).
The recovery
standardrespectively.
deviationsLack of Fit
of the F-values were
predicted found
models as 1.23
were for grade
found and 0.36
0.62 and 1.0 for
forrecovery
grade andmodels
recovery
implying Lack
respectively. that the
ofLack of Fit is not
Fit F-values significant
were found relative
as 1.23 to
forthe pure and
grade error0.36
for both models. The
for recovery predicted
models implying
R2 values of grade and recovery models (0.9714 and 0.9678 respectively) are in reasonable consistency
that the Lack of Fit is not significant relative to the pure error for both models. The predicted R2 values
with the adjusted R2 values (0.9791 and 0.9761), the difference between these two values are less than
of grade and recovery models (0.9714 and 0.9678 respectively) are in reasonable consistency with the
0.2 for both models.
The significance level of the parameters, the coefficient estimates and parameter interactions of
the empirical models are tabulated in Table 5. Values of “Prob > F” less than 0.05 for the models
suggest that the model terms are significant.
Minerals 2016, 6, 5 5 of 11

adjusted R2 values (0.9791 and 0.9761), the difference between these two values are less than 0.2 for
both models.
The significance level of the parameters, the coefficient estimates and parameter interactions
of the empirical models are tabulated in Table 5. Values of “Prob > F” less than 0.05 for the models
suggest that the model terms are significant.
Minerals 2016, 6, 5 5 of 11

Figure 2. Residual plots for predicted grade and recovery values ((a) Grade; (b) Recovery).
Figure 2. Residual plots for predicted grade and recovery values ((a) Grade; (b) Recovery).
Table 4. Analysis of variance (ANOVA) table derived for the grade and recovery models.
Table 4. Analysis of variance (ANOVA) table derived for the grade and recovery models.
Source
Statistics
Cr2O3 (%) Grade Model Cr2O3 (%) Recovery Model
Source
Statistics
Sum of square 569.35 1265.55
Degree of freedom Cr2 O3 (%) Grade
9 Model Cr2 O3 (%)
9 Recovery Model
Sum of Mean
squaresquare 63.26
569.35 140.621265.55
Degree of freedom
F-Value 9
162.72 141.39 9
Mean square
Prob > F 63.26
<0.0001 <0.0001140.62
F-Value
Standard deviation 162.72
0.62 1.00 141.39
Prob > F <0.0001 <0.0001
Adjusted R 2 0.9791 0.9761
Standard deviation 2 0.62 1.00
Predicted
Adjusted R2 R 0.9714
0.9791 0.96780.9761
Predicted R2R 0.9852 0.98300.9678
2
0.9714
R2 0.9852 0.9830
Table 5. Estimated coefficient values for the parameter and parameter interaction effects.

Table 5. Estimated coefficient

Cr2Ovalues for the parameter and parameter
3 (%) Grade interaction effects.
Cr2O3 (%) Recovery
Factor
Coefficient Coefficient
F-Value Prob > F F-Value Prob > F
EstimateCr2 O3 (%) Grade Estimate Cr2 O3 (%) Recovery
A—FWFR Coefficient
4.86 1095.08 <0.0001 07.05 Coefficient
899.85 <0.0001
Factor
B—G Force Estimate
−2.34 F-Value <0.0001
254.362 Prob > F −2.25 Estimate 91.40 F-Value
<0.0001 Prob > F
C—Feed −0.057 0.145 0.70 −1.58 44.96 <0.0001
A—FWFR 4.86 1095.08 <0.0001 07.05 899.85 <0.0001
AB −0.16 0.83 0.37 2.51 75.97 <0.0001
B—G Force ´2.34 254.362 <0.0001 ´2.25 91.40 <0.0001
C—Feed AC 0.41
´0.057 5.21
0.145 0.030.70 −0.15 ´1.58 0.27 0.6115
44.96 <0.0001
AB BC 0.20
´0.16 1.28
0.83 0.270.37 −0.057 2.51 0.04 0.8435
75.97 <0.0001
AC A2 −2.37
0.41 103.18
5.21 <0.0001
0.03 −4.51 ´0.15146.43 <0.0001
0.27 0.6115
BC B2 0.20
0.16 1.28
0.45 0.510.27 −0.42 ´0.0571.27 0.04
0.2719 0.8435
A2 C2 ´2.37
0.16 103.18
0.48 <0.0001
0.49 0.78 ´4.51 4.41 146.43
0.0474 <0.0001
B2 0.16 0.45 0.51 ´0.42 1.27 0.2719
C2 0.16 0.48 0.49 0.78 4.41 0.0474
3.2. The Effect of Process Variables on Grade of the Knelson Separator Concentrate
The estimated coefficient values calculated for main effects of the process variables from the
3.2. The Effect of Process Variables on Grade of the Knelson Separator Concentrate
Equation 2 (Table 5) indicate that fluidization water flow rate (A) has a positive effect and G force (B)
The
has aestimated coefficient
negative effect on the Crvalues calculated
2O3 content for main effects
of the concentrate. It was of
alsothe processthat
observed variables
feed ratefrom
(C) the
has no significant effect on the grade of the concentrate in the range of the parameter levels tested.
Equation 2 (Table 5) indicate that fluidization water flow rate (A) has a positive effect and G force (B)
The Equation 2 dictates that the interaction of variables, fluidization water flow rate (A)—G force
(B) has a negative effect on the Cr2O3 content of the concentrate whereas interactions of fluidization
water flow rate (A)—feed rate (C) and G force (B)—feed rate (C) have a positive effect. On the other
hand, fluidization water flow rate (A) has the highest quadratic effect on the concentrate Cr2O3 grade.
Minerals 2016, 6, 5 6 of 11

