International Journal of Mining Science and Technology 28 (2018) 621–629

Contents lists available at ScienceDirect

International Journal of Mining Science and Technology

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

Application of response surface methodology (RSM) for optimization of

leaching parameters for ash reduction from low-grade coal
Sushanta Kumar Behera a,⇑, Himanshu Meena a, Sudipto Chakraborty a, B.C. Meikap a,b
a
Department of Chemical Engineering, Indian Institute of Technology (IIT) Kharagpur, West Bengal 721302, India
b
Department of Chemical Engineering, School of Engineering, Howard College, University of Kwazulu-Natal (UKZN), Durban 4041, South Africa

a r t i c l e i n f o a b s t r a c t

Article history: Coal is the world’s most abundant energy source because of its abundance and relatively low cost. Due to
Received 2 November 2017 the scarcity in the supply of high-grade coal, it is necessary to use low-grade coal for fulfilling energy
Received in revised form 10 February 2018 demands of modern civilization. However, due to its high ash and moisture content, low-grade coal
Accepted 25 March 2018
exerts the substantial impact on their consumption like pyrolysis, liquefaction, gasification and combus-
Available online 30 April 2018
tion process. The present research aimed to develop the efficient technique for the production of clean
coal by optimizing the operating parameters with the help of response surface methodology. The effect
Keywords:
of three independent variables such as hydrofluoric acid (HF) concentration (10–20% by volume), temper-
Low-grade coal
Hydrofluoric acid (HF)
ature (60–100 °C), and time (90–180 min), for ash reduction from the low-grade coal was investigated. A
Leaching quadratic model was proposed to correlate the independent variables for maximum ash reduction at the
Ash reduction optimum process condition by using central composite design (CCD) method. The study reveals that HF
Response surface methodology concentration was the most effective parameter for ash reduction in comparison with time and temper-
ature. It may be due to the higher F-statistics value for HF concentration, which effects to large extent of
ash reduction. The characterization of coal was evaluated by Fourier transform infrared spectroscopy
(FTIR) analysis and Field-emission scanning electron microscopy with energy-dispersive X-ray (FESEM-
EDX) analysis for confirmation of the ash reduction.
Ó 2018 Published by Elsevier B.V. on behalf of China University of Mining & Technology. This is an open
access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction for power generation. Nowadays various beneficiation technolo-

Coal-fired thermal power sectors are the leading energy provi-
der with significant growth in the world. India is the third largest
country in the world for coal production, and approximately 85%
of the total coal production is used in thermal power plants. Coal
plays the primary dominant energy source for the generation of
electricity, and it is expected to rise dramatically over the next
few decades. It is very difficult to remove the mineral matter from
the coal matrix by washing is due to the drift origin of Indian coal
and wide variation of nature. Therefore, suitable beneficiation
techniques are required for quality up-gradation of low-grade coal.
India is endowed with the vast reserve of non-coking coal. It is hard
to reduce the ash content to a certain limit by conventional coal
washing technique because of the presence of the inherent miner-
als in the coal matrix. Indian coals are usually high ash and mois-
ture content, which exerts a strong influence on their utilization,
and such coals are neither acceptable for iron making nor suitable
⇑ Corresponding author.
E-mail address: skbehera@che.iitkgp.ernet.in (S.K. Behera).
gies are developed for upgrading the low-quality coal that may
be appropriate for different fields. These beneficiation techniques
are alternatively called as coal preparation, in which the undesir-
able impurities are physically separated from the coal [1]. The
physical beneficiation techniques are based on the surface proper-
ties (wettability) and density of the mineral associated with the
coal matrix [2]. It is mandatory to reduce the ash content from
the coal for better utilization by required proper beneficiation
technique to minimize the pollution in the environment. Similarly,
the high sulfur bearing coals are not direct use, which also creates
environmental pollution. Therefore, pre-desulfurization of the coal
is the appropriate method to achieve the clean coal. The chemical
leaching is an effective method to remove both organic and inor-
ganic sulfur minerals from coal, as chemical solvents would easily
diffuse through the coal matrix and dissolve the minerals [3].
Therefore, chemical beneficiation is one of the novel techniques
that can remove inherent mineral and reduce ash-bearing mineral
from the coal matrix. Several authors reported that the ash reduc-
tion was obtained by various leaching methods like acid leaching
[4,5], alkali leaching under different operating conditions [6,7],
leaching by molten caustic baths [8,9] and by leaching with

