APT 1520 No.

of Pages 11, Model 5G

8 February 2017

Advanced Powder Technology xxx (2017) xxx–xxx

1

Contents lists available at ScienceDirect

Advanced Powder Technology

journal homepage: www.elsevier.com/locate/apt

2 Original Research Paper

6
4 Influence of particle size on dry high-intensity magnetic separation
7
5 of paramagnetic mineral
8 Sunil Kumar Tripathy a,⇑, Nikkam Suresh b
9 a
Research and Development Division, Tata Steel Ltd., Jamshedpur 831001, India
10 b
Department of Fuel and Mineral Engineering, Indian School of Mines, Dhanbad 826004, India

12
11
13
a r t i c l e i n f o a b s t r a c t
1
2 5
8
16 Article history: Dry high-intensity magnetic separators are one of the most common magnetic separation methods 29
17 Received 25 August 2016 extensively used for the concentration of paramagnetic minerals based on their magnetic susceptibility, 30
18 Received in revised form 18 January 2017 particle size, particle density, and shape. In this paper, the influence of particle size on the separation of 31
19 Accepted 26 January 2017
the garnet particle in Induced Roll Magnetic Separator (IRMS) was investigated by varying the magnetic 32
20 Available online xxxx
field intensity, rotor speed, feed rate, and splitter position. The influence of particle size on a magnetic 33
fraction of Induced Roll Magnetic Separator was analysed as the D50 (lm) and D80 (lm) size. The results 34
21 Keywords:
show that segregation of the coarser particle is segregated in non-magnetic fraction while finer sized 35
22 Dry separation
23 High-intensity magnetic separation
particles are at the magnetic fraction. To understand further, particle size distribution is analysed by 36
24 Particle size Rosin-Rammler equation and the variation of the constants are explained by statistics. Also, the influence 37
25 Para-magnetic minerals of individual variables and their interactions on the segregation pattern is described by analysis of 38
26 Induced roll magnetic separation variance method along with response surface plots. 39
27 Ó 2017 Published by Elsevier B.V. on behalf of The Society of Powder Technology Japan. All rights 40
reserved. 41
42

43
44
45 1. Introduction Paramagnetic minerals. Prior to this, an attempt has been made 64
to review on IRMS and its application in mineral industry. 65
46 Dry high-intensity magnetic separators (induced roll magnetic
47 separators and permanent roll magnetic separators) are an out- 1.1. Induced roll magnetic separator 66
48 come of the experience gained by the working engineers and scien-
49 tists over the years in the plant, essentially designed to utilize high A dry induced roll magnetic separator is a simple and efficient 67
50 intensity magnetic field for the separation of Paramagnetic parti- unit for upgrading the magnetic materials from an ore by main- 68
51 cles. In induced roll magnetic separator (IRMS), the separation is taining the field intensity up to 2 T (Tesla). The IRMS consists of 69
52 sought due to the difference in magnetic susceptibility, particle a revolving laminated roll formed of alternate magnetisable and 70
53 size, density and the shape of Para-magnetic particles along with non-magnetic discs. Also, another common type of rotor design 71
54 machine operating and design parameters. Although there have is grooved type in which the grooves are filled with copper. The roll 72
55 been several research papers published on IRMS but the exact is placed between specially shaped poles of an electromagnet. The 73
56 information on the proper understanding of the influence of pro- magnetic circuit is completely designed by iron anticipated to vary 74
57 cess variables with the feed material characteristics is still lacking. the clearance between the rolls and the bridge bar; in which the 75
58 Hence, the main objective of the present research investigation is two rolls are placed, one opposite to each pole. The rolls consist 76
59 to identify the key process variables affecting the performance of the alternate lamina of permeable and impermeable material. 77
60 and analyse their influence on separation while treating garnet The roll onto which the ore fed is in a controlled thin stream by 78
61 fines. Further, for predicting the performance of these units, an a vibratory feeder to the top roll. As the roll revolves, the material 79
62 attempt has been made to develop statistical models by carrying passes through a narrow gap between the pole of the magnet and 80
63 out extensive experimental tests while beneficiating different the roll [1,2]. The schematic diagram of particle separation in IRMS 81
for two stages is depicted in Fig. 1. 82
Some of the variables that influence the separation performance 83
of an IRMS can be grouped as design and operating variables. The 84
⇑ Corresponding author. Fax: +91 (0) 657 227 1510. pole shape and roll diameter are the main design variables whereas 85
E-mail address: [email protected] (S.K. Tripathy). pole gap, roll speed, applied current, feed rate, and splitter position 86

http://dx.doi.org/10.1016/j.apt.2017.01.018
0921-8831/Ó 2017 Published by Elsevier B.V. on behalf of The Society of Powder Technology Japan. All rights reserved.

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

2 S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx

Feed ore of Indian origin [13]. A flow sheet comprising of two stage sep- 134

Bridge Bar arations at 1.1 T and 1.7 T was evaluated by using IRMS to reject 135
Primary Pole siliceous and ferruginous gangue minerals for the beneficiation of 136
manganese ore. It was concluded that a concentrated product of 137
42% Mn with 5.5 of Mn: Fe ratio could be obtained with a recovery 138
of 40% from a feed assaying 32.4% Mn and Mn: Fe ratio of 2.3. Ben- 139
Coil eficiation of a low-grade Sudanese iron ore was carried out by 140

1st Magnetics using IRMS [14,15]. It was found that low-grade iron ore can be 141
upgraded to 64% Fe with an acceptable recovery of up to 72% Fe 142
from a low-grade feed assaying 36% Fe in a two-stage IRMS sepa- 143
Primary Pole
ration. Also, the effects of electric current intensity, rotor rotational 144

2nd Magnetics speed and feed particle size were analysed. It was identified that a 145
Non-magnetics
feed size fraction of
150 + 20 lm is an optimum range with 0.3 146

Fig. 1. Schematic diagram of particle separation in a two stage IRMS

[Tripathy et al., 2014].
Table 1
Particle size analysis of the garnet fines.
87 are the main operating variables. In addition to these feed charac-
Particle Size (lm) Wt (%) Retained
88 teristics viz, particle size and magnetic susceptibility of the miner-
89 als, density, shape, and electrostatic charging of particles also play +500 7.0

500 + 250 65.3
90 a vital role during particle separation in IRMS. The mineral which
250 + 212 11.3
91 responds to a particular set of conditions of a machine depends
212 + 150 12.3
92 on physical properties (magnetic susceptibility, density, and size)
150 + 125 1.2
93 and applied external forces. Further, the application of this separa-
125 + 105 2.4

105 + 75 0.4
94 tor in the mineral industry is discussed briefly in proceeding

75 0.1
95 section.
D50 (lm) 272
D80 (lm) 343
96 1.2. Application of IRMS Skewness value (Sk) 2.56
Kurtosis value (k) 6.84
97 At coarser size, IRMS will be advantageous as compared to other Rosin-Rammler Parameters
98 separators which generate the field strength up to 2 T are achiev- m-value 3.7
c-value
9.5
99 able in the gap between the primary pole and the rotor [3]. Recov-
a-value 308.4
100 ery of fines (<45 lm) is very poor in IRMS, and throughput per
101 width of rotor is also low compared to other dry high intensity
102 magnetic separators. The effectiveness of separation on such fine
103 material is severely reduced by the effects of air current, particle- Table 2
104 particle adhesion, and particle-rotor adhesion [3–6]. There is not Chemical analysis of the garnet fines.
105 enough literature available on the application of IRMS and the only Elements/compounds Analysis value (%)
106 limited information available is in generic form. IRMS has been Fe(T) 21.12
107 extensively used for the beneficiation of beach sands and for the SiO2 35.62
108 removal of weakly magnetic impurities from glass sand, andalu- Al2O3 20.76
109 site, feldspar, wollastonite, magnesite and other industrial miner- TiO2 2.87
CaO 2.38
110 als [7]. The IRM separator is also used for the concentration of
MgO 6.29
111 minerals such as chromite, monazite, and wolframite. This partic-
112 ular magnetic separator was also incorporated in the process flow
113 sheet for upgrading the chromite fines while recovering from
114 Indian chromite beneficiation plant tailings [8]. Also, some studies
115 were carried out at Tata Steel to improve the Cr:Fe ratio of low- 2
116 grade chromite fines by using IRMS [9]. IRMS was used to upgrade
117 the magnetic tailings of the Guleman-Sori/Turkey chromite con- 1.5
118 centration plant [10]. The important process variables such as roll
Magnetisation (emu/g)