has a negative effect on the Cr2 O3 content of the concentrate. It was also observed that feed rate (C)
has no significant effect on the grade of the concentrate in the range of the parameter levels tested.
The Equation 2 dictates that the interaction of variables, fluidization water flow rate (A)—G force
(B) has a negative effect on the Cr2 O3 content of the concentrate whereas interactions of fluidization
water flow rate (A)—feed rate (C) and G force (B)—feed rate (C) have a positive effect. On the other
hand, fluidization water flow rate (A) has the highest quadratic effect on the concentrate6 ofCr
Minerals 2016, 6, 5
O grade.
112 3
The predicted response (Y1 ) of the process to the independent variables are presented graphically
Minerals
The2016, 6, 5
predicted response (Y1) of 6 of 11
in three dimensional (3D) space to help to the processthe
visualize to shape
the independent variablessurface
of the response are presented
in Figures 3–5.
graphically in three dimensional (3D) space to help to visualize the shape of the response surface in
The effectThe of the G force
predicted and fluidization
response water flow
(Y1) of the process to therate on gradevariables
independent (Cr2 O3 are%) of the concentrate
presented
Figures 3–5.
graphically
at the center level in feed
of three dimensional
rate is shown(3D) space
in to help3.
Figure to Asvisualize
can the shape
be seen of
fromthe response
Figure surface
3, in
a higher grade
The effect of the G force and fluidization water flow rate on grade (Cr2O3 %) of the concentrate
Figures 3–5.
concentrateat is
theobtained
center levelatofhigher fluidization
feed rate water3.flow
is shown in Figure As canrate
be and FigureG3,force.
at lower
seen from The
a higher grade of the
grade
The effect of the G force and fluidization water flow rate on grade (Cr2O3 %) of the concentrate
concentrateconcentrate
tends to isdecrease
obtained atashigher
the fluidization
applied G water
force flow rate and This
increases. at lower
can G beforce. The gradeby
explained of the
the fact that
at the center level of feed rate is shown in Figure 3. As can be seen from Figure 3, a higher grade
concentrate tends to decrease as the applied G force increases. This can be explained by the fact that
the cut-density of the separation is reduced with increasing G
concentrate is obtained at higher fluidization water flow rate and at lower G force. The grade of the the lower
force which in turn allows
the cut-density of the separation is reduced with increasing G force which in turn allows the lower
concentrate
grade composite tends toto
particles decrease
enter as
thethethe
applied G force
concentrate increases. Thisdecrease
can be explained Crby
2 Othe fact that
grade composite particles to enter concentrate product
product and anddecrease the Cr the
2O3 grade. 3 grade.
the cut-density of the separation is reduced with increasing G force which in turn allows the lower
grade composite particles to enter the concentrate product and decrease the Cr2O3 grade.