https://doi.org/10.1016/j.ijmst.2018.04.014
2095-2686/Ó 2018 Published by Elsevier B.V. on behalf of China University of Mining & Technology.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
622 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629
Table 1
Chemical analyses of original coal sample.
Coal sample Proximate analysis (wt.%, dry basis)
Volatile matter Ash Fixed carbon
Ananta 30.6 39.9 29.5
calcium hydroxides [10,11]. The mineral matters in coal are cate-
gorized into carbonates, sulphates, phosphates, oxides, and sul-
phides. The presence of the major amount of silica and clay
mineral content in the coal is mainly responsible for the formation
of ash. The majority of silica and alumina found in the coal are in
the form of kaolinite, dolomite, and quartz [12].
The effectiveness of the various demineralization depends on
the composition and structure of the minerals in the coal matrix
[13]. Steel et al., investigated the leaching behavior of the mineral
matter in low-temperature ashing of Australian black coal by the
effect HF and HCl. They reported HCl can dissolve the simple com-
pounds like carbonates and phosphates, but it was less dissolute
towards the clay mineral. The investigation result showed that
HF could react with all types of mineral matter, except pyrite,
and mostly all reaction products are water-soluble [14]. Steel and
Patrick reported the production of ultra-clean coal (UCC) by chem-
ical demineralization of high volatile UK coal. The coal was leached
with HF at leaching temperature 65 °C and contact time 3 h, fol-
lowed by HNO3 in the same condition. The ash content of coal
reduced from 7.9% to 2.6% and 67% demineralization was obtained
by HF treatment alone and successive treatment of HF-treated coal
with HNO3. The ash content was decreased up to 0.63% by dissolu-
tion of residual mineral compounds such as CaF2, MgF2, AlF3,
NaAlF4 and FeS2 (pyrite) from HF treatment [15]. Rubiera et al.
investigated the combustibility characteristics demineralized low
ash bituminous coal with aqueous HF followed by HNO3. The result
was obtained from the investigation of over 95% of ash, and the SO2
level was reduced by the leaching treatment. However, NOx emis-
sion was increased with nitric acid treatment [16]. Ning et al. stud-
ied the desulfurization of coal with a chemical reagent in
combination with microwave irradiation [17]. By this investiga-
tion, the pyritic sulfur significantly decreased from 53.6% to
39.2% and organic sulfur reduced from 20.1% to 16.1% without
the changes in coal properties.
Recently many statistical experimental design methods are
used in the different chemical sector for optimization of process
parameters. The optimization of a multivariable system in the con-
ventional technique follows one-factor at a time. Many experi-
ments are required for conventional techniques, and such
methods do not represent the combined effect. It also requires
more data to determine optimum level and take prolonged time,
which is unreliable [18]. The primary purpose of the experimental
design technique is to understand the interactions among the
parameters, which could help in the optimization of experimental
parameters and provide statistical models [19]. Choudhury studied
the statistical and dimensional analysis for demineralized coal
with NaOH and HCl by using fractional and factorial design method
[20]. It has been found from the literature very limited work has
been done on statistical optimization of process condition for ash
reduction from low-grade coal by using HF acid treatment. There-
fore, the present research aims to investigate the statistical opti-
mization to determine the optimum leaching conditions for ash
reduction from low-grade Indian coal by using aqueous HF
leaching.
This paper investigates the combined effects of concentration,
time and temperature on the leaching process, and the process
parameters were optimized using central composite design (CCD)
in conjunction with RSM method. A model was developed by using
Ultimate analysis (wt.%, daf basis)
Carbon Hydrogen Nitrogen Sulfur Oxygen (difference)
39.37 3.21 0 0.37 57.05
the design of experiment to determine the optimum leaching con-
dition where the maximum ash reduction was obtained from the
coal. In addition, this paper presents the comparison between the
characteristic of raw and leached coal at the optimum condition
with the help of FTIR and FESEM-EDS analysis.
2. Material and methods
2.1. Material
The coal sample was collected from Ananta open cast project of
Mahanadi coalfield, Talcher, Orissa. It is one of the major coal pro-
ducers in India, and a subsidiary of Coal India Limited. The lumped
coal sample was crushed and sieved to
30+72 British standard
sieve (BSS) mesh size required for the present study. Before the
start of the experiment, the sample was dried in a hot air oven
for 12 h at 105 °C. The proximate and ultimate analyses of the sam-
ple were determined in ASTM standard [21–24], and the results
were summarized in Table 1. The analytical grade chemical reagent
HF acid was used in this study. In this work, the coal sample was
treated with aqueous hydrofluoric acid (HF) by varying the concen-
tration at different operating parameter.
2.2. Methods
A series of experiments were conducted to find the maximum
ash reduction from the coal where the optimum process condition
maintained. For this purpose, the leaching experiments were car-
ried out in a 1000 ml of Teflon reactor. A magnetic stirrer with a
hot plate was used for the stirring and heating purpose. The exper-
imental leaching set-up was shown in Fig. 1. The different concen-
tration (10%, 12%, 15%, 18% and 20% by volume) of HF solutions
were prepared for the leaching experiment. In each case, 10 g coal
sample was treated with 100 ml HF solution. The reaction resi-
dence time varies from 60 min to 180 min, and temperature ranges
from 60 °C to 100 °C. A plastic lid was covered on the top of the
reactor to prevent the escape of vapours. After subsequent leach-
ing, the treated sample is washed with distilled water to dilute
the HF concentration and filtered in polypropylene funnel by using
Whatman filter paper and dried in the hot air oven at 90 °C for 2–3
h. Finally, the ash reduction was calculated from Eq. (1).
Fig. 1. Leaching experimental setup.
 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629 623

Raw coal ash
Leached coal ash Table 2

Ash reduction ð%Þ ¼  100 ð1Þ Relationship between the coded and actual value of
Raw coal ash
the variables.

