Renewable Energy 77 (2015) 512e520

Contents lists available at ScienceDirect

Renewable Energy
journal homepage: www.elsevier.com/locate/renene

Optimization of rice husk pretreatment for energy production

Alireza Bazargan a, Majid Bazargan b, Gordon McKay a, c, *
a
Department of Chemical and Biomolecular Engineering, Hong Kong University of Science and Technology, Clearwater Bay, Hong Kong
b
Department of Mechanical Engineering, K.N. Toosi University of Science and Technology, 15 Pardis St, Mollasadra Ave, Tehran, Iran
c
Division of Sustainable Development, College of Science, Engineering and Technology, Hamad bin Khalifa University, Qatar Foundation, Doha, Qatar

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

Article history: One of the most widely cultivated crops in the world is rice, leading to millions of tons of rice husks (also
Received 27 December 2013 known as rice hulls or chaffs). The large amount of this lignocellulosic biowaste has resulted in an
Accepted 22 November 2014 extensive search for its utilization. One such usage of this waste is for the production of electricity, such
Available online
as in combined heat and power or gasification units. However, one of the disadvantages of using rice
husks is their high silica content which produces large amounts of undesirable ash upon combustion
Keywords:
leading to operation problems such as slagging and clogging. Here, alkali pretreatment for the extraction
Rice husk/hull/chaff
of silica in the form of sodium silicate has been studied using response surface methodology (RSM) and
Response surface method (RSM)
BoxeBehnken design (BBD)
Analysis of Variance (ANOVA). Three independent variables namely reaction temperature, duration, and
Sodium silicate leaching alkali concentration were considered using a BoxeBehnken design (BBD). The operating conditions were
Alkali extraction of ash optimized under different scenarios. The first optimization focused on the two goals of high ash removal
and high solid yield while the next optimization rounds added the criteria of low NaOH usage and robust
design (using propagation of error (POE)). The final treated rice husks can therefore be more suitably
used as feed for thermal and/or electric units. The developed empirical predictive models were suc-
cessfully validated through additional experimentation.
© 2014 Elsevier Ltd. All rights reserved.

1. Introduction
With the increase in public awareness regarding sustainable
development, the utilization of renewable resources is on the rise.
In the wake of problems associated with fossil fuels, biomass is
commonly recognized as one the most important alternatives [1].
One of the most widely cultivated plants in the world is rice. China
is the world leader in rice production followed by India and
Indonesia. At just under 700 million tons per year, rice is currently
the third most widely produced crop in the world after sugar cane
and maize [2]. The grain is covered and protected by an outer layer
known as the husk or hull. The husk is not edible, is removed in the
first stage of the milling process, and composes approximately
20e22% of the total weight of the milled paddy [3,4]. Rice husks are
particularly known among agricultural biowastes for their low
percentage of protein and available carbohydrates, and high per-
centage of ash containing mostly silica. The exact composition of
* Corresponding author. Department of Chemical and Biomolecular Engineering,
Hong Kong University of Science and Technology, Clearwater Bay, Hong Kong.
Tel.: þ852 23588412.
E-mail addresses: bazargan@ust.hk (A. Bazargan), bazargan@kntu.ac.ir
(M. Bazargan), kemckayg@ust.hk (G. McKay).
rice husk which influences its proximate and ultimate analysis
depends on many factors such as the weather and cultivation
conditions [5e8]. The silica content in the dry season (mean 18%)
has been reported to be higher than in the wet season (mean 15%)
[3].
Various applications have been proposed for rice husk utiliza-
tion [9e11] from which silica production [12e16] and energy
generation [17e23] have been among the most scrutinized. How-
ever, one of the main drawbacks to energy (electricity) generation
from rice husks is their high ash content [24]. Aside from the fact
that the higher heating value of biomass is inversely proportional to
its ash content [25], the silica can form undesired deposits and slags
during combustion, which can lead to operational problems [26].
The melting of the rice husk ashes [27] has been known to cause
agglomeration, fouling, and corrosion of heat transfer surfaces
[28,29].
The removal of silica from rice husks by using sodium hydroxide
has been documented from as early as nearly a century ago [30].
Recent years have also seen numerous sporadic studies regarding
treatment of husks with sodium hydroxide [1,31e40]. Nonetheless,
a parametric study using Response Surface Methodology (RSM) has
never been provided on the alkali leaching of silica from raw rice
husk. For example Singh et al. have previously used RSM to evaluate

http://dx.doi.org/10.1016/j.renene.2014.11.072
0960-1481/© 2014 Elsevier Ltd. All rights reserved.
 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520 513