Figure
Figure 3. The 3. The response
response surface
surface plotplot representingthe
representing the effect
effectofoffluidization waterwater
fluidization flow rate andrate
flow applied
and applied
G force on the Cr2O3 content of the concentrate.
G force on Figure
the Cr3.2The
O3 response
content surface
of theplot
concentrate.
representing the effect of fluidization water flow rate and applied
G force on
Figure the Cr2O3 content
4 illustrates of the
the effect of concentrate.
feed rate and fluidization water flow rate on grade (Cr2O3 %) of
the concentrate
Figure 4 illustrates at the center
the effect level
of feed of G force.
rate and It is seen from Figure
fluidization 4 that there is not grade
any significant
Figure 4 illustrates the effect of feed rate and fluidization water flow
water flow rate
rate on ongrade (Cr2O(Cr
3 %)2O
of3 %) of the
effect of feed rate on the grade of the concentrate at a constant water flow rate in the range of the
concentrate at the center level of G force. It is seen from Figure 4 that there is not
the concentrate at the center level of G force. It is seen from Figure 4 that there is not any significant any significant effect
parameter levels tested. However, increasing the fluidization water flow rate leads to obtaining
effect
of feed ratehigher of
on the feed rate
grade on the grade
of the due of
concentratethe concentrate
at a constant at a constant water flow rate in the range of the
grade concentrate to the increased number of water flow rate
light particles in the to
transported range of theofparameter
the outside
parameter levels tested. However, increasing the fluidization water flow rate leads to obtaining
levels tested. However,
separation cone. increasing the fluidization water flow rate leads to obtaining higher grade
higher grade concentrate due to the increased number of light particles transported to the outside of
concentrate due to the
separation cone. increased number of light particles transported to the outside of separation cone.

Figure 4. The response surface plot representing the effect of fluidization water flow rate and feed
rate on the Cr2O3 content of the concentrate.
Figure
Figure 4. The 4. The response
response surfacesurface plot representing
plot representing thetheeffect
effectof
of fluidization
fluidization water
waterflow rate rate
flow and feed
and feed rate
rate on the Cr2O3 content of the concentrate.
on the Cr2 O3 content of the concentrate.
Minerals 2016, 6, 5 7 of 11
Minerals 2016, 6, 5 7 of 11

Figure 55 represents
Figure represents the
the effect of G
effect of G force
force and
and feed
feed rate
rate on
on grade
grade (Cr(Cr22O %) of
O33 %) of the
the Knelson
Knelson
concentrate at the center level of fluidization water flow rate. It is observed from Figure 55 that
concentrate at the center level of fluidization water flow rate. It is observed from Figure that
changing feed
changing feed rate
rate has
has no
no significant
significant effect
effect on
on the
the concentrate grade in
concentrate grade in the
the range of the
range of the parameter
parameter
levels tested.
levels tested. On
On the
the other
other hand,
hand, the
the use
use of
of higher G force
higher G force at
at aa constant
constant feed
feed rate
rate results
results in
in aa decrease
decrease
in concentrate Cr O grade.
in concentrate Cr22O33 grade.

Figure 5. The response surface plot representing the effect of G force and feed rate on the Cr2O3 content
Figure 5. The response surface plot representing the effect of G force and feed rate on the Cr2 O3 content
of the concentrate.
of the concentrate.

