Application of response surface methodology and central

composite rotatable design for modeling the inuence of some

operating variables of a Multi-Gravity Separator for coal cleaning
N. Aslan
*
Mining Engineering Department, Cumhuriyet University, 58140 Sivas, Turkey
Received 9 March 2006; received in revised form 16 October 2006; accepted 31 October 2006
Available online 27 November 2006
Abstract
In this study, the application of response surface methodology (RSM) and central composite rotatable design (CCRD) for modeling
the inuence of some operating variables on the performance of a Multi-Gravity Separator (MGS) for coal cleaning was discussed. Four
operating variables of MGS, namely drum speed, tilt angle, wash water and feed solids were changed during the tests based on the
CCRD.
In order to produce clean coal with MGS, mathematical model equations were derived by computer simulation programming apply-
ing least squares method using MATLAB 7.1. These equations that are second-order response functions representing ash content and
combustible recovery of clean coal were expressed as functions of four operating parameters of MGS. Predicted values were found to be
in good agreement with experimental values (R
2
values of 0.84 and 0.93 for ash content and combustible recovery of clean coal,
respectively).
This study has shown that the CCRD and RSM could eciently be applied for the modeling of MGS for coal and it is economical
way of obtaining the maximum amount of information in a short period of time and with the fewest number of experiments.
2006 Elsevier Ltd. All rights reserved.
Keywords: Central composite rotatable design; Response surface methodology; MGS
1. Introduction
The Multi-Gravity Separator (MGS) represents the lat-
est development in ne grain mineral concentration. The
parameters that aect the performance of MGS are the
drum speed, tilt angle, shakes amplitude, shakes frequency,
wash water and feed solids [1]. The success of concentra-
tion with MGS depends on the selection of suitable param-
eter levels and minerals. The optimization of these
parameters requires many tests. The total number of exper-
iments required can be reduced depending on the experi-
mental design technique [2].
Process engineers want to determine the levels of the
design parameters at which the response reaches its opti-
mum. The optimum could be either a maximum or a min-
imum of a function of the design parameters. One of the
methodologies for obtaining the optimum results is
response surface methodology (RSM) [3].
It is essential that an experimental design methodology
is very economical for extracting the maximum amount
of complex information, a signicant experimental time
saving factor and moreover, it saves the material used for
analyses and personal costs [4].
The objective of this study was to establish the func-
tional relationships between the some operating parameters
of MGS, namely drum speed, tilt angle, wash water and
feed solid and, ash content and combustible recovery of
clean coal for Yenicubuk/Turkey lignite coal. In the
0016-2361/$ - see front matter 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.fuel.2006.10.020
*
Tel.: +90 346 2191010x1574; fax: +90 346 2191173.
E-mail address: naslan@cumhuriyet.edu.tr
www.fuelrst.com
Fuel 86 (2007) 769776
following sections, the RSM and requirements for CCRD
and its applications for modeling the inuence of some
operating variables on the performance of a MGS for coal
from Yenicubuk/Turkey lignite coal are discussed.
2. Response surface methodology (RSM)
RSM is a collection of statistical and mathematical
methods that are useful for the modeling and analyzing
engineering problems. In this technique, the main objective
is to optimize the response surface that is inuenced by var-
ious process parameters. RSM also quanties the relation-
ship between the controllable input parameters and the
obtained response surfaces [3].
The design procedure for the RSM is as follows [5]:
(i) Designing of a series of experiments for adequate and
reliable measurement of the response of interest.
(ii) Developing a mathematical model of the second-
order response surface with the best ttings.
(iii) Finding the optimal set of experimental parameters
that produce a maximum or minimum value of
response.
(iv) Representing the direct and interactive eects of pro-
cess parameters through two and three-dimensional
(3D) plots.
If all variables are assumed to be measurable, the
response surface can be expressed as follows:
y f x
1
; x
2
; x
3
; . . . ; x
k
1
where y is the answer of the system, and x
i
the variables of
action called factors.
The goal is to optimize the response variable (y). It is
assumed that the independent variables are continuous
and controllable by experiments with negligible errors. It
is required to nd a suitable approximation for the true
functional relationship between independent variables
and the response surface [5].
3. Central composite rotatable design (CCRD)
The experimental design techniques commonly used for
process analysis and modeling are the full factorial, partial
factorial and central composite rotatable designs. A full
factorial design requires at least three levels per variable
to estimate the coecients of the quadratic terms in the
response model. Thus for the four independent variables
81 experiments plus replications would have to be con-
ducted [6]. A partial factorial design requires fewer experi-
ments than the full factorial. However, the former is
particularly useful if certain variables are already known
to show no interaction [7,8].
An eective alternative to the factorial design is the cen-
tral composite rotatable design (CCRD), originally devel-
oped by Box and Wilson [6] and improved upon by Box
and Hunter [9]. The CCRD gives almost as much informa-
tion as a three-level factorial, requires much fewer tests
than the full factorial and has been shown to be sucient
to describe the majority of steady-state process responses
[8,10,11].
The number of tests required for the CCRD includes the
standard 2
k
factorial with its origin at the center, 2k points
xed axially at a distance, say b, from the center to gener-
ate the quadratic terms, and replicate tests at the center;
where k is the number of variables. The axial points are
chosen such that they allow rotatability [9], which ensures
that the variance of the model prediction is constant at
all, points equidistant from the design center. Replicates
of the test at the center are very important as they provide
an independent estimate of the experimental error. For
four variables, the recommended number of tests at the
center is six [9]. Hence the total number of tests required
for the four independent variables is 2
4
+ (2 4) + 6 = 30
[8,9].
Once the desired ranges of values of the variables are
dened, they are coded to lie at 1 for the factorial points,
0 for the center points and b for the axial points. The
codes are calculated as functions of the range of interest
of each factor as shown in Table 1.
When the response data are obtained from the test
work, a regression analysis is carried out to determine the
coecients of the response model (b
1
, b
2
, . . . , b
n
), their stan-
dard errors and signicance. In addition to the constant
(b
0
) and error (e) terms, the response model incorporates
[8]:
Linear terms in each of the variables (x
1
, x
2
, . . . , x
n
).
Squared terms in each of the variables x
2
1
; x
2
2
; . . . ; x
2
n
.
First-order interaction terms for each paired combina-
tion (x
1
x
2
, x
1
x
3
, . . . , x
ni
x
n
).
Thus for the four variables under consideration, the
response model is
y b
0
e
X
4
i1
b
i
x
i

