This article was downloaded by: [NUS National University of Singapore]

On: 16 June 2013, At: 07:03

Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954
Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH,
UK

International Journal of Coal

Preparation and Utilization
Publication details, including instructions for
authors and subscription information:
http://www.tandfonline.com/loi/gcop20

Optimization of A Jigging
Process Using Statistical
Technique
a
Mahmoud M. Ahmed
a
Mining and Metallurgical Engineering Department,
Assiut University, Egypt
Published online: 11 Mar 2011.

To cite this article: Mahmoud M. Ahmed (2011): Optimization of A Jigging Process

Using Statistical Technique, International Journal of Coal Preparation and Utilization,
31:2, 112-123

To link to this article: http://dx.doi.org/10.1080/19392699.2010.549383

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-

and-conditions

This article may be used for research, teaching, and private study purposes.
Any substantial or systematic reproduction, redistribution, reselling, loan,
sub-licensing, systematic supply, or distribution in any form to anyone is
expressly forbidden.
The publisher does not give any warranty express or implied or make any
representation that the contents will be complete or accurate or up to
date. The accuracy of any instructions, formulae, and drug doses should be
independently verified with primary sources. The publisher shall not be liable
for any loss, actions, claims, proceedings, demand, or costs or damages
 whatsoever or howsoever caused arising directly or indirectly in connection
with or arising out of the use of this material.
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013
 International Journal of Coal Preparation and Utilization, 31: 112–123, 2011
Copyright # Taylor & Francis Group, LLC
ISSN: 1939-2699 print=1939-2702 online
DOI: 10.1080/19392699.2010.549383

OPTIMIZATION OF A JIGGING PROCESS USING

STATISTICAL TECHNIQUE
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

MAHMOUD M. AHMED
Mining and Metallurgical Engineering Department,
Assiut University, Egypt

The jigging studies of a synthetic binary mixture of quartz and coal

were carried out. The variables studied were the particle size, bed
thickness, number of strokes per minute, and water level. Experi-
ments were carried out using 24 full factorial design. The main and
interaction effects on quartz recovery were evaluated using Yates’
analysis. The optimum jigging condition was found by the method
of steepest ascent. A product with 81.81% quartz and 97.74% recov-
ery in the lower layer was obtained at the following optimum con-
ditions: particle size of 3.907 mm, bed thickness of 1.87 cm, water
level of 4.41 cm, and 234.3 strokes per minute.

Keywords: Factorial analysis; Harz jig; Interaction effect; t-test;

Yates’ analysis

INTRODUCTION
Jigging is a process of ore concentration carried out in any fluid whose
effectiveness depends on differences in specific gravity of granular min-
eral particles [1]. The jig bed is divided into two zones. The concentrate
zone consists of the bottom layers, where the content of heavy mineral
should be greater than 95%.
Many of the jigging factors are inherently controllable (manipulated
variables), but some uncontrollable factors (disturbance variables)

Address correspondence to Mahmoud M. Ahmed, Mining and Metallurgical Engin-

eering Department, Faculty of Engineering, Assiut University, Assiut 71516, Egypt. E-mail:
mahmoudahmed6666@yahoo.com; mamoah@aun.edu.eg
 OPTIMIZATION OF A JIGGING PROCESS 113

associated with the ore to be treated also play an important role in the
separation process. The basic factors that affect jig performance were
reviewed by some authors [2, 3].
One of the most effective techniques to study process behavior is the
factorial designed tests with analysis of variance [4–6]. There are several
advantages to statistical design of experiments over the classical one
variable at a time method, where one variable is varied at a time. In stat-
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

istical design, experiments can be conducted in an organized manner and

can be analyzed systematically to obtain much needed information. This
information can be utilized for optimization purposes.
A review of jigging separation literature indicates that many authors
studied the analysis of jigging for improved performance [7, 8]. Other
researchers carried out the mathematical modeling and simulation of
particle motion into the jig [9, 10]. There is a lack of statistically based
studies on the effects and=or interactions of different variables on jigging
process.
The main objective of the current work is to determine the main and
interaction effects of jigging variables, using statistical techniques, on
complete stratification process of mineral particles in jig and to find
out optimum conditions. This is done through a 24 factorial design with
midpoint replicates.
The different aims of the optimization strategy used in this study are
to design experimental tests (using factorial design) of separation in a
Harz jig, to perform an analysis of the experimental results by ANOVA
to determine the significant factors influencing the jigging process, and
to find out the optimum conditions of the jigging process.