the saccharification of rice straw and hull by microwaveealkali where Wi and Wf are the weight of the rice husks before and after
pretreatment, but have unfortunately not considered the silica the treatment, and Ashi and Ashf are the initial and final ash content
content as a response [36]. (weight %) of the husk respectively.
RSM is a statistical technique for modeling and analysis of the
effects of designated independent variables on selected response 2.3. Design of experiment
variables. Various RSM designs help obtain statistically valid re-
sults with running limited experimental trials. The powerful and A 3-factor, 3-level BoxeBehnken design (BBD) was used for the
popular tool of RSM has been applied widely across fields [41e43]. Design of Experiments (DOE). The three parameters to evaluate
In this manuscript, RSM is used for the first time to optimize were selected as temperature (X1), treatment duration (X2), and
operating conditions which allow for maximum silica removal NaOH concentration (X3). The NaOH concentration is defined as the
while retaining most organics in the solid phase. If this target is weight ratio of NaOH in the solution to dry rice husk. For example, a
achieved, the rice husks will become more suitable for heat and 25% concentration represents a 1:4 ratio (i.e. the weight of NaOH
energy generation purposes. The extracted silicate can in turn be used is solution is one fourth of the dry weight of the rice husk). Ash
used for the production of silica products with further treatment removal percentage and solid yield were used as response vari-
[44e46] such as in the production of zeolites through hydro- ables. The low, center, and high levels of each factor were denoted
thermal reactions. Elsewhere, Bazargan et al. have used RSM to as 1, 0 and þ1. The corresponding values for these factors are
investigate the effect of alkali treatment on the moisture content shown in Table 1.
and drying kinetics of rice husks which is another important factor A total of 17 experiments were planned by the Design Expert v.7
effecting their usage in energy production [47]. In the literature, statistical software in randomized order. A polynomial equation
the word “ash” is usually used to refer to solid residues of com- was used to model the mathematical relationship between the
bustion. Although no combustion is employed herein, the word variables and responses as follows:
ash is used to refer to the mineral content of the husk removed
with alkali treatment. X
N X
N X
N 1 X
N
Y ¼ b0 þ bi Xi þ bii Xi2 þ bij Xi Xj (2)
i¼1 i¼1 i¼1 j¼iþ1
2. Materials and methods
where Y is the predicted response and N is the number of variables,
2.1. Alkali extraction of silica fixed at 3 in this study. Xi and Xj are the independent variables, and
b0, bi, bii and bij are the intercept term, the linear effects, the squared
The rice husks were harvested in China and donated by Peako effects and the interaction effects, respectively.
Biomass Energy Ltd. The husks were kept in an oven at 90  C and
used as-is without washing or any other pretreatment. In each 3. Results and discussion
experimental run, the husks were treated with a sodium hydroxide
solution at the desired temperature under reflux. The solid loading In this study, three factors with three levels were used in a BBD
was fixed at 10 g/L of solution. A magnetic stirrer was used in every to evaluate alkali leaching of silica from rice husks. The results of
experiment to ensure adequate stirring. After the reaction duration the BBD experiments are displayed in Table 2. A “logit” trans-
was completed, the solids were washed with 4 L of deionized water formation, defined as follows, is applied to the ash removal
per gram of husk to ensure all soluble fractions were removed. The response (y):
as-prepared husk was dried and collected.


y  yl
ytrans ¼ ln (3)
yu  y
2.2. Ash removal measurements
where ytrans is the response after the transformation has been
In the literature, the major constituent of rice husk ash is shown
applied, yl is the lower limit of the response (here set at 0%) and yu
to be silica (SiO2), although other element oxides such as Al, K, P, Ca,
is the upper limit of the response (here set at 100%). The logit
Mg, Na, and Fe are also present is small amounts [28,48]. Since the
transformation is best suited for responses bound between lower
overwhelming majority of the ash is composed of silica, it is safe to
and upper limits.
use the ash content as a good estimate of silica content. The
composition of the rice husk as measured by X-ray fluorescence
(XRF) spectroscopy is provided in the Supplementary Material. The 3.1. Fit summary
XRF data shows that the majority of the ash is composed of silicon
(84.1% wt), followed by potassium (9.4% wt), and calcium (3.2% wt). The fit summary, including the sequential model sum of squares,
Since potassium oxides, nitrate, carbonate, sulfate, and silicate are lack of fit tests, and model summary statistics for ash removal and
all soluble in water, the aqueous sodium hydroxide solution will be solid yield are presented in Tables 3 and 4 respectively. By
capable of removing at least 93.5% of the minerals. Predicting the considering the Sequential Model Sum of Squares, the highest order
solubility of the various other mineral components which may exist polynomial where additional terms are significant and the model is
is less straightforward.
The ash content is measured using thermogravimetric analysis
Table 1
(TGA). The samples are heated to 650  C for a prolonged period Experimental levels of the independent variables.
under air atmosphere until constant weight is reached. The ash
Factor Name Units Low (1) Center (0) High (þ1)
removal efficiency is defined as:
Actual Value Actual Value Actual Value

 
Wf  Ashf X1 Temperature C 25 57.5 90
Ash removal % ¼ 1  100 (1) X2 Time h 4 14 24
Wi  Ashi X3 NaOH Concentration % 1 13 25
(wt alkali/wt husk)
514 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520

Table 2 Table 4
BBD and experimental results. Multi regression analysis of response 2: solid yield %.