X
4
i1
b
ii
x
2
i

X
4
i1
X
4
ji1
b
ij
x
i
x
j
2
The b coecients, which should be determined in the
second-order model, are obtained by the least squares
method. In general Eq. (2) can be written in matrix form
Y bX e 3
Table 1
Relationship between coded and actual values of a variable [12]
Code Actual value of variable
b x
min
1 [(x
max
+ x
min
)/2] [(x
max
x
min
)/2a]
0 (x
max
+ x
min
)/2
+1 [(x
max
+ x
min
)/2] + [(x
max
x
min
)/2a]
+b x
max
x
max
and x
min
= maximum and minimum values of x, respectively;
a = 2
k/4
; k = number of variables.
770 N. Aslan / Fuel 86 (2007) 769776
where Y is dened to be a matrix of measured values and X
to be a matrix of independent variables. The matrices b and
e consist of coecients and errors, respectively. The solu-
tion of Eq. (3) can be obtained by the matrix approach
[3,5].
b X
0
X
1
X
0
Y 4
where X
0
is the transpose of the matrix X and (X
0
X)
1
is
the inverse of the matrix X
0
X.
The coecients, i.e. the main eect (b
i
) and two-factors
interactions (b
ij
) can be estimated from the experimental
results by computer simulation programming applying
least squares method using MATLAB 7.1.
4. Experimental procedure
The central composite rotatable design (CCRD) was
chosen to determine the relationship between four operat-
ing variables of MGS, namely drum speed, tilt angle, wash
water and feed solid and ash content and combustible
recovery of clean coal.
The coal sample, which is lignite, was taken from the
region of Yenicubuk/Turkey. The sample containing
36.1% ash was ground to <500 lm using impact type
crusher and rod mill for MGS tests. The batch MGS tests
were conducted at the mineral processing laboratory of
Cumhuriyet University/Turkey. For each MGS test,
1500 g of the dry coal sample was used. A peristaltic pump
at ow rate of 2 lpm carried out for feeding. Drum speed,
tilt angle, wash water and feed solid were changed during
the tests based on the central composite experimental
design, whilst the other operating parameters of MGS were
kept constant namely 15 mm of shakes amplitude and
4.8 cps of shakes frequency.
The experimental setup used for this study is presented
in Fig. 1. The setup consisted of a feed slurry tank with a
stirrer, a peristaltic pump for supplying feed to the MGS
at consistent rates, a laboratory/pilot scale MGS unit and
sample containers for collecting the clean coal and tailings.
To obtain the required feed solids, measured quantities of
solids and water were mixed in the slurry tank. The MGS
variables were adjusted at the required levels as per the cen-
tral composite experimental design. The feed slurry was
pumped into the MGS drum at the required ow rate using
the peristaltic pump while the MGS was in operation. Sam-
ples from the clean coal and tailing streams were collected
at steady-state conditions. The samples were ltered, dried
and analyzed for ash content and combustible recovery.
The combustible recovery was calculated using Eq. (5).
Combustible recovery; %
M
c
1 A
c