Code The actual level of variable

2.3. Multivariate design of experiment

a Xmin

1 [(Xmax + Xmin)/2]
[(Xmax
Xmin)/2b]
The parameter used for ash reduction from low-grade coal by 0 (Xmax + Xmin)/2
chemical leaching was analyzed by standard response surface +1 [(Xmax + Xmin)/2]+[(Xmax
Xmin)/2b]
methodology (RSM) design called central composite design +a Xmax
(CCD). RSM method is suitable for fitting a quadratic surface and Note: b is 2n/4; n is the number of variables (con-
it helps to optimize the process parameters with a minimum num- centration, time and temperature), and n = 3.
ber of experiments, as well as to analyze the interaction between
the parameters [25]. RSM is a collection of mathematical and sta-
tistical techniques useful for developing the empirical model build- for leaching process; and e the random error. The codes are calcu-
ing, improving and optimizing processes parameter and it can also lated as a function of the range of interest of each factor as shown
be used to find the interaction of several affecting factors [26]. RSM in Table 2 [30]. In developing the regression equation, the test vari-
is a statistical method that uses quantitative data from the related ables were coded according to Eq. (4).
experiment to determine regression model and to optimize a
ðX i
X i Þ
response (output variable) which is influenced by several indepen- xi ¼ ð4Þ
dent variables (input variables).
DX i
Generally, CCD consists of a 2n factorial runs with 2n axial runs, where xi is the ith independent variable of the dimensionless coded
and the experimental error is measured by center runs (nc). This value; the uncoded value of the ith independent variable is denoted
experimental design is composed of 2n factorial with coded by ±1 by Xi; similarly, at the center point, the uncoded value of the ith
notation augmented by 2n axial points (±a, 0, 0. . .0), (0, ±a, 0. . .0) independent variable denoted by X i ; and the step change value
. . . (0, 0, ±a . . .0) and nc center points (0, 0, 0 . . . 0) [27]. Each vari- has been defined by DX i [31].
able is investigated at two levels while as the number of variables The number of tests required for the CCD includes N as a total
(n) increases, the number of runs for a complete replicate of the number of run or experiments, the standard 2n factorial points
design increases rapidly. CCD was employed for the quadratic with its origin at the center. The quadratic terms generate from
effect because individual second order effect cannot be estimated the center at a distance a (in this study a value was fixed at
separately by 2n factorial designs. Therefore, CCD was used to 1.6817), from a fixed 2n axial points and at the center run the repli-
develop the model for ash reduction from coal. In the statistical cates to avoid error; where n is the number of independent vari-
analysis, the response with corresponding parameters was mod- ables. The axial points (2n) are for screening analysis and
elled to optimize the process conditions for the desired response. readability, which checks the variance of model prediction and is
ANOVA is used to calculate the statistical parameters with the constant at all points equidistant from the design center. nc is the
aid of response surface methods. number of central points that provide replicates of the test at the
Optimization by RSM method involves three major steps; these center, and are very necessary for providing an independent esti-
are firstly statistically designed experiments, secondly, estimate mation of the experimental error [32].
the coefficients in a mathematical model and finally predicting
the response and checking the adequacy of the model within the N ¼ 2n þ 2n þ nc ¼ 23 þ ð2  3Þ þ 6 ¼ 20 ð5Þ
setup of the experiment [28]. In this study, three independent vari-
For three variables in the experiments, eight factorial points, six
ables were chosen for the statistical experiment design as follows:
axial points and six replicates at the central points were used for
leachant concentration (X1, vol.%), temperature (X2, °C), and time
the CCD design. The number of points was shown in Fig. 2. Hence,
(X3, min). The range and level of the factors varied accordingly to
the total number of tests (N) required for the three independent
the experimental design. These three variables together with their
variables is 20 as shown in Eq. (5). Once the desired range value
respective ranges were found to be critical parameters for efficient
of the variables is defined, they are coded to ±1 for the factorial
ash reduction from low-grade coal.
points, 0 for center points and ±a for the axial points.
Y ¼ f ðX 1 ; X 2 ; X 3 . . . X n Þ ð2Þ
where Y is the response of the system and Xn is the independent 2.4. Model fitting and statistical analysis
variable of action called factors.
It is assumed in the design, the independent variable in the ANOVA (Analysis of variance) has been used for graphical anal-
experiments are continuous and regulated by experiments with yses of the data to define the interaction between the process vari-
negligible errors. The aim of the experimental design was to opti-
mize the response variables (Y). It needed to find a suitable approx-
imation for the true correlation between independent variables
and response surfaces [29]. The experimental run was randomized
in order to diminish the error and effect of the uncontrolled factors.
The response was used to generate an empirical model, which cor-
relates to the experimental variables using a second-degree poly-
nomial equation Eq. (3).
Xn Xn Xn Xn
Y ¼ b0 þ bX
i¼1 i i
þ b X2
i¼1 ii i
þ i¼1
b XX
j¼iþ1 ij i j
þe ð3Þ

where Y is the predicted response; b0 the constant coefficient; bi the

linear coefficients; bii the quadratic coefficients; bij the interaction
coefficients; n the number of factors studied and optimized in the
experiments; Xi and Xj the coded values of the variable parameters Fig. 2. Location of points in the 23 factorial design.
624 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629

ables and the responses to estimate the statistical parameters. The were removed from the coal by the leaching effect of HF acid
statistical software package Design Expert, Stat-Ease, Inc., Min- and simultaneously reduce the ash-bearing minerals.
neapolis, USA, has been used for the regression analysis of experi-
mental data, to plot the response surfaces, and contour plot at the
Al2 Si2 O5 ðOHÞ4 ðsÞ þ 14HFðaqÞ ! 2AlF3 ðaqÞ þ 2SiF4 ðaqÞ þ 9H2 O
optimized condition. The statistical significance was checked by ð6Þ
the F-test in the same program. The accuracy of the fitted polyno-
mial model was determined by the coefficient of R2. The significant 4HF þ SiO2 ¢ SiF4 þ 2H2 O ð7Þ
model terms were evaluated by the probability value (P-value) at
95% confidence interval.
3.2. Development of model