Run X1 X2 X3 (NaOH Response 1 Response 2 Sequential model sum of squares

(temperature, (time, concentration, %) (ash removal, %) (solid yield,
 Source Sum of Degrees of Mean F-value p-Value Comments
C) hr) %)
squares freedom square
1 57.5 4 1 34 88
Mean 102,028.76 1 102028.76
2 25 14 25 38 83
Linear 1074.75 3 358.25 32.92 <0.0001
3 90 14 25 99 57
2FI 103.25 3 34.42 9.00 0.0034
4 90 14 1 36 81
Quadratic 34.29 3 11.43 20.25 0.0008 Chosen
5 57.5 14 13 91 76
Cubic 2.75 3 0.92 3.06 0.1544 Aliased
6 90 24 13 99 68
Residual 1.20 4 0.30
7 57.5 14 13 96 76
Total 103,245.00 17 6073.24
8 25 4 13 14 90
9 57.5 14 13 93 75 Lack of fit tests
10 90 4 13 96 72
Source Sum of Degrees of Mean F-value p-Value Comments
11 25 24 13 47 81
squares freedom square
12 57.5 24 1 43 87
13 57.5 4 25 92 76 Linear 140.285 9 15.587 51.958 0.0009
14 57.5 14 13 94 76 2FI 37.035 6 6.173 20.575 0.0057
15 57.5 14 13 95 75 Quadratic 2.750 3 0.917 3.056 0.1544 Chosen
16 25 14 1 18 89 Cubic e e e e e Aliased
17 57.5 24 25 95 67 Pure Error 1.200 4.000 0.300

Model summary statistics

not aliased is chosen. Here, a quadratic model is best for both re- Source Std. R- Adj R- Pred. R- PRESS Comments
sponses. One must keep in mind that the values listed in Tables 3 dev. squared Squared squared
and 4 are initial fit summaries before any interaction term is Linear 3.299 0.884 0.857 0.771 278.136
removed and the model is tweaked. 2FI 1.955 0.969 0.950 0.900 121.497
Quadratic 0.751 0.997 0.993 0.962 45.875 Chosen
Cubic 0.548 0.999 0.996 e e Aliased
3.2. Analysis of variance and model fitting

The statistical significance of the models' coefficients was eval-

High F-values and low p-values are sought. When the p-value for
uated using ANOVA tables. Interestingly, as can be seen from Table 5,
a particular coefficient is smaller than 0.05, it is deemed significant.
the response for ash removal showed no significant dependence on
The non-significant coefficients will be removed from the final
non-linear terms containing reaction time. Table 6 shows ANOVA
model. High p-values for lack of fit imply that the lack of fit is not
results for the ash removal response after the non-linear terms
significant relative to the pure error. A high R2 indicates that the data
regarding reaction duration are eliminated. Table 7 shows data
satisfactorily fits the model. However, adding variables to the model
regarding the solid yield response before any terms are removed. In
(irrespective of whether or not they are statistically significant) will
the interest of space, data regarding the ANOVA table for the solid
always increase R2; therefore its high value is not always an accurate
yield after elimination of insignificant terms is not shown.
indication of a favorable regression model. Therefore high values
obtained for Adjusted R2 and Predicted R2 are used to confirm the
Table 3
favorability of the model. Values for CV% are a measure expressing
Multi regression analysis of response 1: ash removal%.
the standard deviation as a percentage of the mean. The smaller the
Sequential model sum of squares CV values are the better the reproducibility of the results will be.
Source Sum of Degrees of Mean F-value p-Value Comments Generally CV values lower than 10 are desired. In addition, low
squares freedom square PRESS values suggest the adequacy of the models for predictive
Mean 39.97 1.00 39.97 applications. The signal-to-noise ratio is assessed by using the
Linear 52.56 3.00 17.52 11.67 0.0005 adequate precision; a ratio greater than 4 is preferred [49].
2FI 4.32 3.00 1.44 0.95 0.4545 Empirical relationships between the response and independent
Quadratic 12.44 3.00 4.15 10.53 0.0055 Suggested
variables were found. The final models obtained in terms of actual
Cubic 2.32 3.00 0.77 7.06 0.0447 Aliased
Residual 0.44 4.00 0.11 factors are given below:
Total 112.05 17.00 6.59

Lack of fit tests
Logit (Ash Removal%) ¼ Ln[(Ash Removal%)/(100  Ash removal
%)] ¼ 5.86149 þ 0.13152*Temperature þ 0.04995*Time þ 0.21581
Source Sum of Degrees of Mean F-value p-Value Comments
*NaOH þ 0.00266*Temperature*NaOH  0.00092*Temperature2 
squares freedom square
0.00926*NaOH2 (4)
Linear 19.075 9.000 2.119 19.355 0.0059
2FI 14.759 6.000 2.460 22.464 0.0048
Quadratic 2.319 3.000 0.773 7.058 0.0447 Suggested Solid Yield% ¼ 100.08644  0.15385*Temperature 
Cubic 0.000 0.000 e e e Aliased 0.87409*Time  0.079265*NaOH þ 0.00384*Temperature*Time 
Pure Error 0.438 4.000 0.110 0.011538*Temperature*NaOH  0.016667*Time*NaOH þ
Model summary statistics 0.020789*Time2 þ 0.012701*NaOH2 (5)
Source Std. R-squared Adj Pred. PRESS Comments
dev. R-squared R-Squared

Linear 1.225 0.729 0.667 0.539 33.229 3.3. Diagnostics and adequacy checking
2FI 1.233 0.789 0.663 0.356 46.418
Quadratic 0.628 0.962 0.913 0.476 37.783 Suggested It is imperative to check the adequacy of the fitted models before
Cubic 0.331 0.994 0.976 e e Aliased
proceeding to optimization in order to avoid misleading results. As
 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520 515

Table 5
ANOVA for response 1 (ash removal) before removal of insignificant terms.