M
f
1 A
f

100 5
where A
c
is ash content of clean coal, A
f
is ash content of
feed, M
c
is mass of clean coal and M
f
is mass of feed.
5. Results and discussion
A four-factor and ve-coded level CCRD was used to
determine the responses (ash content and combustible
recovery of clean coal). The four variables of MGS were
drum speed, tilt angle, wash water and feed solids. The
number of tests at the center points was six, making the
total number of tests required for the four independent
variables (drum speed, tilt angle, wash water and feed sol-
ids): 2
4
+ (2 4) + 6 = 30 [8].
The drum speed (v), tilt angle (a), wash water (w) and
feed solid (s) were independent variables studied to predict
y responses (ash content and combustible recovery of clean
coal). The four independent variables and their levels for
the CCRD used in this study are shown in Table 2.
Using the relationships in Table 2, coded and actual lev-
els of the variables for each of the experiments in the design
matrix were calculated as given in Table 3.
Considering the eects of main factors and also the
interactions between two-factor, Eq. (2) takes the form:
y b
0
b
1
x
1
b
2
x
2
b
3
x
3
b
4
x
4
b
11
x
2
1
b
22
x
2
2
b
33
x
2
3
b
44
x
2
4
b
12
x
1
x
2
b
13
x
1
x
3
b
14
x
1
x
4
b
23
x
2
x
3
b
24
x
2
x
x
b
34
x
3
x
4
6
The coecients, i.e. the main eect (b
i
) and two-factor
interactions (b
ij
) were estimated from the experimental data
obtained by computer simulation programming applying
least squares method using MATLAB 7.1.
From the experimental design in Table 3, experimental
results obtained listed in Table 4 and Eq. (4), the second-
order response functions representing ash content (y
1
)
and combustible recovery (y
2
) of clean coal can be
expressed as a function of four operating parameters of
the MGS, namely drum speed (v), tilt angle (a), wash water
(w) and solid (s). The relationship between responses (ash
content and combustible recovery of clean coal) and oper-
ating parameters were obtained for coded unit as follows:
For ash content of clean coal model equation:
y
1
22:62 1:46x
1
1:04x
2
0:45x
3
0:10x
4
1:08x
2
1
0:25x
2
2
0:23x
2
3
0:12x
2
4
1:52x
1
x
2
0:65x
1
x
3
0:14x
1
x
4
0:51x
2
x
3
0:22x
2
x
4
0:17x
3
x
4
7
Fig. 1. MGS experimental setup.
N. Aslan / Fuel 86 (2007) 769776 771
For combustible recovery of clean coal model equation:
y
2
76:41 3:07x
1
2:48x
2
1:52x
3
0:23x
4
2:49x
2
1
0:09x
2
2
0:40x
2
3
0:83x
2
4
3:57x
1
x
2
1:99x
1
x
3
0:50x
1
x
4
0:55x
2
x
3
0:50x
2
x
4
0:67x
3
x
4
8
The response factors at any regime in the interval of our
experiment design can be calculated from Eqs. (7) and (8).
Experimental results and the predicted values obtained
using model equations (Eqs. (7) and (8)) are given in Table
4 and Figs. 2 and 3. As can be seen, the predicted values
match the experimental values reasonably well, with R
2
of 0.84 for ash content and R
2
of 0.93 for combustible
recovery of clean coal.
5.1. Eect of variables of MGS on ash content
The three-dimensional (3D) response surface plots dem-
onstrate the eect of dierent variables of MGS on ash
content of the clean coal and they are depicted in
Fig. 4(af). The gures show the 3D response surface plots
relationship between two variables of MGS and ash con-
tent of the clean coal at center level of other two variables.
Fig. 4a shows the eect of drum speed and tilt angle on ash
content of clean coal at center level of wash water and