3. Results and discussion

3.1. Leaching mechanism of HF acid treated coal
The typical ash analysis of Indian coal revealed that more
than 50% silica was found in the coal ash [33]. The significant
quantity of silica and alumina are presented in the form of alu-
minosilicate compound such as clays ((NaMgAl)Si4O10(OH)2),
Kaolinite (Al2Si2O5(OH)4), and quartz (SiO2). The leaching process
mechanism may occur by the HF acid diffusing inside of the coal
matrix, which reacts to the major minerals and dissolves almost
the entire Al, Si and K components and other mineral com-
pounds [34]. The dissolution behavior of minerals by leaching
is a function of concentration, time and temperature. HF acid
is the most efficient reagent with greater potential of elution
in comparison to other chemical solvents. The reason behind
for higher degree of dissolution may be due to the high affinity
of fluoride ion (F
) towards the minerals in the coal matrix. The
major minerals (clays, kaolinite) of coal were reacted with HF
acid during the leaching and formed soluble fluoride compounds
like CaF2, MgF2, AlF3 and SiF4, NaAlF4 as shown in Eqs. (6) and
(7). The elements like Na, Mg, Al, Ca from the different minerals
The statistical ‘‘Design-Expert 8.0.7.1” software has been used
to study the regression analysis of experimental data and to draw
the response surface plot. The statistical parameters were esti-
mated by using ANOVA. For the leaching study, the required exper-
imental range and coded level of variables are given in Tables 3 and
4, which shows the design of experiments together with the exper-
imental results. As the output proposed by the Design-Expert soft-
ware, the quadratic model was not aliased. The final empirical
model in terms of a coded factor for ash reduction (Y, %) is shown
in Eq. (8):
Ash reduction ðY; %Þ ¼ 74:48 þ 3:66X 1 þ 0:85X 2 þ 0:43X 3

0:071X 1 X 2
0:064X 1 X 3

0:11X 2 X 3
0:96X 21 þ 0:33X 22
þ 0:28X 23 ð8Þ
where the negative sign indicates the antagonistic effects whereas
the positive sign indicates the synergistic effects.
In order to fit a good model, a test for significance of the regres-
sion model and individual model coefficients with lack of fit test
was performed. Usually, the significant factors were ranked based

Table 3
Experimental independent variables and their coded levels for the central composite design.

Independent variable Symbol Levels of coded variables

a Low Medium High +a

1.6817
1 0 +1 +1.6817
Leachant concentration (vol.%) X1 10 12 15 18 20
Temperature (°C) X2 60 68 80 92 100
Time (min) X3 90 108 135 162 180

Table 4
Experimental factors in coded and actual units and experimental responses.

Standard run no. Run Independent variable in coded Independent variable in actual form Ash reduction Y (%)
form
X1 X2 X3 X1 (vol.%) X2 (°C) X3 (min)
1 20
1
1
1 12 68 108 69.25
2 4 +1
1
1 18 68 108 75.93
3 5
1 +1
1 12 92 108 70.68
4 14 +1 +1
1 18 92 108 78.36
5 1
1
1 +1 12 68 162 70.15
6 19 +1
1 +1 18 68 162 77.86
7 6
1 +1 +1 12 92 162 72.43
8 3 +1 +1 +1 18 92 162 78.57
9 13
a 0 0 9.95 80 135 65.27
10 12 +a 0 0 20.05 80 135 78.21
11 2 0
a 0 15 59.82 135 73.98
12 15 0 +a 0 15 100.18 135 76.84
13 16 0 0
a 15 80 89.59 74.93
14 7 0 0 +a 15 80 180.41 75.61
15 17 0 0 0 15 80 135 74.64
16 10 0 0 0 15 80 135 74.58
17 9 0 0 0 15 80 135 74.24
18 8 0 0 0 15 80 135 74.74
19 11 0 0 0 15 80 135 74.16
20 18 0 0 0 15 80 135 74.55
 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629 625

Table 5
Analysis of variance (ANOVA) for response surface quadratic model for ash reduction of coal.

Source Sum of squares Degree of freedom Mean square F value Prob. > F (P value) Remarks
Model 213.33 9 23.7 120.95 <0.0001 Significant
X1 182.86 1 182.86 933.04 <0.0001 Significant
X2 9.95 1 9.95 50.80 <0.0001 Significant
X3 2.58 1 2.58 13.15 0.0046 Significant
X1X2 0.041 1 0.041 0.21 0.6587
X1X3 0.033 1 0.033 0.17 0.6924
X2X3 0.095 1 0.095 0.48 0.5030
X 21 13.37 1 13.37 68.22 <0.0001 Significant
X 22 1.61 1 1.61 8.22 0.0167 Significant
X 23 1.17 1 1.17 5.97 0.0347 Significant
Residual 1.96 10 0.196
Lack of fit 1.69 5 0.338 6.31 0.0322 Significant
Pure error 0.27 5 0.054
Correlation total 215.29 19

on the F-value or P-value (probability value) with 95% confidence

level. Table 5 shows the ANOVA for the data generated by Eq. (8)
for ash reduction from low-grade coal. The larger F-value and the
smaller ‘P’ value (Prob. > F), show more significant of the corre-
sponding coefficient [35]. The F-value of 120.95 indicates that
the model is significant. Additionally, the model terms are signifi-
cant only when the values of ‘‘Prob. > F” are less than 0.05. In this
case, X1, X2, X3, X 21 , X 22 and X 23 are significant model terms. The ‘‘Lack
of fit F-value” of 6.31 implies that the ‘‘Lack of fit” is significant. The
larger P values (for ‘‘Lack of fit” > 0.05) in Table 5 shows F statistics
value is insignificant. There is only 3.22% chance that a ‘‘Lack of fit
F-value” occurs due to noise.
Level of each variable was calculated based on preliminary
extrusion. The distance of the axial points from the center was
±1.6817 and it was calculated from a = (2n)0.25 where n is the num-
ber of variables.
Fig. 4. Predicted ash reduction of coal and studentized residuals plot.
The signal to noise ratio is measured by adequacy precision,
which comprised the predicted value at the design points and
the average prediction error. In the present study, adequacy preci-
original observation is constant, and this was an indication that
sion ratio is 42.289 and is desirable due to this ratio has been
there was no need for transformation of response variables. The
greater than 4. Hence, the developed model can be used to guide
actual and the predicted percentage for ash reduction was shown
the design space.
in Fig. 5. It was found that the values of R2 and adjusted R2 (Radj)
The adequacy of the developed model was the main part to
were 99% and 98% respectively. The value of R2 describes up to
check the data analysis of the experiment. The normal probability
what extent perfectly model can estimated the experimental data
and studentized residual plot were shown in Fig. 3 for ash reduc-
points and the adjusted R2 measured the amount of variation about
tion percent from coal. It has been found from Fig. 3 that there
mean explained by the model. The predicted R2 value for Eq. (6)
was neither response transformation nor any apparent problem
was 0.9703, which was near close to R2 value. It was revealed that
with normality. Fig. 4 shows the internally studentized residual
versus predicted ash reduction percent. This signifies that for all
values of the response, the random scatter plot, the variance of