1
119 speed and magnetic field intensity were varied and it was found
120 that a product assaying 42.9% Cr2O3 with 85% recovery was
0.5
121 obtained from the feed assaying 32.6% Cr2O3. Magnetic separation
122 of magnesite and dolomite was carried out by using induced roll
0
123 magnetic separator. A magnesite concentrate of 1.12% CaO with -3 -2.5 -2 -1.5 -1 -0.5 0 0.5 1 1.5 2 2.5 3
124 46.02% yield was obtained from a feed of 5.73% CaO. The ore min-
-0.5
125 eralogy reveals that dolomite is associated with goethite, whereas,
126 magnesite is associated with hematite and magnetite. So the mag- -1
127 nesite is collected as a magnetic fraction at the optimum applied
128 current level of 1.0 amp and roll speed of 60 rpm [11]. Effective -1.5
129 removal of iron-bearing minerals (hematite and goethite) from a
130 low grade ferruginous manganese ore was studied using IRMS -2
131 and a model to predict the separation efficiency was developed Applied Magnetic Field (Tesla)
132 based on the artificial neural network [12]. Some more studies
133 were done to beneficiate the low grade ferruginous manganese Fig. 2. VSM result of the garnet fines.

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx 3

3000 928

Garnet
2000
Counts

Garnet

Garnet
Garnet
1000

Garnet

Garnet

Garnet
Garnet

Garnet
Garnet
Garnet
Garnet

Garnet

Garnet
Garnet
0
20 30 40 50 60 70 80
Position [°2Theta] (Copper (Cu))

Fig. 3. XRD pattern of the garnet fines.

147 amp of electric current intensity and at a rotor speed of 100 rpm. rutile, garnet, sillimanite and quartz) of Srikakulam region of 176
148 However, the optimisation study may be misleading as the exper- Andhra Pradesh, India. The as-received samples were subjected 177
149 iments were carried out with one variable at a time. So, a detailed to gravity, magnetic and electrostatic separation for obtaining the 178
150 study for the understanding of interactional effects of the variables respective concentrates. Garnet concentrates were found to pos- 179
151 is required along with statistical analysis. Recently, another study sess more than 99% (purity) by weight in terms of their mineral 180
152 was published on magnetic separation of low-grade iron ore for grains. For feed characterisation, a representative garnet sample 181
153 the concentration of hematite particles [16]. Three variables; mag- was taken and the results obtained are tabulated in subsequent 182
154 netic field intensity, rotor speed and feed rate were varied and tables. Characterisation studies like; particle size analysis, chemi- 183
155 optimum conditions were determined for the effective separation cal analysis, density measurement, X-Ray Diffraction (XRD) stud- 184
156 of hematite particles. Removal of iron bearing impurities from ies, and magnetic property measurement was conducted. 185
157 nepheline syenite ore from Egypt was studied by using IRMS and
158 found that a product of less than 1.5% Fe2O3 can be produced by 2.2. Characterisation studies 186
159 discarding at higher than 80% of the iron impurities [17]. Also,
160 IRMS was used for the pre-concentration of tungsten and tin values Particle size distribution of garnet is carried out in a vibratory 187
161 from a complex ore of wolframite-scheelite-cassiterite ore from sieve shaker and the results are given in Table 1. Further, the par- 188
162 Kyrgyzstan [18]. Two different chromite samples from Sukinda ticle size distributions of the feed samples are expressed in terms 189
163 region have been treated using IRMS and found that Cr: Fe ratio of skewness and kurtosis value and their distributions were 190
164 of these samples was enhanced to 3.98 and 1.9 from 3.09 and simulated by Rosin-Rammler equation. These results are given in 191
165 1.29, respectively [19]. Finer chromite sample of
150 lm particle the same table. Chemical analysis of the samples was carried out 192
166 size was subjected to IRMS and found that a product with Cr: Fe using ICP analyser and the results are given in Table 2. Magnetic 193
167 ratio of 2.2 produced from a feed material of Cr: Fe ratio of 1.6 susceptibility of the feed sample was measured for the sample 194
168 [20]. Low-grade manganese ore with high silica and iron content and it is found to be 11.5  10
7 m3/kg. Further, magnetization 195
169 was investigated using IRMS and concluded that a product with property of ilmenite was measured by using vibratory sample 196
170 Mn: Fe ratio of 3.1 was produced from a feed with Mn: Fe ratio magnetometer (VSM) and the obtained hysteresis curve is given 197
171 of 1.6 [21]. in Fig. 2. From Fig. 2 it is clear that magnetization value for the gar- 198
net is 0.57 and 0.81 emu/g, respectively, at a magnetic field 199
strength of 1 T and 1.5 Tesla. The feed sample was subjected to 200
172 2. Materials and methods
the X-Ray Diffraction for the confirmation of the mineral phases, 201
which is shown in Fig. 3. From Fig. 3, it is revealed that the sample 202
173 2.1. Garnet fines
contains only garnet after beneficiation. Further, the feed sample 203
was analysed for the quantification by Rietveld method and found 204
174 The feed sample for the present investigation was received from
to be 100% of the garnet minerals. 205
175 a placer deposit (consist of different minerals such as ilmenite,

2.3. Induced roll magnetic separator 206

Lepto-kurc Distribuon
In the present study, experiments were carried out to realize 207
the relationship between the separation performance and studied 208
process variables of the IRMS. All the experiments were conducted 209
Frequency

Normal Distribuon Platy-kurc on a bench-scale induced roll magnetic separator (manufactured 210
Distribuon
by Readings of Lismore, Australia) which had a rotor diameter of 211
14 cm and a width of 10 cm, in which the effect of different process 212
variables at different levels was studied. Many preliminary exper- 213
iments were carried out to screen out the levels of the studied vari- 214
ables of IRMS which is very critical in the statistical design of 215
Characteriscs
experimentation. The levels of the variables were had to be fixed 216
Fig. 4. Different types of frequency distribution curve based on kurtosis values. in such a way that the response values of the process should be 217

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

4 S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx

218 varied to extreme boundaries. The considered variables such as 2.4.1. Response surface methodology (RSM) & box-Behnken 238
219 magnetic field intensity (0.8–1.6 T); rotor speed (30–90 rpm); feed experimental design 239
220 rate (0.5–2.0 tph/m width of the rotor) and splitter position (22– RSM was adopted for those tests conducted with IRMS using 240
221 34-degree theta) are varied. The units chosen in defining the vari- garnet fines. Response Surface Methodology (RSM) is a collection 241
222 ables in the present paper are available in the IRMS literature [16]. of statistical and mathematical tools that are useful for modeling 242
223 The levels of the variables used in the present investigation can be and analysing the responses. In this technique, the main objective 243
224 used for the scale up of the process. Magnetic field intensity values is to optimise the response surface that is influenced by various 244
225 are expressed as Tesla (T) which is derived from the standard chart process parameters. It also quantifies the relationship between 245
226 provided by the manufacturer and maintained by adjusting the the controllable input parameters and the response surfaces [22– 246
227 applied current in the separator by keeping 3.2 mm air gap 24]. 247
228 between pole and rotor surface. During the experimentation, the For the three level and four-factorial Box–Behnken experimen- 248
229 humidity of the air was 66% with 35 °C of the environment, and tal designs, a total of 26 experimental runs (with two replicates) 249
230 the moisture content of the feed fines was ensured to below 0.2% are required. The objective is to optimise the response with an 250
231 (by wt.). assumption that the independent variables are continuous and 251
controllable throughout the experimentation. It is also required 252
to find a suitable approximation for the true functional relation- 253
232 2.4. Experimental methodology ship between independent variables and the response surface, typ- 254
ically a second-order model (Eq. (1)) which is utilized in response 255
233 Experiments were carried out in this separator based on statis- surface methodology [22,23]. 256
234 tically designed Box-Behnken experimental method with response 257
235 surface. The levels considered for garnet fines has been mentioned X
k X
k X
k
1 X
k
y ¼ b0 þ bi xi þ bii x2i þ bij xi xj þ e ð1Þ
236 along with the results. There are some other tools used for the 259
i¼1 i¼1 i¼1 j¼2
237 experimental data analysis which is explained further.