Table 2
Four independent variables of MGS and their levels for CCRD
Variable Symbol Coded variable level
Lowest Low Center High Highest
b 1 0 +1 +b
Drum speed (v), rpm x
1
175 200 225 250 275
Tilt angle (a), x
2
1 3 5 7 9
Wash water (w), lpm x
3
1 3 5 7 9
Solid (s), % x
4
10 20 30 40 50
Table 3
Coded and actual levels of four variables of MGS
Run Coded level of variables Actual level of variables
x
1
x
2
x
3
x
4
v (rpm) a () w (lpm) s (%)
1 1 1 1 1 200 3 3 20
2 1 1 1 +1 200 3 3 40
3 1 1 +1 1 200 3 7 20
4 1 1 +1 +1 200 3 7 40
5 1 +1 1 1 200 7 3 20
6 1 +1 1 +1 200 7 3 40
7 1 +1 +1 1 200 7 7 20
8 1 +1 +1 +1 200 7 7 40
9 +1 1 1 1 250 3 3 20
10 +1 1 1 +1 250 3 3 40
11 +1 1 +1 1 250 3 7 20
12 +1 1 +1 +1 250 3 7 40
13 +1 +1 1 1 250 7 3 20
14 +1 +1 1 +1 250 7 3 40
15 +1 +1 +1 1 250 7 7 20
16 +1 +1 +1 +1 250 7 7 40
17 b 0 0 0 175 5 5 30
18 +b 0 0 0 275 5 5 30
19 0 b 0 0 225 1 5 30
20 0 +b 0 0 225 9 5 30
21 0 0 b 0 225 5 1 30
22 0 0 +b 0 225 5 9 30
23 0 0 0 b 225 5 5 10
24 0 0 0 +b 225 5 5 50
25 0 0 0 0 225 5 5 30
26 0 0 0 0 225 5 5 30
27 0 0 0 0 225 5 5 30
28 0 0 0 0 225 5 5 30
29 0 0 0 0 225 5 5 30
30 0 0 0 0 225 5 5 30
772 N. Aslan / Fuel 86 (2007) 769776
solid. Noting that drum speed and tilt angle has a minor
eect on ash content of clean coal. However, it is worth
noting that a lower ash content of clean coal is obtained
at the center level of drum speed. Fig. 4b shows the eect
of drum speed and wash water on ash content of clean coal
at center level of tilt angle and solid. As can be seen in
Fig. 4b, ash content depends more on the drum speed
rather than on wash water. It is also worth noting that
lower ash content is obtained at the center level of drum
speed. Fig. 4c shows the eect of drum speed and solid
on ash content of clean coal at center level of tilt angle
and wash water. The general form of three-dimensional
relationship is similar to the previous gure. Fig. 4d shows
the eect of tilt angle and wash water on ash content at
center level of drum speed and solid. A minimum ash
Table 4
Observed and predicted values of ash content and combustible recovery
Run Variables Ash content, % Combustible recovery, %
v (rpm) a () w (lpm) s (%) Observed Predicted Observed Predicted
1 200 3 3 20 25.90 25.06 83.05 83.40
2 200 3 3 40 25.30 24.87 85.72 85.20
3 200 3 7 20 27.75 26.02 85.61 84.90
4 200 3 7 40 26.83 25.16 87.30 84.02
5 200 7 3 20 23.89 24.69 79.04 81.32
6 200 7 3 40 23.92 25.36 82.90 85.12
7 200 7 7 20 24.11 23.61 80.10 80.62
8 200 7 7 40 23.54 23.61 80.77 81.74
9 250 3 3 20 18.57 17.53 68.30 67.14
10 250 3 3 40 17.89 17.91 67.35 66.94
11 250 3 7 20 23.07 21.10 78.71 76.60
12 250 3 7 40 22.49 20.81 76.18 73.72
13 250 7 3 20 22.03 23.22 75.93 79.34
14 250 7 3 40 23.70 24.46 80.60 81.14
15 250 7 7 20 25.29 24.75 86.25 86.60
16 250 7 7 40 24.97 25.32 85.97 85.72
17 175 5 5 30 29.48 29.87 93.04 92.51
18 275 5 5 30 23.70 24.05 78.80 80.23
19 225 1 5 30 21.05 19.56 72.63 71.81
20 225 9 5 30 24.61 23.70 81.23 81.73
21 225 5 1 30 19.13 20.78 70.95 74.97
22 225 5 9 30 23.78 22.60 81.79 81.05
23 225 5 5 10 23.80 22.90 79.40 79.27
24 225 5 5 50 22.03 23.28 79.28 80.19
25 225 5 5 30 22.62 22.62 76.41 76.41
26 225 5 5 30 22.62 22.62 76.41 76.41
27 225 5 5 30 22.62 22.62 76.41 76.41
28 225 5 5 30 22.62 22.62 76.41 76.41
29 225 5 5 30 22.62 22.62 76.41 76.41
30 225 5 5 30 22.62 22.62 76.41 76.41
R
2
= 0.84
15
20
25
30
35
15 20 25 30 35
Ash content, (%)
Observed