Fig. 3. Studentized residuals and normal percentage probability plot for ash
reduction of coal. Fig. 5. Actual and predicted plot of coal ash reduction (%).
626 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629

Table 6
Statistical parameters obtained from the analysis of variance (ANOVA) for the models
for ash reduction percentage from low-grade coal.

Insignificant factors excluded Ash reduction Y (%)

Standard deviation (S.D.) 0.44
Mean 74.25
Coefficient of variance (C.V) % 0.60
Adeq. precision 42.289
R2 0.9909
Adjusted R2 0.9827
Predicted R2 0.9325
PRESS 14.54

the experimental data for ash reduction percent from low-grade
coal fitted well with the predicted value of the model. The standard
deviation for the model was 0.4. The small value of standard devi-
ation indicates good model that gives near value between pre-
dicted and actual values for the responses. The statistical
parameter obtained from ANOVA was shown in Table 6.
3.3. Combined effect of temperature, concentration and time on
percent ash reduction
The response surface methodology was used to investigate the
individual and interaction effect of the three-factor on ash reduc-
tion from the low-grade coal. Based on ANOVA, the results were
obtained, the effects of experimental factors on ash reduction, cor-
responding three-dimensional response surface plots were shown
in Figs. 6–8 and the response model was represented in Eq. (8).
The HF concentration, time and temperature have significant
effects on the ash reduction from coal. By the use of response sur-
face analysis method, it was observed from Table 4 that the leach-
Fig. 6. Combined effect of concentration and temperature on ash reduction at time
135 min.
Fig. 7. Combined effect of concentration and time on ash reduction at temperature
80 °C.
Fig. 8. Combined effect of temperature and time on ash reduction at 15% HF acid
concentration.
ing parameters influence the ash reduction (Y) from the coal. It has
been found from Table 5, HF concentration (X1) was found to the
major effect on the ash removal from low-grade coal in comparison
to other variables, which may due to the high F value of 933.04 for
HF acid. But the temperature and time have been found less signif-
icant effects for ash reduction from coal. However, the interactions
between the variables have no effect on ash reduction, only the
concentration was found to be more efficient on ash reduction
from coal. The quadratic function of leaching concentration and
temperature demonstrated almost similar effects on the response,
which was the lower level significance compared to the quadratic
function of time.
The response surface methodology (RSM) was used to study the
three-dimensional response plots, which were generated from the
effects of the three variables on ash reduction from low-grade coal
by HF acid leaching. Table 5 shows the ANOVA for a response sur-
face quadratic model for ash reduction from coal. It has been found
from the RSM, the leaching variables were involved more or less in
increasing the effectiveness of ash reduction (Y).
Figs. 6–8 demonstrate the interactions between the variables in
three-dimensional response surface plots. The values of the axis in
the Figs. 6–8 are real values. Fig. 6 indicates the combined effect of
concentration (X1) and temperature (X2) on ash reduction at con-
stant time (135 min). The ash reduction increased due to the influ-
ence of temperature and concentration at constant reaction time.
The maximum ash reduction of 78.12% was obtained. The interac-
tion between the concentration and temperature was the most
effective parameter for ash removal in the leaching process.
It can be observed from Figs. 6 and 8, the ash reduction increase
with the rise of temperature. This may be due to an increase in
temperature results in the increase of the rate of reaction and
the activation energy of the reaction. At the constant time of the
leaching, the concentration increases from 15% to 20% and temper-
ature increases from 80 °C to studied range, which caused the
increase of the ash reduction from 74% to 78.12%. The reason for
increased ash reduction observed from the ANOVA table is the
higher ‘F’ value of HF concentration and temperature, which indi-
cates the greater extent of leaching effect [36]. The combined effect
of concentration (X1) and time (X3) on ash reduction from low-
grade coal at constant temperature (80 °C) was shown in Fig. 7.
The maximum ash reduction has been found to be 77.66% by the
effect of concentration and time at stable temperature.
Similarly, the combined effect of temperature (X2) and time (X3)
on ash reduction at 15% HF acid concentration and corresponding
three-dimensional response surface were illustrated in Fig. 8. It
can be seen from the results, time and temperature have fewer
effects on ash reduction from coal because of the low value of F
statistics. The maximum ash reduction of 76.24% was obtained
by the effect of time and temperature at 15% HF concentration.
 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629
3.4. Optimization by response surface modelling
The vital part of the experimental study was to determine the
optimum leaching process condition where maximum ash reduc-
tion can be obtained from the low-grade coal. Optimization of
the leaching variable parameter was carried out in a numerical
optimization method. The response surface and contour plot at
optimum leaching condition for maximum ash reduction were
shown in Figs. 9 and 10. The optimum operating conditions for
ash reduction from the coal by leaching process were 18% HF acid
concentration by volume, reaction temperature 92 °C and leaching
time 162 min. The predicted and experimental values for the ash
reduction from low-grade coal were obtained as 78.84% and
78.91% at optimum conditions, respectively. A comparison
between the experimental and predicted results indicates the error
was less than 0.08%. From the results, it was concluded that the
developed model could accurately predict the ash reduction. The
detail of optimization condition is shown in Table 7.
3.5. Characteristics of coal
3.5.1. SEM-EDS analysis
The scanning electron microscopy coupled with energy disper-
sive X-ray (SEM-EDX) analysis was used to determine the nature of
mineral matter of the untreated and treated coal samples by Field
Fig. 9. Optimum region on the temperature and concentration for maximization of
ash reduction from low-grade coal.
Fig. 10. Contour plot of temperature and concentration on ash reduction at
optimum condition.
Table 7
Optimal processing conditions and model validation of percentage ash reduction of coal from numerical optimization.
Parameters Leachant concentration X1 (vol.%)
Optimum conditions 18.00
627
Electron Scanning Electron Microscope (FESEM) model JEOL JSM-
7500F, Japan. The SEM sample was prepared by sprinkling pow-
dered coal sample on to a carbon coated metallic holder followed
by gold coating. The reason behind for gold coating is the conduct-
ing behavior of coal samples. Figs. 11 and 12 show the scanning
electron microscope (SEM) and EDX images of the raw coal, at
the optimum condition and at 20% HF acid treated coal. The
SEM-EDX study (Fig. 11a and 12) revealed that the raw coal was
associated with less percentage of carbon content and the major
amount of silica and alumina bearing mineral and also bearing less
percent of sulfur. It can be seen from the results the original carbon