Source Sum of Degrees of Mean square F-value p-Value Comment

squares freedom

Model 69.320 9 7.702 19.558 0.0004 Significant

X1-Temperature 30.946 1 30.946 78.581 <0.0001
X2-Time 1.996 1 1.996 5.068 0.0591
X3-NaOH 19.623 1 19.623 49.828 0.0002
concentration
X1X2 0.019 1 0.019 0.049 0.8310 Not significant
X1X3 4.293 1 4.293 10.900 0.0131
X2X3 0.004 1 0.004 0.009 0.9261 Not significant
X21 3.810 1 3.810 9.674 0.0171
X22 0.499 1 0.499 1.267 0.2974 Not significant
X23 7.118 1 7.118 18.075 0.0038
Residual 2.757 7 0.394
Lack of fit 2.319 3 0.773 7.058 0.0447 Significant

Std. dev. PRESS C.V. % Adeq precision R-squared Adj R-squared Pred R-squared
0.63 37.78 40.93 14.681 0.962 0.913 0.476

Table 6
ANOVA for response 1 (ash removal) after removal of insignificant terms.

Source Sum of squares Degrees of freedom Mean square F-value p-Value Comment

Model 68.798 6 11.466 34.973 <0.0001 Significant

X1-Temperature 30.946 1 30.946 94.387 <0.0001
X2-Time 1.996 1 1.996 6.087 0.0333
X3-NaOH concentration 19.623 1 19.623 59.851 <0.0001
X1X3 4.293 1 4.293 13.093 0.0047
X21 3.967 1 3.967 12.100 0.0059
X23 7.338 1 7.338 22.382 0.0008
Residual 3.279 10 0.328
Lack of Fit 2.841 6 0.473 4.324 0.0890 not Significant

Std. dev. PRESS C.V. % Adeq precision R-squared Adj R-squared Pred R-squared
0.52 16.08 37.34 19.231 0.955 0.927 0.777

Table 7
ANOVA for response 2 (solid yield) before removal of insignificant terms.

Source Sum of Degrees of Mean square F-value p-Value Comment

squares freedom

Model 1212.285 9 134.698 238.706 <0.0001 Significant

X1-Temperature 528.125 1 528.125 935.918 <0.0001
X2-Time 66.125 1 66.125 117.184 <0.0001
X3-NaOH 480.500 1 480.500 851.519 <0.0001
concentration
X1X2 6.250 1 6.250 11.076 0.0126
X1X3 81.000 1 81.000 143.544 <0.0001
X2X3 16.000 1 16.000 28.354 0.0011
X21 0.024 1 0.024 0.042 0.8435 Not significant
X22 18.129 1 18.129 32.127 0.0008
X23 14.024 1 14.024 24.852 0.0016
Residual 3.950 7 0.564
Lack of fit 2.750 3 0.917 3.056 0.1544 Not significant

Std. dev. PRESS C.V. % Adeq precision R-squared Adj R-squared Pred R-squared
0.751 45.875 0.970 57.061 0.997 0.993 0.962

shown in Fig. 1, the normal probability plots of the internally stu-
dentized residuals are linear. In calculating the internally studen-
tized residuals, a residual is compared to the residual variance
(entire set including itself). With this simple graphical method, the
normal distribution of the observed data is confirmed. If the data in
Fig. 1 were not linear, it would indicate that the error terms are not
normally distributed, and suggest that the model's underlying as-
sumptions may have been violated. Fig. 2 satisfactorily shows that
the residuals are randomly scattered when plotted against the
predicted response. Furthermore, Fig. 3 shows that the
experimental data is in fair agreement with the values predicted by
the developed models (with the solid yield response showing a
more accurate predictive capability). Hence the adequacies of the
developed models are confirmed.
3.4. Response surfaces
Fig. 4 depicts the response surfaces showing the effects of the
independent variables on the ash removal and solid yield.
Regarding the solid yield, it is evident that as the temperature,
516 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520

Fig. 1. Normal probability plot for responses (a) ash removal and (b) solid yield.

reaction duration, and alkali concentration increase, there is a At higher temperatures, the NaOH concentration has a drastic effect
larger loss of solids. Although all three factors show significant ef- on the solid yield. This effect is less pronounced at lower temper-
fects, the reaction temperature and alkali concentration appear to atures. Regarding the ash removal, from among the three variables
be more influential on the solid yield than the treatment duration. considered, the duration is of least influence. The combination of

Fig. 2. Plot of internally studentized residuals vs. predicted response for (a) ash removal and (b) solid yield.