A
s
h

c
o
n
t
e
n
t
,

(
%
)
P
r
e
d
i
c
t
e
d
Fig. 2. Relation between experimental and predicted ash content of clean
coal using Eq. (7).
R
2
= 0.93
60
65
70
75
80
85
90
95
100
60 65 70 75 80 85 90 95 100
Combustible recovery (%)
Observed
C
o
m
b
u
s
t
i
b
l
e


r
e
c
o
v
e
r
y

(
%
)

P
r
e
d
i
c
t
e
d
Fig. 3. Relation between experimental and predicted combustible recov-
ery using Eq. (8).
N. Aslan / Fuel 86 (2007) 769776 773
content is obtained with minimum level wash water at the
maximum tilt angle level. Fig. 4e shows the eect of tilt
angle and solid on ash content of clean coal at center level
of drum speed and wash water. The general form of three-
dimensional relationship is similar to the previous gure.
Ash content depends more on the tilt angle rather than
on solid. It is worth noting that lower ash content is
obtained at the maximum wash water level. Fig. 4f shows
the eect of wash water and solid on ash content of clean
coal at center level of drum speed and tilt angle. Noting
that wash water has a signicant eect on ash content of
clean coal whilst solid has a trivial eect.
5.2. Eect of variables of MGS on combustible recovery
Fig. 5(af) show the 3D response surface plots relation-
ship between two variables of MGS and combustible
recovery of clean coal at center level of other two variables.
Fig. 5a shows the eect of drum speed and tilt angle on
combustible recovery of clean coal at center level of wash
water and solid. As can be seen, maximum combustible
recovery is obtained with minimum level drum speed but
maximum tilt angle level. It can be also seen that the center
level of drum speed is not a good condition for getting
higher combustible recovery. Fig. 5b shows the eect of
-2
-1
0
1
2
-2
-1
0
1
2
10
15
20
25
30
Drum speed (v), rpm
Tilt angle (a),
A
s
h