content of coal was not lost during the HF treatment. It was found
from the EDX analysis the sulfur content increased in the HF-
treated coal because the sulfur-bearing minerals were not soluble
during leaching and the results were shown in Fig. 11b, c and 12.
Only the inorganic sulfur-bearing mineral like pyrites (FeS2), was
removed by alkali leaching from Eq. (9). The increased voids on
the surface of treated coal was an indication of the significant
amount of inorganic elements being liberated by the leaching
effect. Major content of carbon was found in optimum condition,
which signified the greater amount of mineral constituents were
liberated by the leaching effect. The leachant is itself as HF acid,
so the fluoride content increased in the leached coal. However,
the residual fluoride can be washed by alkali leaching.
2FeS2 þ 6NaOH ! NaFeO2 þ 4Na2 S þ 2H2 O ð9Þ
3.5.2. FTIR analysis of coal
FTIR (Fourier transform infrared) spectroscopy is a powerful
tool for analyzing the characterization of coal and its products as
it signifies the comprehensive view about the structure/composi-
tion of the minerals in the coal matrix. In this investigation, the
qualitative analysis of before and after leached coal at the opti-
mized condition was determined. Infrared (IR) spectroscopy is
used for the study of interactions between the matter and electro-
magnetic fields in the IR region. The probability of a particular IR
frequency being absorbed depends on the actual interaction
between the frequency and the molecule. Planck energy relation
is used for calculating the energy of an infrared photon using E =
hv, where h (Planck’s constant) = 6.6  10
34 Js and m is the fre-
quency of the photon (s
1). The frequency (m) and speed of light
(c) are related to the relation, c = kv, where c = 3  108 m/s and k
is the wavelength of the light. The c and E are correlated to a stan-
dard spectroscopic unit called wave number (v , cm
1), which is the
reciprocal of the wavelength. E ¼ hck ¼ hcv , both frequency and
wavenumber are directly proportional to energy and this indicates
the higher energy photons have the higher frequency [37]. The
spectrum in the FTIR analysis is mainly presented as transmittance
versus wave number. The presence and elimination of functional
group and bonding characteristic of the raw and upgraded coal
were determined through FTIR measurement using a Nicolet
NEXUS-670 FTIR instrument. The measuring infrared region in FTIR
spectroscopy range from 4000 cm
1 to 450 cm
1 wave number
[38–40]. In the presence of baking lamp, approximately 1 mg of
ground coal sample was properly mixed with 100 mg of dried
KBr powder then it was grounded and pestled with mortar to avoid
moisture absorption. The mixture was pressed into a pellet on a
hydraulic pelleting machine. The software was used for data pro-
Temperature X2 (°C) Time X3 (min) Ash reduction Y (%)
Predicted Experimental
92.00 162 78.84 78.91
628 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629
Fig. 11. SEM image with mapping of elementary composition for raw coal, and leached coal at optimum condition and 20% HF acid.
Fig. 12. EDS analysis of coal samples.
cessing and spectral analysis. To access the presence functional
groups in the sample FTIR analysis was performed. The respective
transmittance value at particular wavelength signifies the presence
of the functional groups. However, several functional groups may
absorb at the same frequency range, and a functional group may
have multiple-characteristic absorption peaks, especially for
1500–650 cm
1 called the fingerprint region.
FTIR profiles of pre- and post-treated coal, and leached coal at
the optimized condition where maximum ash reduction was found
were shown in Fig. 13. These profiles were found similar character-
istic. However, it can be differentiated between structure and
intensity of peaks to corresponding functional groups. The pres-
ence of high volatile matter and oxygen in the raw coal presumed
the absorption of radiation by the organic part of raw coal. The
peak assigned at 3620 cm
1 corresponding to OH vibration, which
disappeared at the 20% HF acid leached coal (Fig. 13c). This was
due to the hydroxyl group was removed from the organic fraction
of coal. The peak at 2921 cm
1 and 3000 cm
1 are assigned to ali-
phatic CAH stretching for both raw and leached coal. These peaks
Fig. 13. FTIR pattern for raw coal, and leached coal at optimum condition and 20%
HF acid.
were indicated to CH3 and CH2 groups. Similarly, from Fig. 13a and
b, there was no change observed for peaks at 1430 cm
1 and 1700
cm
1 which refer to aromatic C@C, aromatic C@O respectively. The
peaks at 400–1100 cm
1 wave number were related to all inor-
ganic mineral matter such as quartz (Q), kaolinite (K), and mont-
morillonite (M). Hence, the peaks at this corresponding wave
number was called as mineral band [41]. The mineral band of coal
usually identified that peaks at 535 cm
1 and 1050 cm
1 belong to
montmorillonite, peaks at 799 cm
1 and 1165 cm
1 were corre-
sponding to quartz, and peaks at 1035 cm
1 was called kaolinite
[42,43]. It was difficult to identify the infrared bands in the spectra
for silica and alumina because the complex and non-constant com-
positions of minerals like montmorillonite, kaolinite, quartz and
other clay minerals were due to the isomorphous replacement of
band positions. These are all predominant minerals in most of
the coals. It may also expect the band varied with the complex
structure and poly-component of coal minerals.
4. Conclusions
A facile single-step method was employed to immobilize HF
acid leaching into low graded Indian coal from Ananta open cast
project, Odisha. The extent of ash reduction was maximized at
optimized leaching conditions to obtain high-grade coal by acid
leaching. Successful immobilization of HF acid leaching was con-
firmed by the structural changes of coal, the liberation of minerals
and surface characteristics of coal before and after the treatment
process. FTIR results stipulated that the mineral bands of coal
reduced by the leaching effect, which subsequently improved coal
quality. FESEM-EDS results found that the carbon content
increased in the leached coal and almost all Si, Al, K components
associated with the minerals were soluble in the aqueous HF acid.
The maximal ash reduction from low-grade coal at optimum leach-
ing conditions by applying RSM in conjunction with CCD was dis-
cussed. The regression analysis and optimization of variables are
calculated by using design expert for predicting the response in
the experimental regions. The statistical significance, regression
analysis and response surface analysis were carried out using the
experimentally obtained values at variable operating conditions.
A model was formulated to correlate the leaching experimental
variables to the responses. Based on models, the response surfaces
were derived. Concentration was found to have the most signifi-
cant effect on ash removal from coal to produce clean coal. Leach-
ing process optimization was carried out and the experimental
values obtained for the ash reduction were found closer to the val-
ues obtained as predicted by the models. Optimal conditions for
maximizing the ash reduction were obtained at HF concentration
18 vol.%, reaction temperature 92 °C and leaching time 162 min.
The absolute error was obtained as 0.08%.
 S.K. Behera et al. / International Journal of Mining Science and Technology 28 (2018) 621–629 629