3.3. Effect of KC-MD3 Knelson Concentrator Variables on Concentrate Recovery

3.3. Effect of KC-MD3 Knelson Concentrator Variables on Concentrate Recovery
The estimated coefficient and Prob > F values obtained for main effects of the independent
The estimated coefficient and Prob > F values obtained for main effects of the independent
variables from Equation 3 indicate that water flow rate (A) has a positive effect on the Cr2O3 recovery
variables from Equation 3 indicate that water flow rate (A) has a positive effect on the Cr2 O3 recovery
of the concentrate, whereas, G force (B) and feed rate (C) negatively affect the Cr2O3 recovery.
of the concentrate, whereas, G force (B) and feed rate (C) negatively affect the Cr2 O3 recovery.
The interaction of variables fluidization water flow rate (A)—G force (B) has a significant positive
The interaction of variables fluidization water flow rate (A)—G force (B) has a significant positive
effect on the Cr2O3 recovery of the concentrate, while interactions of fluidization water flow rate (A)—
effect on the Cr2 O3 recovery of the concentrate, while interactions of fluidization water flow rate
feed rate (C) and G force (B)—feed rate (C) have a negative effect. It is also important to note that
(A)—feed rate (C) and G force (B)—feed rate (C) have a negative effect. It is also important to note that
fluidization water flow rate (A) has a significant quadratic effect on the concentrate Cr2O3 recovery
fluidization water flow rate (A) has a significant quadratic effect on the concentrate Cr2 O3 recovery
compared to the other independent variables.
compared to the other independent variables.
Figure 6 illustrates the effect of G force and fluidization water flow rate on the recovery of Cr2O3
Figure 6 illustrates the effect of G force and fluidization water flow rate on the recovery of Cr2 O3
in the Knelson concentrate. Feed rate was kept constant at the center level during this test. At low
in the Knelson concentrate. Feed rate was kept constant at the center level during this test. At low
level of fluidization water rate, as the G force increases, recovery of Cr2O3 decreases significantly due
level of fluidization water rate, as the G force increases, recovery of Cr2 O3 decreases significantly due
to inefficient fluidization of the separation layer inside the bowl. The highest recovery value was
to inefficient fluidization of the separation layer inside the bowl. The highest recovery value was
obtained by increasing fluidization water level to maximum for providing the appropriate conditions
obtained by increasing fluidization water level to maximum for providing the appropriate conditions
of fluidization.
of fluidization.
Figure 7 shows the effect of fluidization water flow rate and feed rate on recovery of Cr2O3 in
Figure 7 shows the effect of fluidization water flow rate and feed rate on recovery of Cr O3 in the
the concentrate of Knelson separator at the center level of G force. As can be seen from 2Figure 7,
concentrate of Knelson separator at the center level of G force. As can be seen from Figure 7, increasing
increasing fluidization water rate provides better fluidization conditions in the separation layer and
fluidization water rate provides better fluidization conditions in the separation layer and helps to
helps to increase Cr2O3 recovery of concentrate. It should be noted that there is a significant decrease
increase Cr2 O3 recovery of concentrate. It should be noted that there is a significant decrease in the
in the recovery of Cr2O3 at high feed rate when the fluidization water rate is low.
recovery of Cr2 O3 at high feed rate when the fluidization water rate is low.
Figure 8 demonstrates the effect of feed rate and G force on Cr2O3 recovery of the concentrate at
Figure 8 demonstrates the effect of feed rate and G force on Cr2 O3 recovery of the concentrate at the
the center level of fluidization water flow rate. As the G force decreases the transport of the
center level of fluidization water flow rate. As the G force decreases the transport of the lighter particles
lighter particles to the tailing fraction degreases which in increases recovery of Cr2O3 to the
to the tailing fraction degreases which in increases recovery of Cr2 O3 to the concentrate fraction.
concentrate fraction.
Figure 8 was plotted representing the recovery of Cr O concentrate vs. feed rate and G force. The
Figure 8 was plotted representing the recovery of 2Cr32O3 concentrate vs. feed rate and G force.
Cr2 O3 recovery of the concentrate increases with decrease in feed rate and G force. The lowest level of
The Cr2O3 recovery of the concentrate increases with decrease in feed rate and G force. The lowest
both independent variables appears to be providing favorable conditions for an efficient separation at
level of both independent variables appears to be providing favorable conditions for an efficient
separation at this level of fluidization water flow. Increasing any of these variables to a higher level
seems to change the fluidization conditions and lower the Cr2O3 recovery due to the disruption of the
separation layer.
Minerals 2016, 6, 5 8 of 11

this level of fluidization water flow. Increasing any of these variables to a higher level seems to change
the fluidization conditions
Minerals 2016, 6, 5 and lower the Cr2 O3 recovery due to the disruption of the separation
8 of 11 layer.
Minerals 2016, 6, 5 8 of 11
Minerals 2016, 6, 5 8 of 11

Figure 6. The response surface plot representing the effect of fluidization water flow rate and applied
Figure 6. The response surface plot representing the effect of fluidization water flow rate and applied
G force6.on
Figure theresponse
The Cr2O3 recovery
surface of therepresenting
plot concentrate.the effect of fluidization water flow rate and applied
G force on the Cr2 O3 recovery of the concentrate.
G force 6.
Figure onThe
theresponse
Cr2O3 recovery
surfaceof therepresenting
plot concentrate. the effect of fluidization water flow rate and applied
G force on the Cr2O3 recovery of the concentrate.