c
o
n
t
e
n
t
,

%
-2
-1
0
1
2
-2
-1
0
1
2
22
22
24
26
28
30
32
Drum speed (v), rpm
Wash water (w), lpm
A
s
h

c
o
n
t
e
n
t
,

%

-2
-1
0
1
2
-2
-1
0
1
2
16
18
20
22
24
26
Drumspeed (v), rpm
Solid (s), %
A
s
h

c
o
n
t
e
n
t
,

%
-2
-1
0
1
2
-2
-1
0
1
2
10
15
20
25
30
Tilt angle (a),
Wash water (w), lpm
A
s
h

c
o
n
t
e
n
t
,

%
-2
-1
0
1
2
-2
-1
0
1
2
20
25
30
35
20
25
30
Tilt angle (a),
Solid (s), %
A
s
h

c
o
n
t
e
n
t
,

%

-2
-1
0
1
2
-2
-1
0
1
2
15
16
17
18
19
20
21
Wash water (w), lpm
Solid (s), %
A
s
h

c
o
n
t
e
n
t
,

%
Fig. 4. Response surface plots showing the eect of two variables on ash content of clean coal. Other two variables are held at center level. (a) Drum speed
and tilt angle; (b) drum speed and wash water; (c) drum speed and solid; (d) tilt angle and wash water; (e) tilt angle and solid; (f) wash water and solid.
774 N. Aslan / Fuel 86 (2007) 769776
drum speed and wash water on combustible recovery of
clean coal at center level of tilt angle and solid. The general
form of three-dimensional relationship is similar to the pre-
vious gure, however the eect of drum speed is more pow-
erful than previous. Fig. 5c shows the eect of drum speed
and solid on combustible recovery of clean coal at center
level of tilt angle and wash water. As can be seen from
Fig. 5c, combustible recovery depends more on the drum
speed rather than on solid. Fig. 5d shows the eect of tilt
angle and wash water on combustible recovery of clean
coal at center level of drum speed and solid. Both variables
have same eect on combustible recovery of clean coal.
As the tilt angle is increased, combustible recovery is
increased, just as wash water. Fig. 5e shows the eect of tilt
angle and solid on combustible recovery of clean coal at
center level of drum speed and wash water. Noting that,
as the tilt angle is increased, combustible recovery is
increased steadily, noting also that center level of solid is
not good for getting a higher combustible recovery. Fig. 5f
shows the eect of wash water and solid on combustible
recovery of clean coal at center level of drum speed and tilt
angle. Noting that the general form of three-dimensional
-2
-1
0
1
2
-2
-1
0
1
2
60
70
80
90
100
Drum speed (v), rpm
Tilt angle (a),
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
-2
-1
0
1
2
-2
-1
0
1
2
50
60
70
80
90
100
Drum speed (v), rpm
Wash water (w), lpm
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
-2
-1
0
1
2
-2
-1
0
1
2
65
70
75
80
85
90
Drum speed (v), rpm
Solid (s), %
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
-2
-1
0
1
2
-2
-1
0
1
2
65
70
75
80
85
Tilt angle (a),
Wash water (w), lpm
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
-2
-1
0
1
2
-2
-1
0
1
2
50
60
70
80
90
Tilt angle (a),
Solid (s), %
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
-2
-1
0
1
2
-2
-1
0
1
2
60
62
64
66
68
70
72
Wash water (w), lpm
Solid (s), %
C
o
m
b
u
s
t
i
b
l
e