Fig. 3. Diagnostic plots of the model precision for (a) ash removal and (b) solid yield.
 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520
Fig. 4. Response surfaces for (a) ash removal and (bed) solid yield. For all surfaces, the input variables not shown are fixed at the center level.
high temperature and high alkali concentration leads to highest ash
removal. It is also noteworthy that 13% NaOH is high enough and
removes more than 90% of the ash at moderate temperatures
(57.5  C). Any further increase of temperature or concentration
does not significantly increase silica ash removal. The slightly
higher ash content when the NaOH concentration is increased
above 13% could be ascribed to the fact that the sodium which does
not react with the silica to form soluble silicate, can attach to the
fiber in other forms. This, in turn, adds to the final measured ash
amount upon thermal degradation.
Hsieh et al. [1] concluded that at a fixed reaction duration, the
reaction temperature can enhance SiO2 removal slightly while a
change of NaOH concentration has remarkable influence. They
found that a 1:5 ratio of NaOH:husk at 50  C for 3 h was adequate to
remove nearly all the soluble silica while retaining 63% of the
organic matter as solid residue. Further increase of the temperature
or alkali concentration did not greatly enhance the silica removal.
Approximately 14% of the rice husk was soluble when the treat-
ment was done with pure water. Johar et al. [37] have reported that
almost all the silica was removed with alkali treatment under reflux
for 6 h (3 runs, 2 h each). The alkali treatment was reported to
efficiently remove hemicellulose as well. The smooth surface of the
untreated rice husk becomes rough after alkali treatment. This was
ascribed to the removal the outer non-cellulosic layer composed of
materials such as hemicelluloses, lignin, pectin (protective layer),
wax, and other impurities. The diameter of the husk fiber bundles
were reported to remain constant after alkali treatment indicating
that a large fraction of the lignin (which acts as a cementing agent)
517
persisted. According to Ndazi et al. [31] higher temperatures, longer
treatment, and higher concentration removes more lignin, but at
lower temperatures the lignin might not degrade. In rice husks,
alkali treatment has previously been shown to break the cell wall
by dissolving hemicellulose, lignin, and silica; by hydrolyzing
uronic and acetic acid esters; and by swelling cellulose [50].
Endwise degradation, also known as endwise depolymerization or
peeling, of cellulose can also occur in alkali media [51]. The
hydrogen bond of cellulose is disrupted as a result of alkali treat-
ment and thereby the surface roughness of the samples increases.
When cellulose is mercerized, the hydrolysis rate increases mark-
edly (70%). This means that the number of accessible glycosidic
linkages in mercerized cellulose is higher than the native one [52].
Hemicelluloses contain a variety of sugar units in addition to
glucose and can act as a physical barrier which surrounds and
protects the cellulose fibers [53]. The nonglucose units of hemi-
cellulose exhibit distinctly different reactivities when compared to
glucose units because of their different ring structures and/or hy-
droxyl configurations. Generally, hemicelluloses are more reactive
than cellulose. Their degradation is similar to that of cellulose but
proceeds much more readily and extensively [52]. If the alkali
concentration is increased above the limits of this paper, the rice
husk will break down to a puffy (mercerized) material which can be
used as a sorbent for liquids such as in oil spills [54,55].
Silica has a melting point above 1400  C. In addition to adhesion
effects of the sintered ash, in combustion, the alkali oxides and salts
can react with silica to form eutectic mixtures. These compounds
such as the one formed through the following reaction have
518 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520
melting points below that of the individual components (below
800  C) [28]:
4SiO2 þ K2CO3 / K2O $ 4SiO2 þ CO2
Hence the more minerals are removed from the rice husks, the
less the operational problems arising from agglomeration, fouling,
slagging, and clogging will be. For a review of possible operation
problems arising from excessive ash content in combustion sys-
tems, refer to the work of Van Caneghem et al. [26].
The silicon in rice husk is thought to be bound to the organic
molecules. Since cellulose and lignin do not possess considerable
bonding ability with silicon, hemicellulose has been speculated to
be responsible for this binding. This can be attributed to the polar
electrometric effect in the aldehyde groups [48,56]:
The nature of the silicon bonding is believed to be complex. It
has been speculated that it is bound to four carbohydrates as shown
below [48,56]:
When these bonds are broken, the SieO groups become
attached to each other and form low temperature silica. The
removal of silica by alkali treatment is thought to be the result of
the following reaction:
SiO2 þ 2NaOH / Na2SiO3 þ H2O
The leached silica in the liquid can then be used for the pro-
duction of porous silica-based materials.
3.5. Optimization
Optimization has been defined as improving a system's perfor-
mance in order to obtain the maximum benefit from it [57]. When a
set of responses are influenced by several variables, the system is
optimized with the objective to simultaneously adjust the levels of
the input variables in order to attain the most desirable overall
response.
In the previous section we have already established the effects
of temperature, duration, and alkali ratio on the solid yield and
silica removal. First, the system will be optimized with the
following criteria:
1. To minimize the loss of solid residue due to the treatment
2. To maximize the removal of silica ash from the solid residue
The following two criteria will be added for a second round of
optimization:
3. To minimize the use of alkali (for cost-effectiveness of process)
4. To have the most robust design (i.e. minimum propagation of
error for both ash removal and solid yield)
Errors in input variables such as temperature, duration, and
concentration can transmit to the response (silica removal and solid
yield). In examining the Propagation of Error (POE), conditions are
sought which minimize the transmitted variation of inputs thus
“creating a process that's robust to the factor settings” [58]. The
lower POE values are, “the less of the error in control factors will be
transmitted to the selected response which results in a more robust
process” [58]. If the response surface is linear, POE plots become
valueless. This is because under linear conditions errors will be
transmitted equally throughout the region. Here, in POE analysis, a
standard deviation of 0.5 is chosen for each input variable.
Table 8
Operation parameters under different optimization scenarios.
Proposed values for input variables Predicted responses
under optimized
conditions
Optimization Temperature Time Concentration Ash Solid