Figure 7. The response surface plot representing the effect of fluidization water flow rate and feed
rate on the Cr2response
O3 recovery of theplot
concentrate.
Figure 7. The 7.
Figure The
response surface
surface representing
plot representing theeffect
the effectof
offluidization
fluidization water
waterflow rate
flow andand
rate feedfeed rate
rate on 7.
Figure theThe
Cr2O 3 recovery
response of theplot
surface concentrate.
representing the effect of fluidization water flow rate and feed
on the Cr2 O3 recovery of the concentrate.
rate on the Cr2O3 recovery of the concentrate.

Figure 8. The response surface plot representing the effect of G force and feed rate on the Cr2O3
recovery
Figure 8. of
The theresponse
concentrate.
surface plot representing the effect of G force and feed rate on the Cr2O3
recovery of the concentrate.
Figure 8. The response surface plot representing the effect of G force and feed rate on the Cr2O3
Figure 8. The response
recovery surface plot representing the effect of G force and feed rate on the Cr2 O3
of the concentrate.
recovery of the concentrate.
Minerals 2016, 6, 5 9 of 11

3.4. Validation of the Models

For validation
Minerals 2016, 6, 5of the obtained models for predicting the grade and recovery of Cr2 O3 9content
of 11 in
concentrate, nine additional tests were conducted using different levels of independent variables. The
3.4. Validation of the Models
test conditions and results are presented in Table 6.
For validation of the obtained models for predicting the grade and recovery of Cr2O3 content in
concentrate,Table
nine 6.
additional tests were
The conditions andconducted
actual andusing different
predicted levels
results of independent
of the variables. The
validation tests.
test conditions and results are presented in Table 6.
Conditions Grade, Cr2 O3 % Recovery, Cr2 O3 %
Run Number Table 6. The conditions and actual and predicted results of the validation tests.
A B C Actual Predicted Actual Predicted
Conditions Grade, Cr2O3 % Recovery, Cr2O3 %
1
Run Number 4 31.8 33.1 57.5 60.5
2 A8 B 60 C Actual Predicted 40.3 Actual72.7
41.3 Predicted
69.6
31 4 12 31.8 41.8 33.1 42.8 57.5 69.1 60.5
69.7
2 8 60 41.3 40.3 72.7 69.6
43 12 4 41.8 29.5 42.8 30.7 69.1 53.0 56.2
69.7
54 18
48 90 29.5 37.6 30.7 37.7 53.0 68.7 67.8
56.2
65 812 90 18 37.6 40.5 37.7 40.0 68.7 72.6 70.4
67.8
76 12 4 40.5 28.7 40.0 28.6 72.6 50.8 70.4
51.0
87 48 120 28.7 35.1 28.6 35.4 50.8 64.5 51.0
65.2
98 812 120 35.1 41.1 35.4 37.5 64.5 72.3 65.2
70.3
9 12 41.1 37.5 72.3 70.3

The The
observed
observed results and
results andthe
thepredicted resultsfor
predicted results forboth
bothresponses
responses
are are illustrated
illustrated in Figure
in Figure 9. 9.
2
Figure 9 shows
Figure 9 shows that
thatthe
the predicted values
predicted values areare consistent
consistent with
with the the observed
observed values,
values, with 0.9254R of
R2 of with
2 of 0.9635 percent for recovery of chromite concentrates.
0.9254 percent
percent for for grade
grade R2 ofR0.9635
and and percent for recovery of chromite concentrates.

Figure 9. Relationship between actual and predicted results of the validation tests. ((a) Grade; (b) Recovery)
Figure 9. Relationship between actual and predicted results of the validation tests. ((a) Grade;
(b) Recovery)
3.5. Optimization
The optimized independent variable levels for making the grade and recovery values maximum
3.5. Optimization
were also determined using the optimization module of Design Expert version 9.0.3.1. The numerical
The optimized
optimization independent
option variable
of the module levels for
was utilized by making
employingthethe
grade and recovery
developed models values maximum
to search the
factor space for the best trade-offs to achieve the highest grade and recovery in
were also determined using the optimization module of Design Expert version 9.0.3.1. The numericalthe range of the
experimental
optimization optiondata. Themodule
of the results was
showed that by
utilized a concentrate
employing the assaying 43% Cr
developed 2O3 with
models 72% Crthe
to search 2O3factor