Fig. 5. Different types of frequency distribution curve based on skewness values.

Table 3
Statistics considered for the statistical analysis for the experiment carried out in IRMS using garnet fines.

Factor Name Units Low value High value

A Magnetic Field Intensity T 0.8 1.2
B Rotor Speed Rpm 30 90
C Splitter Position Degree 26 34
D Feed Rate Tph 1 2
Response Name Units Minimum Maximum
Y1 Yield to Magnetic % 0.3 99.9
Y2 D50 Size lm 145 270
Y3 D80 Size lm 188 320
Y4 Skewness 0.94 2.59
Y5 Kurtosis
1.13 6.95
Y6 m 2.7 4
Y7 c
10.1 6.6
Y8 a 179.2 309.7

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx 5

Table 4
Experimental results of IRMS along with the operating levels for treating garnet fines.

Test No. A:Magnetic B:Rotor C:Splitter D:Feed Yield (%) D50 Size (lm) D80 Size (lm) Skewness Kurtosis Rosin-Rammler
Field Intensity (T) Speed Position Rate (tph) values
(rpm) (Degree)
m c a
1 1.0 60 30 1.5 36.3 198 285 0.94
1.13 3.7
9.3 242.4
2 0.8 30 30 1.5 87.1 260 315 2.38 5.94 3.8
9.7 289.4
3 1.2 30 30 1.5 98.9 265 319 2.47 6.36 3.7
9.6 295.1
4 0.8 90 30 1.5 0.3 168 198 1.47 1 2.7
6.6 200.4
6 1.2 90 30 1.5 7.3 158 188 2.17 5.34 2.9
6.9 189.2
7 1.0 60 26 1.0 2.9 158 215 1.88 4.03 2.8
6.8 194.9
8 1.0 60 34 1.0 86 261 318 2.41 6.11 3.3
8.7 306.2
9 1.0 60 26 2.0 1.1 155 198 1.59 2.76 2.8
6.6 179.3
10 1.0 60 34 2.0 87 262 319 2.41 6.09 3.8
9.7 287.4
11 0.8 60 30 1.0 20 179 262 1.25 0.51 3.3
8.3 234
12 1.2 60 30 1.0 76.5 261 315 2.4 6.01 4
10.1 290
13 0.8 60 30 2.0 1.4 145 190 1.32 1.56 2.8
6.7 179.2
14 1.2 60 30 2.0 52.8 240 315 1.75 2.71 3.5
8.9 259.7
15 1.0 30 26 1.5 2.3 157 212 1.88 4.03 2.8
6.8 194.9
16 1.0 90 26 1.5 0.9 159 200 1.3 1.82 2.9
7 192.5
17 1.0 30 34 1.5 99.9 270 320 2.59 6.95 3.7
9.6 309.7
18 1.0 90 34 1.5 15.3 202 289 0.97
0.84 3.3
8.1 245.5
19 0.8 60 26 1.5 1.4 159 195 1.38 1.83 2.8
6.7 185.5
20 1.2 60 26 1.5 4.7 158 198 2.07 4.89 2.9
6.9 194.3
21 0.8 60 34 1.5 63.2 258 312 2.25 5.34 3.5
9 285.8
22 1.2 60 34 1.5 97.4 265 320 2.43 6.14 3.6
9.3 302.4
23 1.0 30 30 1.0 97.4 211 296 2.42 6.15 3.5
8.9 299.6
24 1.0 90 30 1.0 4.6 159 208 1.98 4.5 2.9
6.9 196
25 1.0 30 30 2.0 92.7 269 319 2.49 6.47 3.5
9 300.7
26 1.0 90 30 2.0 1.0 154 189 1.32 1.56 2.8
6.7 179.2

Table 5
Analysis of variance for the different responses of IRMS for garnet fines.

Source Yield (%) D50 size (lm) D80 size (lm) Skewness value
Sum of Mean F Sum of Mean F Sum of Mean F Sum of Mean F
squares square Value squares square Value squares square Value squares square Value
A 2246.8 2246.8 7.7 2640.3 2640.3 3.7 2790.8 2790.8 3.1 0.9 0.9 10.3
B 16792.6 16792.6 57.3 15552.0 15552.0 22.0 21590.1 21590.1 24.0 2.1 2.1 24.8
C 15805.0 15805.0 53.9 27265.3 27265.3 38.6 36300.0 36300.0 40.4 0.7 0.7 8.6
D 220.2 220.2 0.8 1.3 1.3 0.0 588.0 588.0 0.7 0.2 0.2 2.1
A2 102.3 102.3 0.3 643.3 643.3 0.9 698.2 698.2 0.8 1.7 1.7 19.7
B2 82.1 82.1 0.3 1.9 1.9 0.0 2498.2 2498.2 2.8 1.7 1.7 20.6
C2 29.8 29.8 0.1 77.6 77.6 0.1 1317.2 1317.2 1.5 1.5 1.5 18.0
D2 197.7 197.7 0.7 31.6 31.6 0.0 441.5 441.5 0.5 1.6 1.6 18.5
AB 5.8 5.8 0.0 56.2 56.2 0.1 49.0 49.0 0.1 0.1 0.1 1.1
AC 238.7 238.7 0.8 16.0 16.0 0.0 6.3 6.3 0.0 0.1 0.1 0.8
AD 6.5 6.5 0.0 42.3 42.3 0.1 1296.0 1296.0 1.4 0.1 0.1 1.5
BC 1730.6 1730.6 5.9 1225.0 1225.0 1.7 90.3 90.3 0.1 0.3 0.3 3.2
BD 0.3 0.3 0.0 992.2 992.2 1.4 441.0 441.0 0.5 0.1 0.1 1.6
CD 2.0 2.0 0.0 4.0 4.0 0.0 81.0 81.0 0.1 0.0 0.0 0.2
Source Kurtosis value m-Value c-Value a-Value
Sum of Mean F Sum of Mean F Sum of Mean F Sum of Mean F
squares square Value squares square Value squares square Value squares square Value
A 19.4 19.4 10.8 0.2 0.2 3.7 1.8 1.8 3.7 2038.4 2038.4 3.8
B 42.3 42.3 23.5 1.0 1.0 15.5 10.8 10.8 21.5 19731.6 19731.6 37.0
C 9.1 9.1 5.1 1.5 1.5 22.4 15.4 15.4 30.6 29561.6 29561.6 55.5
D 3.2 3.2 1.8 0.0 0.0 0.5 0.4 0.4 0.7 1523.3 1523.3 2.9
A2 39.5 39.5 22.0 0.1 0.1 2.1 0.9 0.9 1.8 2.6 2.6 0.0
B2 46.3 46.3 25.8 0.5 0.5 7.2 3.5 3.5 7.0 6.0 6.0 0.0
C2 45.8 45.8 25.5 0.6 0.6 9.4 4.3 4.3 8.5 46.8 46.8 0.1
D2 42.8 42.8 23.9 0.3 0.3 4.3 2.0 2.0 3.9 4.0 4.0 0.0
AB 3.8 3.8 2.1 0.0 0.0 0.3 0.0 0.0 0.1 71.4 71.4 0.1
AC 1.3 1.3 0.7 0.0 0.0 0.0 0.0 0.0 0.0 15.2 15.2 0.0
AD 4.7 4.7 2.6 0.0 0.0 0.0 0.0 0.0 0.1 150.1 150.1 0.3
BC 7.8 7.8 4.3 0.1 0.1 1.0 0.7 0.7 1.4 954.8 954.8 1.8
BD 2.7 2.7 1.5 0.0 0.0 0.0 0.0 0.0 0.0 80.1 80.1 0.2
CD 0.4 0.4 0.2 0.1 0.1 1.0 0.4 0.4 0.7 2.6 2.6 0.0