References [23] ASTM Standard D3175-11. Standard test method for volatile matter in the
analysis sample of coal and coke. West Conshohocken, PA: ASTM
International; 2011.
[1] Aaron N, Gerald HL. A review of state of the art processing operations in coal
[24] ASTM Standard D5373-08. Standard test methods for instrumental
preparation. Int J Min Sci Technol 2015;25:511–21.
determination of carbon, hydrogen, and nitrogen in laboratory samples of
[2] Meshram P, Purohit BK, Sinha MK, Sahu SK, Pandey BD. Demineralization of
coal. West Conshohocken, PA: ASTM International; 2008.
low grade coal – a review. Renew Sustain Energy Rev 2015;41:745–61.
[25] Azargohar R, Dalai AK. Production of activated carbon from Luscar char:
[3] Longjun X, Tiefeng P, Dingyue Z, Fukai Z. Fractal desulfurization kinetics of high
experimental and modeling studies. Micropor Mesopor Mater
sulfur coal. Int J Min Sci Technol 2012;22:111–4.
2005;85:219–25.
[4] Mukherjee S, Borthakur PC. Chemical demineralization/desulfurization
[26] Montgomery DC. Design and analysis of experiments. 8th ed. New York: John
of high sulfur coal using sodium hydroxide and acid solutions. Fuel
Wiley and Sons; 2014.
2001;80:2037–40.
[27] Myers RH. Response surface methodology. New York: Allyn and Bacon; 1971.
[5] Ochoa J, Bonelli PR, Cassanello MC, Cukierman AL. Effect of thermal and acid
[28] Mahalik K, Sahu JN, Patwardhan A, Meikap BC. Statistical modeling and
treatments on properties of Argentinean low rank coals. Energy Sources
optimization of hydrolysis of urea to generate ammonia for flue gas
2001;23:383–92.
conditioning. J Hazard Mater 2010;182:603–10.
[6] Behera SK, Chakraborty S, Meikap BC. Chemical demineralization of high ash
[29] Gunaraj V, Murugan N. Application of response surface methodologies for
Indian coal by using alkali and acid solutions. Fuel 2017;196:102–9.
predicting weld base quality in submerged arc welding of pipes. J Mater
[7] Balaz P, LaCount RB, Kern DK, Turcaniova L. Chemical treatment of coal by
Process Technol 1999;88:266–75.
grinding and aqueous caustic leaching. Fuel 2001;80:665–71.
[30] Napier-Munn TJ. The central composite rotatable design JKMRC. Brisbane: The
[8] Duz MZ, Erdogan S, Saydut A, Merdivan M, Hamamci C. Effect of molten caustic
University of Queensland; 2000.
leaching on demineralization and desulfurization of asphaltite. Energy
[31] Sen RK. Response surface optimization of the critical media components for
Sources, Part A: Recov Util Environ Eff 2008;30:1637–44.
the production of surfactin. J Chem Tech Biotechnol 1997;68:263–70.
[9] Lee SH, Shon EK. Effect of molten caustic leaching on the combustion
[32] Box GEP, Hunter JS. Multi-factor experimental design for exploring response
characteristics of anthracite. Fuel 1997;76:241–6.
surfaces. Ann Math Stat 1957;28:195–241.
[10] Wang J, Tomita A. Hydro thermal reaction of Ca(OH)2 with quartz in
[33] Choudhury R, Bhaktavatsalam AK. Beneficiation of Indian coal by chemical
connection with coal demineralization. Ind Eng Chem Res 1997;36:1464–9.
techniques. Energy Convers Manage 1997;38:173–8.
[11] Wang J, Zhang Z, Kobayashi Y, Tomita A. Chemistry of Ca(OH)2 leaching on
[34] Steel KM, Besida J, O’Donnell TA, Wood DG. Production of Ultra Clean Coal Part
mineral matter removal from coal. Energy Fuels 1996;10:386–91.
II – ionic equilibria in solution when mineral matter from black coal is treated
[12] Steel KM, Besida J, O’Donnell TA, Wood DG. Production of Ultra Clean Coal Part
with aqueous hydrofluoric acid. Fuel Process Technol 2001;70:193–219.