recovery can be obtained by adjusting the variables to the following levels; fluidization water
space for the best trade-offs to achieve the highest grade and recovery in the range of the experimental rate of
11 L/min, G force of 60 G, feed rate of 12 kg/h.
data. The results showed that a concentrate assaying 43% Cr2 O3 with 72% Cr2 O3 recovery can be
obtained by adjusting the variables to the following levels; fluidization water rate of 11 L/min, G force
4. Conclusions
of 60 G, feed rate of 12 kg/h.
In this study, the effect of feed rate (kg/h), G force and fluidization water flow rate (L/min) on
the chromite beneficiation process by KC-MD3 Knelson concentrator was investigated. Full factorial
4. Conclusions
experimental design and response surface methodology were used to develop mathematical models
In this
for bothstudy, the recovery
grade and effect ofof
feed rate
Cr2O (kg/h), GMathematical
3 concentrate. force and fluidization waterwere
model equations flow rate (L/min)
derived using on
the chromite
experimental data and mathematical software package Design Expert 9.0.3.1. It was found that factorial
beneficiation process by KC-MD3 Knelson concentrator was investigated. Full the
experimental design and response surface methodology were used to develop mathematical models
Minerals 2016, 6, 5 10 of 11

for both grade and recovery of Cr2 O3 concentrate. Mathematical model equations were derived using
experimental data and mathematical software package Design Expert 9.0.3.1. It was found that the
predicted values were in reasonable consistency with the experimental data which were also confirmed
by the validation experiments. The ANOVA results indicated that fluidization water flow rate was the
dominant independent variable affecting both grade and recovery of the concentrate. The results of
the numerical optimization in the range of the experimental data showed that it is possible to produce
a concentrate assaying 43% Cr2 O3 with 72% Cr2 O3 recovery by operating Knelson concentrator at the
fluidization water rate of 11 L/min, G force of 60 G and feed rate of 12 kg/h.

Conflicts of Interest: The authors declare no conflict of interest.

References
1. U.S. Geological Survey. Mineral Commodity Summaries. Available online: http://minerals.usgs.gov/
minerals/pubs/mcs/2015/mcs2015.pdf (accessed on 8 November 2015).
2. Cicek, T.; Cocen, I.; Engin, V.T.; Cengizler, H.; Sen, S. Technical and economical applicability study of
centrifugal force gravity separator (MGS) to Kef chromite concentration plant. Miner. Process. Extr. Metall.
2008, 117, 248–255. [CrossRef]
3. Grieco, G.; Pedrotti, M.; Moroni, M. Metamorphic redistribution of Cr within chromitites and its influence
on chromite ore enrichment. Miner. Eng. 2011, 24, 102–107. [CrossRef]
4. Kapure, G.; Kari, C.; Rao, S.M.; Rao, N.D. The feasibility of a slip velocity model for predicting the enrichment
of chromite in a Floatex density separator. Int. J. Miner. Process. 2007, 82, 86–95. [CrossRef]
5. Kumar, C.R.; Tripathy, S.K.; Rao, D.S. Characterisation and pre-concentration of chromite values from plant
tailings using floatex density separator. J. Miner. Mater. Charact. Eng. 2009, 8, 367–378. [CrossRef]
6. Tripathy, S.K.; Ramamurthy, Y.; Singh, V. Recovery of chromite values from plant tailings by gravity
concentration. J. Miner. Mater. Charact. Eng. 2011, 10, 13–25. [CrossRef]
7. Tripathy, S.K.; Ramamurthy, Y. Modeling and optimization of spiral concentrator for separation of ultrafine
chromite. Powder Technol. 2012, 221, 387–394. [CrossRef]
8. Guney, A.; Onal, G.; Celik, M.S. A new flowsheet for processing chromite fines by column flotation and the
collector adsorption mechanism. Miner. Eng. 1999, 12, 1041–1049. [CrossRef]
9. Feng, D.; Aldrich, C. Adsorption of heavy metals by biomaterials derived from the marine alga Ecklonia
maxima. Hydrometallurgy 2004, 73, 1–10. [CrossRef]
10. Gallios, G.P.; Deliyanni, E.A.; Pelek, E.N.; Matis, K.A. Flotation of chromite and serpentine. Sep. Purif.
Technol. 2007, 55, 232–237. [CrossRef]
11. Smith, R.W.; Allards, S.G. Effects of pretreatment and aging on chromite flotation. Int. J. Miner. Process. 1983,
11, 163–174. [CrossRef]
12. Sysila, S.; Laapas, H.; Heiskanen, K.; Ruokonen, E. The effect of surface potential on the flotation of chromite.
Miner. Eng. 1996, 9, 519–525. [CrossRef]
13. Aslan, N.; Kaya, H. Beneficiation of chromite concentration waste by multi-gravity separator and high
intensity induced-roll magnetic separator. Arab. J. Sci. Eng. 2009, 34, 285–297.
14. Tripathy, S.K.; Murthy, Y.R.; Singh, V. Characterisation and separation studies of Indian chromite beneficiation
plant tailing. Int. J. Miner. Process. 2013, 122, 47–53. [CrossRef]
15. Ucbas, Y.; Bozkurt, T.V.; Bilir, K.; Ipek, H. Separation of chromite from serpentine in fine sizes using magnetic
carrier. Sep. Purif. Technol. 2014, 49, 946–956. [CrossRef]
16. Ucbas, Y.; Bozkurt, T.V.; Bilir, K.; Ipek, H. Concentration of chromite by means of magnetic carrier using
sodium oleate and other reagents. Physicochem. Probl. Miner. Process. 2014, 50, 767–782.
17. Tripathy, S.K.; Banerjee, P.K.; Suresh, N. Magnetic separation studies on ferruginous chromite fine to enhance
Cr:Fe ratio. Int. J. Miner. Metall. Mater. 2015, 22, 217–224. [CrossRef]
18. Gence, N. Beneficiation of Elazığ-Kefdağı chromite by multi-gravity separator. Turk. J. Eng. Environ. Sci.
1993, 23, 473–475.
19. Knelson, B.; ve Jones, R. A new generation of Knelson concentrators a totally secure system goes on line.
Miner. Eng. 1993, 7, 201–207. [CrossRef]
Minerals 2016, 6, 5 11 of 11