EXPERIMENTAL
Materials
The tests were run on a batch basis using a synthetic binary mixture of
heavy and light minerals. The mixture consists of quartz (specific gravity
[SG] ¼ 2.65 g=cm3) and coal (SG ¼ 1.30 g=cm3) with a percent of 1:1 by
weight. The concentration criterion of the mixture is about 5.5, which
means that high promise separation efficiency will be expected for the
current process. Vijayendra [1] stated that if the concentration criterion
is a negative or positive number greater than 2.5, separation in water is
easy at all sizes down to the finest sands. Two size fractions of (
8 þ 6.3)
 114 M. M. AHMED

mm and (
1.6 þ 1.25) mm of two minerals were prepared to be used in

the experiments. Two different weights of the mixture were used, i.e.,
200 gm with the high level and 100 gm with the low level.

Methods
A laboratory fixed-sieve jig was used. It is a single-hutch Harz-type
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

machine. It consists of one rectangular-shaped hopper compartment

called an external cell with a size of 10  8  12 cm. The internal cell con-
sists of a rectangular box with a size of 6  6  8 cm, ending with a wire
screen of 1 mm size.
In each experiment, the required weights of two minerals are well
mixed and put into the internal jig cell. The cell is then connected with
the drive unit and put into the external cell, which is filled with water.
The bed thickness, number of strokes per minute, and water level are
set according to the required values for each particular experiment. The
stroke length is fixed to about 12 mm. The jigging time is set at 45 seconds.
These two variables are maintained constant during all experiments.
In each experiment and after the separation is attained, the machine
is stopped. The internal cell is separated from the drive unit and taken out
of the water. The level of cutting (or removal) of the jigging products is
taken as the height of heavy mineral when complete separation is attained.
Both the upper layer and the lower layer are collected separately,
dried and analyzed for quartz percent. The content of the heavy mineral
(quartz) in the jigging products is determined by the heavy-liquid separ-
ation process. The raw data of experimental results are tabulated in
Table 1.

Variables
The variables considered in this study are: particle size (X1), bed thick-
ness (X2), water level (X3), and number of strokes per minute (X4). The
levels of variables are given in Table 2.

Coding and General Form of Response with Main and

Interaction Effects
The statistical design of experiments is useful in that the simultaneous
assessment of several factors can be made by determining the main
 OPTIMIZATION OF A JIGGING PROCESS 115

Table 1. Raw data of experimental results

Light layer Heavy layer

Wt. (g) Coal % Quartz % Wt. (g) Coal % Quartz %

44.00 91.09 8.91 51.50 11.31 88.69

35.50 97.18 2.82 64.50 23.26 76.74
121.00 61.88 38.12 71.00 28.41 71.59
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

120.00 53.33 46.67 73.50 40.82 59.18

41.50 89.24 10.76 53.00 15.21 84.79
35.50 88.73 11.27 61.50 25.20 74.80
91.50 87.49 12.51 100.50 12.88 87.12
111.50 61.88 38.12 82.50 31.52 68.48
47.00 87.48 12.52 48.00 9.40 90.60
37.50 97.33 2.67 61.50 17.89 82.11
87.00 97.37 2.63 104.50 6.71 93.29
93.00 90.86 9.14 106.50 12.68 87.32
40.50 90.82 9.18 54.50 15.92 84.08
40.50 95.06 4.94 58.00 18.10 81.90
92.50 93.88 6.12 100.50 6.56 93.44
93.50 93.05 6.95 105.00 10.95 89.05

and interaction effects. Each value of jigging variables was converted

into the following three coding levels:
1, þ1, and 0 for low, high, and
center point, respectively. The center point is the arithmetic mean of
the high and low levels. Quartz recovery in the lower layer has been
treated as ‘‘response.’’
The matrix for four variables varied at two levels (þ,
) and the cor-
responding recovery and grade of quartz in the lower layer are shown in
Table 3. According to the basic principle of the design of experiments,
four experiments were carried out at the base level to estimate error
and standard deviation.