260 where x1, x2, . . . , xk are input factors which influence the response which should be determined in the second-order model, are 263
261 ‘y’; b0, bii (i = 1, 2, . . . , k), bij (i = 1, 2, . . . , k; j = 1, 2, . . . , k) are obtained by the least square method. The detailed procedure has 264
262 unknown parameters and e is a random error. The b coefficients, been well explained in the published literature [22–24]. 265

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

6 S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx

Table 6
Summary of the estimated effects, coefficients for all the responses of coded units for a confidence level of 95% (alpha = 0.05).

Factor Yield (%) D50 Size (lm) D80 Size(lm) Skewness

Coeff. P Signi. Coeff. P Signi. Coeff. P Signi. Coeff. P Signi.
Intercept 36.3 0.00 Yes 198.0 0.00 Yes 285.0 0.00 Yes 0.9 0.00 Yes
A 13.7 0.02 Yes 14.8 0.07 No 15.3 0.10 No 0.3 0.01 Yes
B
37.4 0.00 Yes
36.0 0.00 Yes
42.4 0.00 Yes
0.4 0.00 Yes
C 36.3 0.00 Yes 47.7 0.00 Yes 55.0 0.00 Yes 0.2 0.01 Yes
D
4.3 0.40 No
0.3 0.97 No
7.0 0.43 No
0.1 0.17 No
A2 4.0 0.56 No 10.0 0.36 No
10.4 0.39 No 0.5 0.00 Yes
B2 3.6 0.60 No
0.5 0.96 No
19.6 0.12 No 0.5 0.00 Yes
C2
2.1 0.75 No 3.5 0.75 No
14.3 0.25 No 0.5 0.00 Yes
D2 5.5 0.43 No 2.2 0.84 No
8.3 0.49 No 0.5 0.00 Yes
AB
1.2 0.89 No
3.7 0.78 No
3.5 0.82 No 0.2 0.31 No
AC 7.7 0.38 No 2.0 0.88 No 1.3 0.93 No
0.1 0.40 No
AD
1.3 0.88 No 3.3 0.81 No 18.0 0.25 No
0.2 0.24 No
BC
20.8 0.03 Yes
17.5 0.21 No
4.8 0.76 No
0.3 0.10 No
BD 0.3 0.97 No
15.8 0.26 No
10.5 0.49 No
0.2 0.23 No
CD 0.7 0.94 No 1.0 0.94 No 4.5 0.77 No 0.1 0.63 No
Std. Deviation 17.1 26.6 30 0.29
Corr. Coeff. 0.9 0.83 0.84 0.88

Factor Kurtosis m-value c-value a-value

Coeff. P Signi. Coeff. P Signi. Coeff. P Signi. Coeff. P Signi.
Intercept
1.1 0.00 Yes 3.7 0.01 Yes
9.3 0.00 Yes 242.4 0.00 Yes
A 1.3 0.01 Yes 0.1 0.08 No
0.4 0.08 No 13.0 0.07 No
B
1.9 0.00 Yes
0.3 0.00 Yes 1.0 0.00 Yes
40.6 0.00 Yes
C 0.9 0.04 Yes 0.4 0.00 Yes
1.1 0.00 Yes 49.6 0.00 Yes
D
0.5 0.21 No
0.1 0.51 No 0.2 0.41 No
11.3 0.11 No
A2 2.5 0.00 Yes
0.1 0.17 No 0.4 0.20 No 0.6 0.94 No
B2 2.7 0.00 Yes
0.3 0.02 Yes 0.7 0.02 Yes
1.0 0.92 No
C2 2.7 0.00 Yes
0.3 0.01 Yes 0.8 0.01 Yes
2.7 0.77 No
D2 2.6 0.00 Yes
0.2 0.06 No 0.6 0.07 No 0.8 0.93 No
AB 1.0 0.17 No 0.1 0.57 No
0.1 0.78 No
4.2 0.72 No
AC
0.6 0.41 No 0.0 1.00 No 0.0 0.94 No 2.0 0.87 No
AD
1.1 0.13 No 0.0 1.00 NO
0.1 0.78 No 6.1 0.60 No
BC
1.4 0.06 No
0.1 0.35 No 0.4 0.25 No
15.5 0.20 No
BD
0.8 0.24 No 0.0 0.85 No 0.1 0.84 No
4.5 0.70 No
CD 0.3 0.65 No 0.1 0.35 No
0.3 0.41 No
0.8 0.95 No
Std. Deviation 1.3 0.26 0.71 23.1
Corr. Coeff. 0.89 0.81 0.84 0.88

266 2.4.2. Rosin-Rammler distribution function the nature of distribution [27]. Kurtosis is a measure of the peaked- 292
267 After each experiment, the feed is separated into two fractions ness of particle size distribution. Higher the value of Kurtosis indi- 293
268 such as higher in the magnetic susceptibility i.e. magnetic product cates that the distribution is peaked relative to the normal (lepto- 294
269 and the other one is a non-magnetic product. For predicting the kurtic). If the Kurtosis is lower in value, the distribution is flatter 295
270 particle size distribution of magnetic products obtained from than the ideal normal curve (platy-kurtic). For better understand- 296
271 experiments carried out with garnet samples, Rosin-Rammler ing on this, different types of distribution curves based on kurtosis 297
272 equation was used. This equation is also called Weibull distribu- values are depicted in Fig. 4. This was estimated by: 298
273 tion function, well-known distribution in mineral processing and 299
274 is widely used to analyse all types of materials whether crushed, nðn þ 1Þ X X 1
X 0 4
K¼ ð3Þ
275 ground or sized [25,26]. It is described in Eq. (2) as: ðn
1Þðn
2Þðn
3Þ s 301
276
x m
278 R ¼ 100  e
ð Þ a ð2Þ where ‘K’ is the kurtosis value of the garnet fines, ‘n’ is the number 302
of data points, ‘x1’ is the 1st data point in the sample, ‘x’ is the arith- 303
279 where, ‘R’ is the mass retained in%, ‘x’ is the size in lm, ‘m’ is the metic mean of the sample, and ‘s’ is the standard deviation of the 304
280 slope of the line in log-log plot, ‘a’ is the size at 36.8% of particles data point. Skewness is a measure of the spread of distribution. If 305
281 retained (36.8 is the ratio 100/e and e is the Neper number the size distribution is ‘‘Normal” than skewness value will be equal 306
282
2.718). The size parameter ‘a’ can be determined by classifying a to zero. A distribution with a significant positive skewness value has 307
283 given material on a mesh size ‘a’ is equal to ‘x’. a long right tail. For better understanding on this, different types of 308
skewed distribution curves are depicted in Fig. 5. In other words, 309
284 2.4.3. Skewness and kurtosis there are one or more extreme large values. It can be determined as: 310
285 After each experiment of pure minerals, the separated products 311
 3
286 (magnetic and non-magnetic fractions) were collected and 1 X X1
X0
SK ¼ ð4Þ
287 weighed, from which a few representative samples were drawn n s 313
288 and subjected to particle size analysis. From the particle size distri-
289 bution data, D50 and D80 (50% and 80% passing particle size, respec- where ‘Sk’ is the skewness value of the studied garnet fines. Further, 314
290 tively) values were determined. The particle size distribution, the effect of different process variables was analysed and their 315
291 skewness, and kurtosis were computed as an index to understand influence on particle size separation in IRMS is discussed. 316