20. Guney, A.; Önal, G.; Atmaca, T. New aspect of chromite gravity tailings re-processing. Miner. Eng. 2001, 14,
1527–1530. [CrossRef]
21. Sen, S. Gold recovery by KC from grinding circuit of Bergama CIP plant. Rem Rev. Esc. Minas 2010, 63,
539–545. [CrossRef]
22. Uslu, T.; Sahinoglu, E.; Yavuz, M. Desulphurization and deashing of oxidized fine coal by Knelson
concentrator. Fuel Process. Technol. 2012, 101, 94–100. [CrossRef]
23. Kökkılıç, O.; Langlois, R.; Waters, K.E. A design of experiments investigation into dry separation using a
Knelson concentrator. Miner. Eng. 2015, 72, 73–86. [CrossRef]
24. Aslan, N. Multi-objective optimization of some process parameters of a multi-gravity separator for chromite
concentration. Sep. Purif. Technol. 2008, 64, 237–241. [CrossRef]
25. Altun, N.E.; Sakuhuni, G.; Klein, B. The use of continuous centrifugal gravity concentration in grinding
circuit. Modified approach for improved metallurgical performance and reduced grinding requirements.
Physicochem. Probl. Miner. Process. 2015, 51, 115–126.
26. Obeng, D.P.; Morrell, T.S.; Napier-Munn, T.J. Application of central composite rotatable design to modelling
the effect of some operating variables on the performance of the three-product cyclone. Int. J. Miner. Process.
2005, 76, 181–192. [CrossRef]
27. Dehghan, R.; Noaparast, M.; Kolahdoozan, M.; Mousavi, S.M. Statistical evaluation and optimization of
factors affecting the leaching performance of a sphalerite concentrate. Int. J. Miner. Process. 2008, 89, 9–16.
[CrossRef]
28. Sharma, S.; Sharma, N.; Gupta, G.D. Optimizing of promethazine theoclate fast dissolving tablet using pore
forming technology by 3-factor, 3-level response surface-full factorial design. Arch. Pharm. Res. 2010, 33,
1199–1207. [CrossRef] [PubMed]
29. Martinez, A.L.; Uribe, A.S.; Carrillo, F.R.P.; Coreno, J.A.; Ortiza, J.C. Study of celestite flotation efficiency
using sodium dodecyl sulfonate collector: Factorial experiment and statistical analysis of data. Int. J. Miner.
Process. 2003, 70, 83–97. [CrossRef]
30. ASTM, E342-04. Standard Test Method for Determination of Chromium Oxide in Chrome Ores by Permanganate
Titrimetry; American Society for Testing and Materials (ASTM): West Conshohocken, PA, USA, 1999.

© 2016 by the author; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons by Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).