criteria ( C) (hr) (%) removal (%) yield (%)
1 and 2 only 59.5 4 15.7 92.6 78.7
(equal weight)
1, 2, 3, and 4 44.3 22.5 9.8 84.4 79.7
(equal weight)
1 and 2 (more 48.4 24 10.2 89.2 78.8
weight), 3 and
4 (less weight)
With the inclusion of additional optimization criteria, or
designating different weights to imposed criteria, various opti-
mized operation conditions emerge. A trade-off occurs during
optimization as all criteria cannot be met 100%. The optimization
process implies trying to meet the criteria (even when conflicting)
as best as possible. As a case in point, adding criteria 3 to the
optimization process leads to a decrease in silica removal which is
unwanted. This could be expected because designating ‘using less
alkali’ as an optimization target logically leads to less soluble so-
dium silicate formation. Ultimately the optimized operating con-
ditions and the weight of each criterion depend on the choice of the
end-user. For example, under conditions in which access to NaOH is
difficult and costly, a higher silica content in the final residue may
be tolerated. Meanwhile, if cheap alkali is accessible and/or the CHP
unit to be used is highly sensitive to ash content, the importance of
criterion 2 could be increased while the importance of criterion 3 is
diminished. For a more clear understanding, the results of various
optimization scenarios have been tabulated in Table 8. Note that
numerous solutions are provided during optimization, and Table 8
only shows the ones with the highest desirability. The results from
Table 8 show that with optimum operating conditions, nearly 80%
of the solid could be kept while approximately 90% of the ash is
removed. Such conditions could be very advantageous for more
favorable utilization of rice husks as solid fuel. Note that the
amount of silica in the raw husk sample used was approximately
12% dry weight. Hence if the husks used have a higher initial silica
content, the results would be expected to shift.
3.6. Model validation
In order to verify whether the proposed models can accurately
estimate the behavior of the system additional experiments are
performed and compared to the estimated values from the model.
These points are chosen within the design area and not super-
imposing any previous experimental condition used for developing
the model. The values for 95% PI low and 95% PI high (not shown
here) designate the range in which the actual response should be. If
the actual experimental result for any point is out of the corre-
sponding range, the model is negated. Here, all validation experi-
ments led to response values within the appropriate range and
hence validated the model. Table 9 shows two experimental runs
used for validation.
 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520
Table 9
Validation experiments for the proposed models.
Validation Validation
experiment 1 experiment 2
Temperature ( C) 70 58
Duration (hr) 14 16
Alkali concentration (%) 14 23
Actual ash removal (%) 99 92
Predicted ash removal (%) 96.7 95.8
Actual solid yield (%) 71 71
Predicted solid removal (%) 71.7 69.4
4. Conclusion
Satisfactory utilization of rice husks for electricity or heat gen-
eration is often hampered due to the high ash content of these
wastes. Here, a simple alkali treatment is proposed for the removal
of the silica ash, so that problems such as clogging, fouling, and
slagging could be avoided during energy generation processes.
Response Surface Methodology (RSM) with a BoxeBehnken design
(BBD) has been used to consider the three variables of reaction
temperature, duration, and alkali concentration. The results
showed that with optimum operating conditions, nearly 80% of the
rice husk can be retained in the solid phase while approximately
90% of the ash is removed. Such pretreatment could therefore be
very suitable for rendering the rice husks more favorable for energy
production, while retaining a large portion of the organics in the
solid phase. Future work may focus on a more detailed under-
standing of the ash removal mechanism from the rice husks. This
could include mass and energy balances for the extraction of
minerals (in particular silica), kinetics analyses, and chemical
equilibria.
Acknowledgments
The authors would like to sincerely thank the staff at the MCPF
and AEMF facilities at the Hong Kong University and Technology, as
well as Mr. Ronnie Lo of Peako Biomass Energy Co. for their
assistance.
Appendix A. Supplementary data
Supplementary data related to this article can be found at http://
dx.doi.org/10.1016/j.renene.2014.11.072.
References
[1] Hsieh Y, Du Y, Jin F, Zhou Z, Enomoto H. Alkaline pre-treatment of rice hulls
for hydrothermal production of acetic acid. Chem Eng Res Des 2009;87:13e8.
[2] FAOSTAT. Food and Agriculture Organization of the United Nations Statistics.
2013.
[3] Juliano BO. Rice hull and rice straw. In: Juliano BO, editor. Rice: chemistry and
technology. 2nd ed. American Association of Cereal Chemists; 1985.
p. 689e755.