Table 2. The variables and levels of 24 factorial design for jigging process

Variables Code Low level (
) High level (þ)

Particle size (mm) X1 1.425 7.150

Bed thickness (cm) X2 1.570 03.26
Water level (cm) X3 03.10 05.25
Number of strokes (rpm) X4 151.1 249.9
 116 M. M. AHMED

Table 3. 24 Full factorial design matrix for jigging process

Coded factors Response

Observation X1 X2 X3 X4 Recovery % Grade %

1



92.09 88.69
2 þ


98.02 76.74
3
þ

52.43 71.59
4 þ þ

43.72 59.18
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

5

þ
90.96 84.79
6 þ
þ
92.00 74.80
7
þ þ
88.44 87.12
8 þ þ þ
57.07 68.48
9


þ 88.08 90.60
10 þ

þ 98.06 82.11
11
þ
þ 97.70 93.29
12 þ þ
þ 91.63 87.32
13

þ þ 92.49 84.08
14 þ
þ þ 95.96 81.90
15
þ þ þ 94.31 93.44
16 þ þ þ þ 93.50 89.05

RESULTS AND DISCUSSION

The experimental data were analyzed statistically. The effect of the vari-
ables were quantified and interpreted.
Minitab statistical software was used for the analysis of experimental
data from the randomized tests with designed conditions, which yielded
the main and interaction effects that are specific to the jigging system
under investigation. The main effect of a factor is given as the change
in a response produced by the change between the upper and lower level
of that factor [11].

Statistical Analysis
In the present work, four variables were taken into consideration to
evaluate their main and interaction effects on the recovery of quartz in
the lower layer to study the separation of quartz from coal. In other
words, the main goal has been to establish the best set of variables that
could be used in a jig to obtain maximum recovery in the lower layer with
an acceptable grade.
 OPTIMIZATION OF A JIGGING PROCESS 117

A 24 full factorial unreplicated jigging experiment was carried out in

order to evaluate the main and interaction effects of variables on jigging
process. Yates’ notation has been used in this work to name each treat-
ment [12]. For example, treatment ‘‘ab’’ is the experimental run in which
the variables X1 and X2 are set at their highest levels whereas the vari-
ables X3 and X4 are at their lowest levels. Treatment ‘‘acd’’ is the experi-
mental run in which the variables X1, X3, and X4 are set at their highest
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

levels whereas the variable X2 is at its lowest level, and so on.

To study the main and interaction effects of the variables on the recov-
ery of quartz, Yates’ analysis and an analysis of variance have been carried
out [12]. The total variance (total mean square) of a factorial experiment
can be divided into several sources using Yates’ analysis. In case of unre-
plicated experiments, all the variance is subdivided between the effects.
A 24 experiment has (24–1) degrees of freedom, and Yates’ analysis
divides the total variation in the results into the 15 effects. It follows that
each effect has one degree of freedom; hence, for any effect, the mean
square equals the sum of squares. In Yates’ analysis, the standard
addition and subtraction in pairs is carried out by n times for n factors.
The test of statistical significance of each effect necessitates estimation
of experimental error.
A suitable confidence interval of 99% is chosen for determination of
significance of main and interaction effects. In the current analysis with
k ¼ 4 factors, 4 center points have been used to estimate the experi-
mental error and the variance, r2. The results of replicated experiments
from the central points are tabulated in Table 4. The calculated Student’s
t test is given by [13, 14]:

tcal: ¼ ½Calculated main or interaction effect=

pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð1Þ
ðVarianceðEffectsÞÞ  t0:01;3:

Table 4. The results of replicated experiments

Light layer Heavy layer

Wt. (g) R(Coal) % R(Quartz) % Wt. (g) R(Coal) % R(Quartz) %

53.14 69.39 3.43 89.47 30.61 95.57

60.17 74.83 6.85 84.80 25.17 93.15
61.15 78.20 4.32 84.85 21.80 95.68
60.57 75.66 7.05 82.74 24.34 92.95
 118 M. M. AHMED

The value of t0.01,3 is 4.54, which can be obtained from the Student’s
t distribution table and if the estimated main, and interaction effects are
significant at 99% confidence level, then they will satisfy the above cri-
teria [14]. In other words, an effect is considered to be significant if its
significance level is greater than 99%. The Yates’ analysis and analysis
of variance for quartz recovery are given in Table 5.
On eliminating the coefficients that are not significant, the statistical
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

model can be built up for prediction of quartz recovery using Yates’

analysis data (Table 5). This model can be used to perform analysis of
the residues to check the assumption on the experimental error distri-
bution of the factorial design [15]. The model formed for quartz recov-
ery, using the effects of variables significant at the 99% confidence
level or more, is given below [6, 14]:

Rprd ¼ 85:404
8:054 X2
4:211 X1  X2 þ 2:688 X3 þ 3:293 X2  X3

þ 8:563 X4 þ 2:479 X1  X4 þ 8:373 X2  X4
2:588 X3  X4

3:772 X2  X3  X4 ð2Þ

Table 5. Results of statistical analysis and test of significance of main

and interaction coefficients

Code Rexp, % Effects tcal. Sign.

Average 92.09 85.404 – –

X1 98.02
1.659
3.66 NS
X2 52.43
8.054
17.75 99.9
X1X2 43.72
4.211
9.28 99.5
X3 90.96 2.688 5.92 99
X1X3 92.00
1.799
3.97 NS
X2X3 88.44 3.293 7.26 99.5
X1X2X3 57.07
0.374
0.83 NS
X4 88.08 8.563 18.87 99.9
X1X4 98.06 2.479 5.46 99
X2X4 97.70 8.373 18.45 99.9
X1X2X4 91.63 1.669 3.68 NS
X3X4 92.49
2.588
5.70 99
X1X3X4 95.96 1.644 3.62 NS
X2X3X4 94.31
3.772
8.31 99.5
X1X2X3X4 93.50 1.847 4.07 NS

Note. NS ¼ nonsignificant.
 OPTIMIZATION OF A JIGGING PROCESS 119

R2 ¼ 0:94748

X1, X2, X3, X4 are expressed in coded form of
1 or þ1.

The coefficient of determination, R2, is used to check the model’s
ability to predict the response (recovery) accurately. The empirical
model was found to accurately estimate the response variable as indi-
cated by the R2 value (0.95). The residual analysis for recovery is given
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

in Figure 1.
The main effects of all the variables on the recovery are significant
at the 99% confidence level except the particle size. The order of influ-
ence is X4>X2>X3>X1. The most important effect is the number of
strokes (X4). It is highly significant and positive. The effect of bed
thickness (X2) is also highly significant, but negative. Of course this
variable will influence contrary to the mass productivity. The variable
water level (X3) has a positive effect. The interpretation of variables
effects on recovery in the lower layer are explained in the following
sections.
A heavier and coarser feed requires a stronger cycle than a finer
one [16]. The coarser the ore, the deeper the whole bed and hence the
greater the output is. But when the bed is too deep, the separation by
gravity is hindered. When the jig contains fine particles, there are smaller
voids between these smaller particles. This will improve hindered settle-
ment, i.e., shortening the time of stratification [1]. The motion should be

Figure 1. Residuals analysis of quartz recovery.