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx 7

100.0
a 100.0 100.0

75.0 75.0 b 75.0 c

Yield to Magnetic

Yield to Magnetic

Yield to Magnetic
50.0 50.0 50.0

25.0 25.0 25.0

0.0 0.0 0.0

90 34.00 2.00
1.2 1.2 1.2
75 32.00 1.75
1.1 1.1 1.1
60 30.00 1.50
1.0 1.0 1.0
B: Rotor Speed 45 0.9 C: Splitter Position 28.00 0.9 D: Feed Rtae 1.25 0.9
A: M ag netic Field Intensity A: M ag netic Field Intensity A: M ag netic Field Intensity
30 0.8 26.00 0.8 1.00 0.8

100.0 87.3 100.0

75.0 65.9 e 75.0

f
d
Yield to Magnetic

Yield to Magnetic

Yield to Magnetic
50.0 44.5 50.0

25.0 23.1 25.0

0.0 1.7 0.0

34.00 2.00 2.00

90 90.00 34.00
32.00 1.75 1.75
75 75.00 32.00
30.00 1.50 1.50
60 60.00 30.00
C: Splitter Position 28.00 45 D: Feed Rtae 1.25 D: Feed Rtae 1.25
45.00 28.00
B: Rotor Speed B: Rotor Speed C: Splitter Position
26.00 30 1.00 30.00 1.00 26.00

Fig. 6. Response surface plots showing the effects on yield (%) of magnetic product of IRMS: (a) between magnetic field intensity (A) and rotor speed (B), (b) between
magnetic field intensity (A) and splitter position (C), (c) between magnetic field intensity (A) and feed rate (D), (d) between rotor speed (B) and splitter position (C), (e)
between rotor speed (B) and feed rate (D), and splitter position (C) and feed rate (D).

317 3. Results and discussion lower levels magnetic field intensity, intermediate levels of rotor 347
speed and splitter position, but at a higher level of feed rate. To 348
318 Magnetic separation studies were carried out by using high- analyse the obtained results further, a predictive second order 349
319 intensity IRMS. The main process parameters of the IRMS viz. mag- quadratic model were constructed and analysed. 350
320 netic field intensity (A), rotor speed (B), splitter position (C) and
321 feed rate (D) were varied, and the obtained results are discussed 3.1. Statistical analysis 351
322 in the subsequent sections. The levels of the variables considered
323 in these experiments are as per the Box-Behnken experimental Statistically designed experiments were considered for the sep- 352
324 design shown in Table 3, along with the statistical values. From aration of garnet fines in IRMS to investigate the effect of process 353
325 Table 3, it can be found that yield (%) to the magnetic product of variables on separation performance. The levels of variables studied 354
326 IRMS was varied from 0.3% to 99.9%. Similarly, D50 particle size and the test results considered for the statistical analysis are given 355
327 (lm) of the magnetic product was varied from 145 to 270 lm. Sim- in Table 4. However, it should be noted here that, the results ana- 356
328 ilarly, all the response values were varied significantly while there lysed were aimed to understand the effect of each variable, their 357
329 is a change in the levels of the variables. interactional effects and their significance through the use of statis- 358
330 A four-factor with three-level Box–Behnken design was used to tics [28,29]. The results of the statistical analysis are presented in 359
331 determine the responses of IRMS viz.: particle size characteristics Tables 5 and 6, and the main findings are described in the following 360
332 of the magnetic products in the IRMS. The independent process sections. From Table 5 it is evident that among the variables, the 361
333 variables and the results obtained from the (Box–Behnken) rotor speed (B) is found to have more influence on the yield (%) of 362
334 designed experiments are given in Table 4. It is observed here that the magnetic product as compared to others. This is due to the 363
335 all the studied variables influence these responses. It is possible to higher F-value, which is an indicator to understand and analyse 364
336 achieve the maximum yield of 99.9% in the magnetic product at the significance of each variable. Also, when there is a change in 365
337 lower level of rotor speed, intermediate levels of magnetic field the rotor speed of the IRMS, there will be dominance of centrifugal 366
338 intensity and feed rate and at a higher level of splitter position force acting on a particle. Similarly, the sequences of the influence 367
339 whereas a possibility of obtaining a minimum of 0.3% yield is of the variables studied on different responses were analysed based 368
340 reported at lower levels of magnetic field intensity, at a higher on the F-value and the order of the significance is: 369
341 level of rotor speed and intermediate levels of both splitter posi-
342 tion and feed rate. Similarly, the D50 (lm) particle size of the mag- Main Effects: 370
343 netic product is reported to be as maximum as of 270 lm at the B > C > A > D (for yield (%) in magnetic product) 371
344 lower level of rotor speed, intermediate levels of both magnetic C > B > A > D (for D50 (lm) size in magnetic product) 372
345 field intensity and feed rate and a higher level of splitter position. C > B > A > D (for D80 (lm) size in magnetic product) 373
346 On the other hand, a minimum D50 (lm) of 145 lm is reported at B > A > C > D (for skewness value) 374

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

8 S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx

262.0 a 262.0 262.0

236.1 236.1 b 236.1 c

210.2 210.2 210.2
D50 Size

D50 Size

D50 Size
184.3 184.3 184.3

158.4 158.4 158.4

90 34.00 2.00
1.2 1.2 1.2
75 32.00 1.75
1.1 1.1 1.1
60 30.00 1.50
1.0 1.0 1.0
B: Rotor Speed 45 0.9 C: Splitter Position 28.00 0.9 D: Feed Rtae 1.25 0.9
A: Magnetic Field Intensity A: Magnetic Field Intensity A: Magnetic Field Intensity
30 0.8 26.00 0.8 1.00 0.8

262.0 262.0
262.0
236.1 d 236.1 e 236.1
210.2 210.2
210.2
D50 Size

D50 Size

D50 Size
184.3 184.3
184.3
158.4 158.4
158.4

34.00 2.00
2.00
90 90
32.00 1.75 34.00
75 75 1.75
32.00
30.00 1.50
60 60 1.50
30.00
C: Splitter Position 28.00 45 D: Feed Rtae 1.25 45
B: Rotor Speed B: Rotor Speed D: Feed Rtae 1.25 28.00
26.00 30 1.00 30 C: Splitter Position
1.00 26.00

Fig. 7. Response surface plots showing the effects on D50 size (lm) of magnetic product of IRMS: (a) between magnetic field intensity (A) and rotor speed (B), (b) between
magnetic field intensity (A) and splitter position (C), (c) between magnetic field intensity (A) and feed rate (D), (d) between rotor speed (B) and splitter position (C), (e)
between rotor speed (B) and feed rate (D), and (f) between splitter position (C) and feed rate (D).