I – dissolution behavior of mineral matter in black coal toward hydrochloric
[35] Yi S, Su Y, Qi B, Su Z, Wan Y. Application of response surface methodology and
and hydrofluoric acids. Fuel Process Technol 2001;70:171–92.
central composite rotatable design in optimization; the preparation condition
[13] Mukherjee S, Borthakur PC. Effect of leaching high sulfur subbituminous coal
of vinyltriethoxysilane modified silicalite/polydimethylsiloxane hybrid
by potassium hydroxide and acid on removal of mineral matter and sulfur.
pervaporation membranes. Sep Purif Technol 2010;71:252–62.
Fuel 2003;82:783–8.
[36] Niya AA, Daud WMAW, Mjalli FS, Abnisa F, Shafeeyana MdS. Production of
[14] Steel KM, Besida J, O’Donnell TA, Wood DG. Production of ultra clean coal Part
microporous palm shell based activated carbon for methane adsorption:
III – effect of coal’s carbonaceous matrix on the dissolution of mineral matter
Modeling and optimization using response surface methodology. Chem Eng
using hydrofluoric acid. Fuel Process Technol 2002;76:51–9.
Res Des 2012;90:776–84.
[15] Steel KM, Patrick JW. The production of ultra clean coal by chemical
[37] Behera SK, Chakraborty S, Meikap BC. Removal of ash from low grade Indian
demineralization. Fuel 2001;80:2019–23.
coal by chemical leaching technique. Miner Process Extr Metall Rev
[16] Rubiera F, Arenillas A, Arias B, Pis JJ, Suárez-Ruiz I, Steel KM, et al. Combustion
2018;39:59–67.
behavior of ultra clean coal obtained by chemical demineralization. Fuel
[38] Ge L, Zhang Y, Wang Z, Zhou J, Cen K. Effects of microwave irradiation
2003;82:2145–51.
treatment on physicochemical characteristics of Chinese low-rank coals.
[17] Ning X, Tao XX. Changes in sulfur form during coal desulfurization with
Energy Convers Manage 2013;71:84–91.
microwave: effect on coal properties. Int J Min Sci Technol 2015;25:435–8.
[39] Ye Q, Jia ZZ, Zheng CS. Study on hydraulic-controlled blasting technology for
[18] Sahu JN, Acharya J, Meikap BC. Response surface modeling and optimization of
pressure relief and permeability improvement in a deep hole. J Petrol Sci Eng
chromium (VI) removal from aqueous solution using Tamarind wood activated
2017;159:433–42.
carbon in batch process. J Hazard Mater 2009;172:818–25.
[40] Tao YJ, Zhang L, Tao DP, Xian YS, Sun QX. Effects of key factors of rotary
[19] Alam MZ, Muyibi SA, Toramae J. Statistical optimization of adsorption
triboelectrostatic separator on efficiency of fly ash decarbonization. Int J Min
processes for removal of 2,4-dichlorophenol by activated carbon derived
Sci Technol 2017;27(6):1037–42.
from oil palm empty fruit bunches. J Environ Sci 2007;19:674–7.
[41] Dash PS, Sriramoju SK, Kargupta K, Banerjee PK, Ganguly S. Characterization of
[20] Choudhury S. Studies on demineralization of coal: fractional factorial design.
chemical beneficiated Indian coals. Int J Coal Prep Util 2015;35:257–72.
Int J Inno Tech Res 2013;1:13–8.
[42] Painter PC, Coleman MM, Jenkins RG, Whang Jr PW, Walker PL. Fourier
[21] ASTM Standard D3173-11. Standard test method for moisture in the analysis
transform Infrared study of mineral matter in coal. A novel method for
sample of coal and coke. West Conshohocken, PA: ASTM International; 2011.
quantitative mineralogical analysis. Fuel 1978;57:337–44.
[22] ASTM Standard D3174-11. Standard test method for ash in the analysis sample
[43] Painter PC, Snyder RW, Pearson DE, Kwong J. Fourier transform infrared study
of coal and coke from coal. West Conshohocken, PA: ASTM International; 2011.
of the variation in the oxidation of the coking coal. Fuel 1980;59:282–6.