 120 M. M. AHMED

stronger with coarse ore than with fine ore. Although relatively short fall
strokes are used to separate fine materials, more control and better strati-
fication can be achieved by using longer, slower strokes, especially with
the coarser particle sizes. It is therefore good practice to screen the feed
to jigs into different size ranges and treat these separately [3].
The stratification process is rapid when the thickness of the jig bed is
thin. This may probably be due to excessive mobility of the bed. Thus,
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

the mineral particles of the bottom layers have the opportunity to rise
and penetrate the top layers [2].
The higher the water level, the better separation results are. This
could be attributed to a reduction in intensity and duration of suction
at higher water levels as well as to greater mobility of the settling of heavy
mineral grains [2]. At the point of transition between the pulsion and
suction stroke, the bed will be completely compacted. In a closely sized
ore, the heavy grains will have difficulty penetrating through the bed and
may be lost to tailings. Severe compaction of the bed can be reduced by
the addition of hutch water, a constant volume of water, which creates a
constant upward flow through the bed. The coarser ore then penetrates
the bed more easily and the horizontal transport of the feed over the jig is
also improved [3].
There are two opposite forces that determine the final rate of separ-
ation at changing the number of strokes. The first one favors the strati-
fication and the second delays its completion. A deeper study into the
mechanism of jigging using hydrodynamics as well as some concepts
of theory of stochastic processes reveals that jigging is a combination
of two separate, contracting processes. The first one favors the separ-
ation while the second remixes stratified particles at the same time [2].
Optimum frequency and stroke depend on the feed rate, the specific
gravity of the feed, its granulometry, bed thickness, the type of jigging
cycle employed, and the intensity of suction (the supply rate of back
water) [16].
From Equation 2, it can be also revealed that, although the particle
size (X1) has no significant effect on the recovery, its interaction with
other variables is significant at the 99% level and has a clear effect on
the recovery. It interacts with the variable bed thickness (X1  X2) and
decreases the recovery significantly. The variable (X1) interacts also with
the number of strokes (X1  X4) and increases the recovery. The variable
water level (X3), which has a positive effect, interacts with the negative
effect variable (X2) and increases the recovery. This variable (X3)
 OPTIMIZATION OF A JIGGING PROCESS 121

interacts with the variable (X4) and decreases recovery with negative
interaction. The most important effect of all interactions is X2  X4.
This interaction is highly significant and increases the recovery posi-
tively. The interaction (X2  X3  X4) is significant at the 99% confidence
level and has negative affect; i.e., it decreases the recovery.

Optimization
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

One of the techniques of optimization is the method of steepest ascent, in

which the base point is assumed and the next set of values is selected,
which is proportional to product of the coefficient and step size. The
selected values are incremented successively and objective function is
evaluated each time until the optimum point is reached.
In this work, our objective was to maximize recovery of quartz in the
lower layer of jig product. Equation 2 was used to determine the
increment size for recovery. The variables having positive effects were
increased and the variables having negative effects were decreased
according to the increment size and evaluated by carrying out successive
experiments. The results obtained with their variables are given in
Table 6.
Equation 2 indicates that the effects of the number of strokes per
minute (X4) and the water level (X3) are positive, whereas that of the
bed thickness (X2) and particle size (X1) are negative. The results of
the experiments are given in Table 6. The optimum condition was found
to be at a particle size 3.907 mm, bed thickness 1.87 cm, water level
4.41 cm, and 234.3 strokes per minute. At these conditions, a lower
layer product with 81.81% quartz at 97.74% quartz recovery was
obtained.

Table 6. Optimization of quartz recovery in the lower layer product

Variables Response

Particle size Bed thickness Water level Number of strokes Recovery Grade
(X1), mm (X2), cm (X3), cm (X4), rpm % %

4.192 2.279 4.233 208.960 91.54 85.18

4.097 2.143 4.291 217.421 95.94 82.57
4.002 2.007 4.348 225.881 97.38 82.24
3.907 1.871 4.406 234.341 97.74 81.81
 122 M. M. AHMED

CONCLUSIONS
A confidence interval of 99% was chosen for determination of significance
of main and interaction effects. Four experiments at the center points were
carried out to estimate the experimental error and variance. The main
effects of all the variables on the recovery were significant at the 99% con-
fidence level except the particle size. The order of influence was number of
strokes > bed thickness > water level > particle size. The most important
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

effect was the number of strokes that had a positive response. The effect
of bed thickness was also highly significant, but negative. The water level
had a positive effect. The empirical model was found to accurately predict
the quartz recovery where the coefficient of determination was about 0.95.
An optimum product with 81.81% grade and 97.74% recovery of quartz in
the lower layer was obtained at a particle size of 3.907 mm, a bed thickness
of 1.87 cm, a water level of 4.41 cm, and 234.3 strokes per minute.