375 B>A>C>D (for kurtosis value) cally guided by centrifugal force while leaving the particle from 406
376 C>B>A>D (for m – value) the rotor surface to form a trajectory. Similarly, all the significance 407
377 C>B>A>D (for c – value) of variables and their interactions estimated are mentioned for all 408
378 C>B>A>D (for a – value) the responses in the same table. To quantify the influence of each 409

379 Interactional Effects variable on the specific response, the co-efficient values can be 410

380 BC > AC > AD > AB > CD > BD (for yield (%) in magnetic product) referred. Among the variables studied, magnetic field intensity 411

381 BC > BD > AB > AD > AC > CD (for D50 (lm) size in magnetic (A) was found to be directly proportional to all the responses that 412

382 product) can be noticed from the positive sign of the coefficient. For 413

383 AD > BD > BC > CD > AB > AC (for D80 (lm) size in magnetic yield (%), magnetic field intensity (A) and rotor speed (B) were 414

384 product) directly proportional due to the positive sign of the coefficient, 415

385 BC > BD > AD > AB > AC > CD (for skewness value) whereas, the other variables (rotor speed (B) and feed rate (D)) 416

386 BC > AD > AB > BD > AC > CD (for kurtosis value) showed an inverse relationship as witnessed from the negative 417

387 BC > CD > AB > BD > AC > AD (for m – value) sign of the coefficient. Similarly, the influence of other responses 418

388 BC > CD > AB > AD > BD > AC (for c – value) has been quantified and is mentioned in Table 6. 419

389 BC > AD > BD > AB > AC > CD (for a – value) Further, it was desirable to investigate the behaviour of the 420

390 IRMS to the simultaneously changing operating parameters, utiliz- 421

391 Further, Table 6 summarizes the estimated effects and coeffi- ing the aforementioned empirical model equations and to discuss 422

392 cients for all the response (yield, D50 (lm) size, D80 (lm) size, it further of the influence of process variables, the predicted mod- 423

393 skewness, kurtosis, m-value, c-value and a-value) of coded units els are described in terms of three-dimensional (3D) response sur- 424

394 for a confidence level of 95% (alpha = 0.05). Based on p-values pre- face plots that show the effect of process variables of IRMS on 425

395 sented in Table 6, rotor speed and splitter position of IRMS was different responses are shown in Figs. 4–7. Each figure shows the 426

396 found to be significant variables for all the responses studied. 3D response surface plots between two variables of IRMS and a 427

397 The yield (%) of the magnetic product stream is influenced by split- particular response at the centre level of the third and fourth 428

398 ter position (C), rotor speed (B) and also by the magnetic field variable. 429

399 intensity (A) whereas, among the interactional effects, the only
400 interaction between rotor speed and splitter position (BC) was 3.2. Effect of variables on yield (%) of magnetic product 430
401 found to be significant. This can be explained as there is a change
402 in the particle trajectory for differently-sized fractions with a Effect of the variables on yield (%) of magnetic product in IRMS 431
403 change in the rotor speed as we well as splitter position. In this is depicted in Fig. 6, in which the effect of magnetic field intensity 432
404 particular case, there are mono-magnetic susceptibility materials (A) and rotor speed (B) on yield (%) of the magnetic product when 433
405 with different particle size fractions in the feed which automati- analysed at the centre level of the splitter position (C) and feed rate 434

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx 9

2.50 a 2.50 2.50 c

2.08 2.08 b 2.08

1.65 1.65 1.65

Skewness
Skewness

Skewness
1.23 1.23 1.23

0.80 0.80 0.80

90.00 34.00 2.00

1.2 1.2 1.2
75.00 32.00 1.75
1.1 1.1 1.1
60.00 30.00 1.50
1.0 1.0 1.0
B: Rotor Speed 45.00 C: Splitter Position 28.00 D: Feed Rtae 1.25 0.9
0.9 0.9
A: Magnetic Field Intensity A: Magnetic Field Intensity A: Magnetic Field Intensity
30.00 0.8 26.00 0.8 1.00 0.8

2.50 2.50 2.50

e f
2.08
d 2.08 2.08

1.65 1.65 1.65

Skewness
Skewness

Skewness

1.23 1.23 1.23

0.80 0.80 0.80

34.00 2.00 2.00

90.00 90.00 34.00
32.00 1.75 1.75
75.00 75.00 32.00
30.00 1.50 1.50
60.00 60.00 30.00
C: Splitter Position 28.00 D: Feed Rtae 1.25 D: Feed Rtae 1.25 28.00
45.00 45.00
B: Rotor Speed B: Rotor Speed C: Splitter Position
26.00 30.00 1.00 30.00 1.00 26.00

Fig. 8. Response surface plots showing the effects on skewness value of magnetic product of IRMS: (a) between magnetic field intensity (A) and rotor speed (B), (b) between
magnetic field intensity (A) and splitter position (C), (c) between magnetic field intensity (A) and feed rate (D), (d) between rotor speed (B) and splitter position (C), (e)
between rotor speed (B) and feed rate (D), and (f) between splitter position (C) and feed rate (D).

435 (D) of solid. It is observed from Fig. 6(a) that, higher yield (%) is of yield (%) is observed at the lower level of splitter position (C) and 467
436 obtained at the lower level of rotor speed (B) and a higher level a higher level of rotor speed (B). Fig. 6(e) shows the effect of rotor 468
437 of magnetic field intensity (A). This is due to the lower centrifugal speed (B) and feed rate (D) on yield (%) of the magnetic product of 469
438 force that is acting on the particles at low rotor speed that directs IRMS at centre level of both the magnetic field intensity (A) and 470
439 the particle flow to the magnetic product. Similarly, lower value of splitter position (C). It is noticed from the figure that yield (%) to 471
440 yield (%) to the magnetic product was reported at the higher level the magnetic product increases with a decrease in the rotor speed 472
441 of rotor speed (B) and lower level of magnetic field intensity (A). whereas there is a trivial effect with a change in the feed rate (D). 473
442 Fig. 6(b) explains the effects of magnetic field intensity (A) and In the same way, Fig. 6(f) shows the effect of splitter position (C) 474
443 splitter position (C) on yield (%) of the magnetic product of IRMS and feed rate (D) on yield (%) of the magnetic product of IRMS at 475
444 at centre levels of both the feed rate and rotor speed. The centre level of both the magnetic field intensity (A) and rotor speed 476
445 yield (%) of the magnetic product increases with increase in the (B). It is observed that as there is a decrease in the yield (%) with a 477
446 splitter position. The increase in the yield is due to the shifting of decrease in the splitter position (C) and there is a trivial effect on 478
447 the material collection away from the centre of the roll, and in turn, yield of the magnetic product with changed solids feed rate (D). 479
448 it accommodates a number of particle flows towards the magnetic It is found that there is a trivial effect on yield of the magnetic pro- 480
449 product bin. From the figure, it is also evident that there is a duct of IRMS with change in the feed rate. This can be explained 481
450 decreasing effect on yield of the magnetic product with a decrease that at a higher feed rate of separation, the magnetic field required 482
451 in magnetic field intensity (A). In IRMS, the magnetic field gradient to magnetize the garnet particle is sufficient at the top layer of the 483
452 is maximum at the rotor surface and decreases gradually towards feed bed. The top layer of the feed bed passes through a low mag- 484
453 the pole of the separator. With an increase in the field gradient, netic field gradient and in this case the garnet is of mono-density 485
454 there will be enhanced magnetization of the coarser particle, and as well as magnetic susceptibility particles. So, the there is an 486
455 the applied magnetic force is higher than the resultant force acting insignificant change in the yield of the magnetic product of IRMS 487
456 on the coarse particle increases. Similarly, Fig. 6(c) shows the while there is a shift in the feed rate. 488
457 effects of magnetic field intensity (A) and feed rate (D) on yield
458 (%) of the magnetic product of IRMS at centre level of both the rotor 3.3. Effect of variables on D50 (lm) size of magnetic product 489
459 speed (B) and splitter position (C). It can be observed that there is
460 an insignificant effect on the yield (%) of the magnetic product as Effect of the variables on D50 size (lm) of magnetic product in 490
461 the feed rate (D) changes. Fig. 6(d) shows the effect of rotor speed IRMS is depicted in Fig. 7(a) which illustrates the effects of mag- 491
462 (B) and splitter position (C) on yield (%) of the magnetic product of netic field intensity (A) and rotor speed (B), on D50 size (lm) of 492
463 IRMS at centre level of both the magnetic field intensity (A) and the magnetic product, at the centre level of the splitter position 493
464 feed rate (D). This figure clearly depicts that maximum yield (%) (C) and feed rate (D) of solids. An important observation that can 494
465 to the magnetic product is observed at the higher level of splitter be here is higher D50 size (lm) is obtained at the lower level of 495
466 position (C) and lower level of rotor speed (B). Likewise, lower level rotor speed (B) and a higher level of magnetic field intensity (A). 496