[4] Adam F, Appaturi JN, Iqbal A. The utilization of rice husk silica as a catalyst:
review and recent progress. Catal Today 2012;190:2e14.
[5] Wang W, Martin J, Zhang N, Ma C, Han A, Sun L. Harvesting silica nano-
particles from rice husks. J Nanopart Res 2011;13:6981e90.
[6] Virtanen T, Svedstro€m K, Andersson S, Tervala L, Torkkeli M, Knaapila M, et al.
A physico-chemical characterisation of new raw materials for microcrystalline
cellulose manufacturing. Cellulose 2012;19:219e35.
[7] Nazemi L, Nazmara S, Eshraghyan MR, Nasseri S, Djafarian K, Yunesian M,
et al. Selenium status in soil, water and essential crops of Iran. Iran J Environ
Health Sci Eng 2012;9.
[8] Mehdinia SM, Latif PA, Taghipour H. A comparative evaluation of dried acti-
vated sludge and mixed dried activated sludge with rice husk silica to remove
hydrogen sulfide. Iran J Environ Health Sci Eng 2013;10.
[9] Chowdhury S, Das Saha P. Adsorption of malachite green from aqueous so-
lution by NAOH-modified rice husk: fixed-bed column studies. Environ Prog
Sustain Energy 2013;32:633e9.
519
[10] Mehdinia SM, Latif PA, Taghipour H. Removal of hydrogen sulfide by physico-
biological filter using mixed Rice husk silica and dried activated sludge. CLEAN
Soil Air Water 2013;41:949e54.
[11] Kordatos K, Ntziouni A, Iliadis L, Kasselouri-Rigopoulou V. Utilization of
amorphous rice husk ash for the synthesis of ZSM-5 zeolite under low tem-
perature. J Mat Cycl Waste Manage 2013;15:571e80.
[12] Xing A, Tian S, Tang H, Losic D, Bao Z. Mesoporous silicon engineered by the
reduction of biosilica from rice husk as a high-performance anode for lithium-
ion batteries. RSC Adv 2013;3:10145e9.
[13] Chen H, Wang W, Martin JC, Oliphant AJ, Doerr PA, Xu JF, et al. Extraction of
lignocellulose and synthesis of porous silica nanoparticles from rice husks: a
comprehensive utilization of rice husk biomass. ACS Sustain Chem Eng
2012;1:254e9.
[14] Ugheoke B, Mamat O, Ari-Wahjoedi B. Thermal expansion behavior, phase
transitions and some physico-mechanical characteristics of fired doped rice
husk silica refractory. J Adv Ceram 2013;2:79e86.
[15] Alyosef HA, Eilert A, Welscher J, Ibrahim SS, Denecke R, Schwieger W, et al.
Characterization of biogenic silica generated by thermo chemical treatment of
rice husk. Part Sci Technol 2013;31:524e32.
[16] Gu S, Zhou J, Luo Z, Wang Q, Ni M. A detailed study of the effects of pyrolysis
temperature and feedstock particle size on the preparation of nanosilica from
rice husk. Ind Crops Prod 2013;50:540e9.
[17] Gupta R, Pandit A, Nirjar A, Gupta P. Husk power systems: bringing light to
rural India and tapping fortune at the bottom of the pyramid. Asian J Manag
Cases 2013;10:129e43.
[18] Mustonen S, Raiko R, Luukkanen J. Bioenergy consumption and biogas po-
tential in cambodian households. Sustain Switz 2013;5:1875e92.
[19] Yoon SJ, Son Y-I, Kim Y-K, Lee J-G. Gasification and power generation char-
acteristics of rice husk and rice husk pellet using a downdraft fixed-bed
gasifier. Renew Energy 2012;42:163e7.
[20] Shackley S, Carter S, Knowles T, Middelink E, Haefele S, Haszeldine S. Sus-
tainable gasificationebiochar systems? A case-study of rice-husk gasification
in Cambodia, part II: field trial results, carbon abatement, economic assess-
ment and conclusions. Energy Policy 2012;41:618e23.
[21] Shackley S, Carter S, Knowles T, Middelink E, Haefele S, Sohi S, et al. Sus-
tainable gasificationebiochar systems? A case-study of rice-husk gasification
in Cambodia, part I: context, chemical properties, environmental and health
and safety issues. Energy Policy 2012;42:49e58.
[22] Abu Bakar MS, Titiloye JO. Catalytic pyrolysis of rice husk for bio-oil pro-
duction. J Anal Appl Pyrolysis 2013;103:362e8.
[23] Qi P, Lin N, Liu Y, Zhao J. Improvement of oil/water selectivity by stearic acid
modified expanded perlite for oil spill cleanup. J Shanghai Jiaot Univ Sci
2013;18:500e7.
[24] Karmakar MK, Mandal J, Haldar S, Chatterjee PK. Investigation of fuel gas
generation in a pilot scale fluidized bed autothermal gasifier using rice husk.
Fuel 2013;111:584e91.
[25] Lim JS, Abdul Manan Z, Wan Alwi SR, Hashim H. A review on utilisation of
biomass from rice industry as a source of renewable energy. Renew Sustain
Energy Rev 2012;16:3084e94.
[26] Van Caneghem J, Brems A, Lievens P, Block C, Billen P, Vermeulen I, et al.
Fluidized bed waste incinerators: design, operational and environmental is-
sues. Prog Energy Combust Sci 2012;38:551e82.
[27] Virmond E, De Sena RF, Albrecht W, Althoff CA, Moreira RFPM, Jose  HJ.
Characterisation of agroindustrial solid residues as biofuels and potential
application in thermochemical processes. Waste Manag 2012;32:1952e61.
[28] Armesto L, Bahillo A, Veijonen K, Cabanillas A, Otero J. Combustion behaviour
of rice husk in a bubbling fluidised bed. Biomass Bioenergy 2002;23:171e9.
[29] Bakker RR, Jenkins BM, Williams RB. Fluidized bed combustion of leached rice
straw. Energy Fuels 2002;16:356e65.
[30] Archibald JG. The effect of sodium hydroxid on the composition, digestibility,