NOMENCLATURE

k ¼ number of factors
Rexp ¼ experimental response (recovery of quartz in the lower layer), %
R2 ¼ coefficient of determination
t ¼ Student’s t test
X1 ¼ particle size, mm
X2 ¼ bed thickness, cm
X3 ¼ water level, cm
X4 ¼ number of strokes per minute, rpm
r2 ¼ variance

REFERENCES
1. Vijayendra, H. G. 2001. Gravity concentration. In A Handbook on Mineral
Dressing. 2nd ed., ed. H. G. Vijayendra, 149–183. New Delhi: Vikas Publish-
ing House PVT Ltd.
2. Herbest, J. A. 1984. Contribution to the modeling of the jigging process. In
Control ’84 Mineral=Metallurgical Processing, 97–105. New York: Society of
Mining Engineers.
3. Wills, B. A. 2006. Gravity concentration. In Mineral Processing Technology.
7th ed., ed. T. Napier-Munn, 225–245. Amsterdam: Elsevier.
4. Naik, P. K., L. B. Sukla, and S. C. Das. 2000. Aqueous SO2 leaching studies
on nishikhal manganese ore through factorial experiment. Hydrometallurgy
54: 217–228.
 OPTIMIZATION OF A JIGGING PROCESS 123

5. Naik, P. K., L. B. Sukla, and S. C. Das. 2002. Application of statistical design

in the leaching study of low grade manganese ore using aqueous sulfur diox-
ide. Separation Science and Technology 37(6): 1375–1389.
6. Naik, P. K., P. S. R. Reddy, and V. N. Misra. 2004. Optimization of coal flo-
tation using statistical technique. Fuel Processing Technology 85: 1473–1485.
7. Mukherjee, A. K., and B. K. Mishra. 2006. An integral assessment of the role
of critical process parameters on jigging. International Journal Mineral
Processing 81: 187–200.
Downloaded by [NUS National University of Singapore] at 07:03 16 June 2013

8. Mukherjee, A. K., V. K. Dwivedi, and B. K. Mishra. 2005. Analysis of a

laboratory jigging system for improved performance. Minerals Engineering
18: 1037–1044.
9. Ya-li, K., Z. Jun, Z. Hai-yang, and G. Jun. 2009. Mathematic simulation
of þ13 mm particles motion in jig. Procedia Earth and Planetary Science 1:
819–829.
10. Xia, Y., F. F. Peng, and E. Wolfe. 2007. CFD simulation of fine coal segre-
gation and stratification in jigs. International Journal of Mineral Processing 82:
164–176.
11. Montgomery, D. C., and G. C. Runger. 2002. Applied Statistics and Probability
for Engineers. 3rd ed. New York: Wiley.
12. Davies, L. 1993. Efficiency in research, development, and production. In The
Statistical Design and Analysis of Chemical Experiments, ed. L. Davies, 180.
Cambridge, UK: The Royal society of Chemistry.
13. Naik, P. K., P. S. R. Reddy, and V. N. Misra. 2005. Interpretation of interac-
tion effects and optimization of reagent dosages for fine coal flotation. Inter-
national Journal of Mineral Processing 75: 83–90.
14. Kelebek, S., U. Demir, O. Sahbaz, A. Ucar, M. Cinar, C. Karaguzel, and B.
Oteyaka. 2008. The effects of dodecylamine, kerosene and pH on batch flo-
tation of Turkey’s Tuncbilek coal. International Journal of Mineral Processing
88: 65–71.
15. Montgomery, D. C. 1991. Design and Analysis of Experiments. 3rd ed.
Singapore: Wiley.
16. Tarjan, G. 1986. Concentration by gravity. Mineral Processing 2: 7–112.
Budapest: Akademiai Kiado.