r
e
c
o
v
e
r
y
,
%
Fig. 5. Response surface plots showing the eect of two variables on combustible recovery of clean coal. Other two variables are held at center level. (a)
Drum speed and tilt angle; (b) drum speed and wash water; (c) drum speed and solid; (d) tilt angle and wash water; (e) tilt angle and solid; (f) wash water
and solid.
N. Aslan / Fuel 86 (2007) 769776 775
relationship is similar to the previous gure. Namely, as the
wash water is increased, combustible recovery is increased
progressively, noting also that center level of solid is not
good condition to be obtained a higher combustible recov-
ery but the extreme levels of solid are good for combustible
recovery.
6. Summary and conclusions
The application of response surface methodology (RSM)
and central composite rotatable design (CCRD) for model-
ing the inuence of some operating variables on the perfor-
mance of the Multi-Gravity Separator (MGS) treating coal
has been discussed.
The central composite rotatable design (CCRD) was
used to design an experimental program to provide data
to model the eects of drum speed, tilt angle, wash water
and feed solids on the performance of Multi-Gravity Sepa-
rator treating coal from Yenicubuk/Turkey lignite coal
containing approximately 36.1% ash. The ranges of values
of variables of MGS used in the design were; drum speed:
175275 rpm, tilt angle: 19, wash water: 19 lpm and feed
solids: 1050%. A total of 30 tests including center points
were conducted. The mathematical model equations were
derived for both ash content and combustible recovery by
using sets of experimental data and a mathematical soft-
ware package (MATLAB 7.1).
The predicted values match the experimental values rea-
sonably well, with R
2
of 0.84 for ash content and R
2
of 0.93
for combustible recovery of clean coal.
In order to gain a better understanding of the eect of
the variables of MGS on ash content and combustible
recovery of clean coal, the predicted models were presented
three-dimensional (3D) response surface graphs.
This study demonstrates that the central composite
rotatable design (CCRD) and response surface methodol-
ogy (RSM) can be successfully used for modeling the some
operating parameters of Multi-Gravity Separator for Yen-
icubuk/Turkey coal and that it is economical way of
obtaining the maximum amount of information in a short
period of time and with the fewest number of experiments.
References
[1] Aslan N, Canbazoglu M. Processing of thickener underow from
celestite concentrator by Multi-Gravity Separator. In: Kemal M,
Arslan V, Akar A, editors. Changing scopes in mineral process-
ing. A.A Balkema; 1996. p. 1036.
[2] O

zbayoglu G, Atalay MU. Beneciation of bastnaesite by a multi-

gravity separator. J Alloy Compd 2000;303304:5203.
[3] Kwak JS. Application of Taguchi and response surface methodolo-
gies for geometric error in surface grinding process. Int J Mach Tool
Manuf 2005;45:32734.
[4] Kincl M, Turk S, Vrecer F. Application of experimental design
methodology in development and optimization of drug release
method. Int J Pharm 2005;291:3949.
[5] Gunaraj V, Murugan N. Application of response surface methodol-
ogies for predicting weld base quality in submerged arc welding of
pipes. J Mater Process Technol 1999;88:26675.
[6] Box GEP, Wilson KB. On the experimental attainment of optimum
conditions. J R Stat Soc, Ser B Stat Meth 1951;13:145.
[7] Box GEP, Hunter WG. The 2
kp
fractional factorial designs part I
and II. J Technometrics 1961;3:311458.
[8] Obeng DP, Morrell S, Napier TJN. Application of central composite
rotatable design to modeling the eect of some operating variables on
the performance of the three-product cyclone. Int J Miner Process
2005;769:18192.
[9] Box GEP, Hunter JS. Multi-factor experimental design for exploring
response surfaces. Ann Math Stat 1957;28:195241.
[10] Cilliers JJ, Austin RC, Tucker JP. An evaluation of formal exper-
imental design procedures for hydrocyclone modelling. In: Svarovsky
L, Thew MT, editors. Proc fourth Int Conf Hydrocyclones. South-
ampton: Kluwer Academic Publishers; 1992. p. 3149.
[11] Crozier RD. Flotation theory, reagents and ore testing. New
York: Pergamon Press; 1992.
[12] Napier-Munn TJ. The central composite rotatable design, JKMRC.
The University of Queensland Brisbane, Australia, 2000. p. 19.
776 N. Aslan / Fuel 86 (2007) 769776