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

10 S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx

497 Correspondingly lower value of D50 size (lm) to the magnetic (A) and splitter position (C). It is observed that higher D50 (lm) size 528
498 stream to get reported at the higher level of rotor speed (B) and is reported at the higher level of feed rate (D) and lower level of 529
499 lower level of magnetic field intensity (A). Further, it also estab- rotor speed (B). Similarly, Fig. 7(f) shows the effects of splitter posi- 530
500 lishes that all the interactional effects are found to be resembled tion (C) and feed rate (D) on D50 (lm) size of the magnetic product 531
501 with the yield (%) to the magnetic product. Fig. 7(b) explains the of IRMS at centre level of both the magnetic field intensity (A) and 532
502 effects of magnetic field intensity (A) and splitter position (C) on rotor speed (B). It is observed that there is a decrease in the D50 533
503 D50 (lm) size of the magnetic product of IRMS at centre levels of (lm) size with decrease in the splitter position (C), which gets 534
504 both feed rate (D) and rotor speed (B). The D50 (lm) size of the influenced marginally with increased feed rate (D) of solids to 535
505 magnetic product increases with increase in the splitter position the magnetic separator. 536
506 (C). It is also observed that there is a decreasing effect on D50
507 (lm) size of the magnetic product with a decrease in the magnetic
508 field intensity (A). It is evident from the results that there is a clear 3.4. Effect of variables on skewness and kurtosis value of magnetic 537

509 gradation of particle size and their trajectory from finer to coarser product 538

510 particle size while shifting the splitter position from 22 degrees to
511 34 degrees in IRMS. So, in other words, particle trajectory is firmly Effect of the variables on skewness value of the magnetic pro- 539

512 governed by the particle magnetization value of a particle in a duct of IRMS is depicted in Fig. 8. Fig. 8(a) shows the effects of 540

513 given magnetic field intensity. Similarly Fig. 7(c) shows the effects magnetic field intensity (A) and rotor speed (B) on skewness value 541

514 of magnetic field intensity (A) and feed rate (D) on D50 (lm) size of of the magnetic product at the centre levels of both splitter posi- 542

515 the magnetic product of IRMS at centre level of both the rotor tion (C) and feed rate (D). It is observed that higher skewness value 543

516 speed (B) and splitter position (C). From the nature of the plots, is obtained at the lower level of rotor speed (B) and a higher level 544

517 that a minimal effect on the D50 (lm) size of the magnetic product of magnetic field intensity (A). A similarly lower value of skewness 545

518 as the feed rate (D) changes. Fig. 7(d) shows the effects of rotor for the magnetic product is reported at the higher level of rotor 546

519 speed (B) and splitter position (C) on D50 (lm) size of the magnetic speed (B) and lower magnetic field intensity (A). Fig. 8(b) explains 547

520 product of IRMS at centre levels of the magnetic field intensity (A) the effects of magnetic field intensity (A) and splitter position (C) 548

521 and feed rate (D). From the plot, maximum D50 (lm) size to the on the skewness of magnetic product of IRMS at centre levels of 549

522 magnetic product is observed at the higher level of splitter position feed rate (D) and rotor speed (B). The skewness of the magnetic 550

523 (C) and lower level of rotor speed (D). Similarly, lower level of D50 product increases as there is an increase in both splitter position 551

524 (lm) size is observed at the lower level of splitter position (C) and (C) and magnetic field intensity (A). It is also observed that lower 552

525 a higher level of rotor speed (B). Fig. 7(e) shows the effect of rotor value of skewness for the magnetic product is reported at a lower 553

526 speed (B) and feed rate (D) on D50 (lm) size of the magnetic pro- value of both magnetic field intensity (A) and splitter position (C). 554

527 duct of IRMS at centre level of both the magnetic field intensity Similarly Fig. 8(c) shows the effects of magnetic field intensity (A) 555
and feed rate (D) on skewness value of the magnetic product of 556

6.18
a 6.18
b 6.18 c
4.20 4.20 4.20
Kurtosis

2.22 2.22 2.22

Kurtosis

Kurtosis

0.25 0.25 0.25

-1.73 -1.73 -1.73

90 34.00 2.00
1.2 1.2 1.2
75 32.00 1.75
1.1 1.1 1.1
60 30.00 1.50
1.0 1.0 1.0
B: Rotor Speed 45 0.9 C: Splitter Position 28.00 0.9 D: Feed Rtae 1.25 0.9
A: Magnetic Field Intensity A: Magnetic Field Intensity A: Magnetic Field Intensity
30 0.8 26.00 0.8 1.00 0.8

6.18 6.18 6.18

4.20 d e f
4.20 4.20
Kurtosis

2.22 2.22 2.22

Kurtosis

Kurtosis

0.25 0.25 0.25

-1.73 -1.73 -1.73

34.00 2.00 2.00

90 90 34.00
32.00 1.75 1.75
75 75 32.00
30.00 1.50 1.50
60 60 30.00
C: Splitter Position 28.00 45 D: Feed Rtae 1.25 D: Feed Rtae 1.25
45 28.00
B: Rotor Speed B: Rotor Speed C: Splitter Position
26.00 30 1.00 30 1.00 26.00

Fig. 9. Response surface plots showing the effects on kurtosis value of magnetic product: (a) between magnetic field intensity (A) and rotor speed (B), (b) between magnetic
field intensity (A) and splitter position (C), (c) between magnetic field intensity (A) and feed rate (D), (d) between rotor speed (B) and splitter position (C), (e) between rotor
speed (B) and feed rate (D), and (f) between splitter position (C) and feed rate (D).

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018
 APT 1520 No. of Pages 11, Model 5G
8 February 2017