and feeding value of grain hulls and their fibrous material. J Agric Res
1924;27:245e65.
[31] Ndazi BS, Nyahumwa C, Tesha J. Chemical and thermal stability of rice husks
against alkali treatment. BioResources 2008;3:1267e77.
[32] Ndazi BS, Karlsson S, Tesha JV, Nyahumwa CW. Chemical and physical mod-
ifications of rice husks for use as composite panels. Compos Part A Appl Sci
Manuf 2007;38:925e35.
[33] Favaro SL, Lopes MS, Vieira de Carvalho Neto AG, Roge rio de Santana R,
Radovanovic E. Chemical, morphological, and mechanical analysis of rice
husk/post-consumer polyethylene composites. Compos Part A Appl Sci Manuf
2010;41:154e60.
[34] Chakraborty S, Chowdhury S, Das Saha P. Adsorption of Crystal Violet from
aqueous solution onto NaOH-modified rice husk. Carbohydr Polym 2011;86:
1533e41.
[35] Salgado-Delgado R, Vazquez SR, Alvarez-Castillo A, Licea-Claveríe A,
Castan~ o VM. Chemical modification of rice husk surface through alkaline
treatments. Mater Technol 2005;20:26e9.
[36] Singh A, Tuteja S, Singh N, Bishnoi NR. Enhanced saccharification of rice straw
and hull by microwave-alkali pretreatment and lignocellulolytic enzyme
production. Bioresour Technol 2011;102:1773e82.
[37] Johar N, Ahmad I, Dufresne A. Extraction, preparation and characterization of
cellulose fibres and nanocrystals from rice husk. Ind Crops Prod 2012;37:
93e9.
[38] Kuan C-Y, Yuen K-H, Liong M-T. Physical, chemical and physicochemical
characterization of rice husk. Br Food J 2012;114:853e67.
520 A. Bazargan et al. / Renewable Energy 77 (2015) 512e520
[39] Huntanuwatr N, Hunds FC, Davis C. An evaluation of methods for improving
the in vitro digestibility of rice hulls. J Anim Sci 1974;38:140e8.
[40] McManus WR, Choung CC. Studies on forage cell walls: 2. Conditions for alkali
treatment of rice straw and rice hulls. J Agric Sci 1976;86:453e70.
[41] Taheri M, Alavi Moghaddam MR, Arami M. Optimization of Acid Black 172
decolorization by electrocoagulation using response surface methodology.
Iran J Environ Health Sci Eng 2012;9.
[42] Usharani K, Muthukumar M. Optimization of aqueous methylparathion
biodegradation by Fusarium sp in batch scale process using response surface
methodology. Int J Environ Sci Technol 2013;10:591e606.
[43] Jadhav SB, Surwase SN, Phugare SS, Jadhav JP. Response surface methodology
mediated optimization of Remazol Orange decolorization in plain distilled
water by Pseudomonas aeruginosa BCH. Int J Environ Sci Technol 2013;10:
181e90.
[44] Mohamed RM, Mkhalid IA, Abdel Salam M, Barakat MA. Zeolite Y from rice
husk ash encapsulated with Ag-TiO2: characterization and applications for
photocatalytic degradation catalysts. Desalin Water Treat 2013;51:7562e9.
[45] Bohra S, Dey K, Kundu D, Naskar M. Synthesis of zeolite T powders by direct
dissolution of rice husk ash: an agro-waste material. J Mater Sci 2013;48:
7893e901.
[46] Braga R, Barros JF, Melo DA, Melo MF, Aquino FM, Freitas JO, et al. Kinetic
study of template removal of MCM-41 derived from rice husk ash. J Therm
Anal Calorim 2013;111:1013e8.
[47] Bazargan A, Gebreegziabher T, Hui CW, McKay G. The effect of alkali treatment
on rice husk moisture content and drying kinetics. Biomass and Bioenergy
2014;70:468e75. http://dx.doi.org/10.1016/j.biombioe.2014.08.018.
[48] Patel M, Karera A, Prasanna P. Effect of thermal and chemical treatments on
carbon and silica contents in rice husk. J Mater Sci 1987;22:2457e64.
[49] Samavati V, Adeli M. Isolation and characterization of hydrophobic com-
pounds from carbohydrate matrix of Pistacia atlantica. Carbohydr Polym
2014;101:890e6.
[50] Jackson MG. Review article: the alkali treatment of straws. Anim Feed Sci
Technol 1977;2:105e30.
[51] Li X, Tabil LG, Panigrahi S. Chemical treatments of natural fiber for use in
natural fiber-reinforced composites: a review. J Polym Environ 2007;15:
25e33.
[52] Lai Y-Z. Reactivity and accessibility of cellulose, hemicelluloses, and lignins.
In: Hon DN-S, editor. Chemical modification of lignocellulosic materials. New
York: Marcel Dekker; 1996. p. 35e95.
[53] Taherzadeh MJ, Karimi K. Pretreatment of lignocellulosic wastes to improve
ethanol and biogas production: a review. Int J Mol Sci 2008;9:1621e51.
[54] Bazargan A, Tan J, Hui C, McKay G. Utilization of rice husks for the production
of oil sorbent materials. Cellulose 2014;21:1679e88.
[55] Bazargan A, Hui C, Mckay G. Marine residual fuel sorption and desorption
kinetics by alkali treated rice husks. Cellulose 2014;21:1997e2006.
[56] Vlaev L, Petkov P, Dimitrov A, Genieva S. Cleanup of water polluted with crude
oil or diesel fuel using rice husks ash. J Taiwan Inst Chem Eng 2011;42:
957e64.
[57] Bezerra MA, Santelli RE, Oliveira EP, Villar LS, Escaleira LA. Response surface
methodology (RSM) as a tool for optimization in analytical chemistry. Talanta
2008;76:965e77.
[58] Design expert 7 user's guide, multifactor RSM tutorial e part 3: adding
propagation of error (POE) to the optimization, design expert 7 user's guide.
2004. p. 1e10.