S.K. Tripathy, N. Suresh / Advanced Powder Technology xxx (2017) xxx–xxx 11

557 IRMS at centre levels of both the rotor speed (B) and splitter posi- Rammler distribution equation and wherein the equation con- 619
558 tion (C). It can further be observed that there is a minimal effect on stants can be computed and estimated by different regression 620
559 the skewness value of the magnetic product as the feed rate (D) equations. 621
560 changes. The maximum value of skewness reported at the
561 higher level of magnetic field intensity (A) and lower level of feed References 622
562 rate (D). Fig. 8(d) shows the effect of rotor speed (B) and splitter
[1] J. Svoboda, T. Fujita, Recent developments in magnetic methods of material 623
563 position (C) on skewness for the magnetic product of IRMS at cen-
separation, Miner. Eng. 16 (2003) 785–792. 624
564 tre levels of both the magnetic field intensity (A) and feed rate (D). [2] W.A.P.J. Premaratne, N.A. Rowson, Development of a magnetic hydro cyclone 625
565 It is observed that maximum skewness value to the magnetic pro- separation for the recovery of titanium from beach sands, Phys. Separat. Sci. 626
Eng. 12 (2003) 215–222. 627
566 duct is observed at the higher level of splitter position (C) and
[3] J. Svoboda, Development in Mineral Processing, vol. 8, Elsevier Science 628
567 lower level of rotor speed (B). Similarly, lower level of skewness Publishers, Amsterdam, BV, 1987, p. 692. 629
568 is observed at the lower level of splitter position (C) and a higher [4] D.G. Morgan, W.J. Bronkala, The selection and application of magnetic 630
569 level of rotor speed (B). Fig. 8(e) shows the effect of rotor speed separation equipment. Part I, Magn. Electr. Separat. 3 (1991) 5. 631
[5] V.V. Karmazin, Theoretical assessment of technological potential of magnetic 632
570 (B) and feed rate (D) on the skewness of magnetic product of IRMS and electrical separation, Magn. Electr. Separat. 8 (1997) 139–159. 633
571 at centre levels of both the magnetic field intensity (A) and splitter [6] M. Dobbins, P. Dunn, I. Sherrell, Recent advances in magnetic separator designs 634
572 position (D). It is observed that higher skewness value is reported and applications, in: The 7th International Heavy Minerals Conference ‘What 635
next’, The Southern African Institute of Mining and Metallurgy, 2009, pp. 63– 636
573 at the higher level of feed rate (D) and lower level of rotor speed 637
70.
574 (B). Similarly, Fig. 8(f) shows the effect of splitter position (C) and [7] P.K. Naik, Quantification of induced roll magnetic separation of mineral sand, 638
575 feed rate (D) on skewness value of the magnetic product of IRMS Scand. J. Metall. 31 (2002) 367–373. 639
[8] S.K. Tripathy, Y.R. Murthy, V. Singh, Characterisation and separation studies of 640
576 at centre level of both the magnetic field intensity (A) and rotor 641
Indian chromite beneficiation plant tailing, Int. J. Miner. Process. 122 (2013)
577 speed (B). It is observed that there is a decrease in the skewness 47–53. 642
578 value with a decrease in the splitter position(C). It is also observed [9] S.K. Tripathy, V. Singh, Y.R. Murthy, Improvement in Cr: Fe ratio of indian 643
chromite ore for ferro chrome production, Int. J. Mining Eng. Miner. Process. 1– 644
579 that there is a trivial change in skewness value of the magnetic
3 (2013) 101–106. 645
580 product with a change in the feed rate (D). [10] N. Aslan, H. Kaya, Beneficiation of chromite concentration waste by multi- 646
581 Effect of the variables on kurtosis value of the magnetic product gravity separator and high intensity Induced-roll magnetic separator, Arab. J. 647
Sci. Eng. 34-2B (2009) 285–297. 648
582 in IRMS is depicted in Fig. 9. It is found that all the interactional
[11] A. Yehia, M. Al-Wakeel, Role of ore mineralogy in selecting beneficiation route 649
583 effects are found to be identical with the skewness value of the for magnesite-dolomite separation, Physicochem. Probl. Miner. Process. 49 (2) 650
584 magnetic product. (2013) 525–534. 651
[12] V. Singh, Connectionist approach for modeling the dry roll magnetic separator, 652
Miner. Metall. Process. 26 (3) (2009) 127–132. 653
585 4. Conclusion [13] V. Singh, T.K. Ghosh, Y. Ramamurthy, V. Tathavadkar, Beneficiation and 654
agglomeration process to utilize low-grade ferruginous manganese ore fines, 655
Int. J. Miner. Process. 99 (2011) 84–86. 656
586 The performance of Induced Roll Magnetic Separator (IRMS) has 657
[14] A.A.S. Seifelnassr, E.M. Moslim, A.Z.M. Abouzeid, Effective processing of low-
587 investigated by conducting experiments using uniform magnetic grade iron ore through gravity and magnetic separation techniques, 658
588 susceptibility and density material (garnet fines) by varying the Physicochem. Probl. Miner. Process. 48 (2) (2012) 567–578. 659
[15] A.A.S. Seifelnassr, E.M. Moslim, A.Z.M. Abouzeid, Concentration of a Sudanese 660
589 operating variables; viz.: magnetic field intensity (A), rotor speed
low-grade iron ore, Int. J. Miner. Process. 122 (2013) 59–62. 661
590 (B), splitter position(C), feed rate (D). The role of particle size in [16] S.K. Tripathy, P.K. Banerjee, N. Suresh, Separation analysis of dry high intensity 662
591 magnetic separation with garnet fines is investigated by analysing induced roll magnetic separator for concentration of hematite fines, Powder 663
592 the particle size distribution of the magnetic fraction. To under- Technol. 264 (2014) 527–535. 664
[17] S.S. Ibrahim, H.A. Mohamed, T.R. Boulos, Dry magnetic separation of nepheline 665
593 stand the variation, particle size distribution indices (D50 and D80 syenite ores, Physicochem. Probl. Miner. Process. 36 (2002) 173–183. 666
594 sizes) and the estimated constants of Rosin-Rammler equation [18] T. Sreenivas, K. Srinivas, R. Natarajan, N.P.H. Padmanabhan, An integrated 667
595 were analysed statistically. It is concluded that role of particle size process for the recovery of tungsten and tin from a combined wolframite- 668
scheelite-cassiterite concentrate, Miner. Process. Extractive Metall. Rev.: Int. J. 669
596 is a significant factor which needs to be considered in this type of 25 (2004) 193–203. 670
597 separator as visualized from the test results of garnet fines. It is [19] S.K. Tripathy, Y.R. Murthy, V. Singh, N. Suresh, Processing of ferruginous 671
598 also found that segregation of finer particles to take place in the chromite ore by dry high intensity magnetic separation, Miner. Process. Extr. 672
Metall. Rev. (2014), http://dx.doi.org/10.1080/08827508.2016.1168418. 673
599 magnetic products whereas; the non-magnetic stream product is 674
[20] S.K. Tripathy, P.K. Banerjee, N. Suresh, Magnetic separation studies on
600 affected by coarser particles. ferruginous chromite fine to enhance Cr: Fe ratio, Int. J. Miner., Metall. 675
601 It is concluded that by the increase in the rotor speed of IRMS, Mater. 22 (3) (2015) 217. 676
[21] S.K. Tripathy, P.K. Banerjee, N. Suresh, Effect of desliming on the magnetic 677
602 would result in the throwing of some magnetic particles to the 678
separation of low-grade ferruginous manganese ore, International Journal of
603 non-magnetic stream resulting in poor yield (%) of the magnetic Minerals, Metall. Mater. 22 (7) (2015) 1. 679
604 product. Amongst, all the variables studied for the case of IRMS, [22] N. Aslan, Application of response surface methodology and central composite 680
rotatable design for modeling and optimization of a multi-gravity separator 681
605 the rotor speed (B) is found to be the most influencing parameter
for chromite concentration, Powder Technol. 185 (2008) 80–86. 682
606 as for as yield (%) of the magnetic product is concerned. Similarly, [23] N. Aslan, Y. Cebeci, Application of Box-Behnken design and response surface 683
607 interactional effects between rotor speed (B) and splitter position methodology for modeling of some Turkish coals, Fuel 86 (2007) 90–97. 684
608 (C) are found to be most significant on yield (%) of the magnetic [24] S. Ozgen, A. Yıldız, A. Çalısßkan, E. Sabah, Modeling and optimizatıon of hydro 685
cyclone processıng of low grade bentonites, Appl. Clay Sci. 46 (3) (2009) 305– 686
609 product compared to other interactions of IRMS. For segregating 313. 687
610 of garnet fines (lower magnetic susceptibility) in IRMS, the order [25] P.A. Vesilind, The Rosin-Rammler particle size distribution, Resource Recov. 688
611 of significance among the variables is as given below: Conserv. 5 (1980) 275–277. 689
[26] A. Tasdemir, Fractal evaluation of particle size distributions of chromites in 690
different comminution environments, Miner. Eng. 22 (2009) 156–167. 691
612 B>C>A>D (for yield (%) in magnetic product) [27] P. Pourghahramani, Effects of ore characteristics on product shape properties 692
613 C>B>A>D (for D50 (lm) size in magnetic product) and breakage mechanisms in industrial SAG mills, Miner. Eng. 32 (2012) 30– 693
37. 694
614 B>A>C>D (for skewness value) 695
[28] E.W. Ronald, H.M. Raymond, Probability and Statistics for Engineers and
615 B>A>C>D (for kurtosis value) Scientists, fourth ed., Macmillan Publishers, London, 1989, p. 578. 696
616 [29] D.P. Patil, K.U. Bhaskar, M.R. Jakhu, T.C. Rao, Removal of graphite from lead 697
rougher concentrate using water-only cyclones, Int. J. Miner. Process. 49 698
617 Based on the studied variables, the particle size distribution of 699
(1997) 87–96.
618 the magnetic product of IRMS can be predicted by Rosin-
700

Please cite this article in press as: S.K. Tripathy, N. Suresh, Influence of particle size on dry high-intensity magnetic separation of paramagnetic mineral,
Advanced Powder Technology (2017), http://dx.doi.org/10.1016/j.apt.2017.01.018