Paper 1

Full text

Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=tent20
Download by: [Hrudayanath thatoi] Date: 24 November 2015, At: 00:49
Environmental Technology
ISSN: 0959-3330 (Print) 1479-487X (Online) Journal homepage: http://www.tandfonline.com/loi/tent20
Optimization of delignification of two Pennisetum
grass species by NaOH pretreatment using Taguchi
and ANN statistical approach
Sonali Mohaptra, Preeti Krishna Dash, Sudhanshu Shekar Behera &
Hrudayanath Thatoi
To cite this article: Sonali Mohaptra, Preeti Krishna Dash, Sudhanshu Shekar Behera &
Hrudayanath Thatoi (2015): Optimization of delignification of two Pennisetum grass species by
NaOH pretreatment using Taguchi and ANN statistical approach, Environmental Technology,
DOI: 10.1080/09593330.2015.1093034
To link to this article: http://dx.doi.org/10.1080/09593330.2015.1093034
Accepted author version posted online: 19
Nov 2015.
Published online: 20 Nov 2015.
Submit your article to this journal
Article views: 8
View related articles
View Crossmark data

Optimization of delignification of two Pennisetum grass species by NaOH
pretreatment using Taguchi and ANN statistical approach
Sonali Mohaptra
a
, Preeti Krishna Dash
a
, Sudhanshu Shekar Behera
b
and Hrudayanath Thatoi
a
a
Department of Biotechnology, College of Engineering and Technology, Biju Patnaik University of Technology, Bhubaneswar, India;
b
Department of Biotechnology and Medical Engineering, National Institute of Technology, Rourkela, India
ABSTRACT
In the bioconversion of lignocelluloses for bioethanol, pretreatment seems to be the most important
step which improves the elimination of the lignin and hemicelluloses content, exposing cellulose to
further hydrolysis. The present study discusses the application of dynamic statistical techniques like
the Taguchi method and artificial neural network (ANN) in the optimization of pretreatment of
lignocellulosic biomasses such as Hybrid Napier grass (HNG) (Pennisetum purpureum)and
Denanath grass (DG) (Pennisetum pedicellatum), using alkali sodium hydroxide. This study analysed
and determined a parameter combination with a low number of experiments by using the
Taguchi method in which both the substrates can be efficiently pretreated. The optimized
parameters obtained from the L
16
orthogonal array are soaking time (18 and 26 h), temperature
(60°C and 55°C), and alkali concentration (1%) for HNG and DG, respectively. High performance
liquid chromatography analysis of the optimized pretreated grass varieties confirmed the presence
of glucan (47.94% and 46.50%), xylan (9.35% and 7.95%), arabinan (2.15% and 2.2%), and galactan/
mannan (1.44% and 1.52%) for HNG and DG, respectively. Physicochemical characterization studies
of native and alkali-pretreated grasses were carried out by scanning electron microscopy and
Fourier transformation Infrared spectroscopy which revealed some morphological differences
between the native and optimized pretreated samples. Model validation by ANN showed a good
agreement between experimental results and the predicted responses.
ARTICLE HISTORY
Received 18 May 2015
Accepted 3 September 2015
KEYWORDS
Lignocellulose; Pennisetum
purpureum;Pennisetum
pedicellatum; Taguchi
design; ANN
1. Introduction
The economy of the world is largely dependent on
energy sources in which fossil fuel plays an important
role. In recent years, the price of fossil fuels has increased
enormously due to externalities in fossil fuel consump-
tion and inadequate supply.[1] Therefore, attention has
been given to industrial biotechnology, to develop bio-
fuels from non-conventional raw materials/renewable
energy sources such as agricultural biomass. In this
context, the use of ethanol as an alternative of fossil
fuel has been steadily increasing around the world.
Extraction of bioethanol from renewable resources has
stimulated studies on efficient fermentation technology.
For an increased yield of ethanol by fermentation, an
ideal fermentation substrate and suitable process tech-
nology are also required apart from an efficient microbial
strain with high fermenting ability.[2,3] Bioethanol can
be obtained by fermentation of various carbohydrate-
rich raw materials including agricultural products such
as sugar (sugarcane and molasses), starch (cassava,
cereals, and potatoes), and lignocelluloses (rice straw,
corn cob, grasses, and sugarcane waste).[4] Since the
sugary and starchy biomasses are mainly obtained
from food crops and cannot be diverted for non-food
purposes, lignocellulosic materials remain an important
source for bioethanol production for a sustainable and
environmentally friendly system. Among the different
lignocellulosic materials, grasses show potential as a
biofuel crop as they are sufficiently abundant, generate
very low net greenhouse emissions, and range from
extremely persistent to short life cycles.[5]
Besides, grass is the world’s cheapest lignocellulosic
biomass and is available in every region of the world.
Moreover, farmers are cultivating different varieties of
grasses to make them available throughout the seasons.
Being a lignocellulosic perennial crop, grass is a promising
feedstock for producing bioethanol because of its high
yields, low costs, better utilization of low-quality land,
and having no adverse impact on the environment.
Besides, the lignin content of grasses is distinctly low in
comparison to other lignocellulosic materials such as
hardwood and softwood [6] and hence they can be
used as an ideal substrate for bioethanol production.
Generally, the lignocellulosic biomasses are com-
posed of lignin, cellulose, and hemicellulose where
© 2015 Taylor & Francis
CONTACT Hrudayanath Thatoi hn_thatoi@rediffmail.com
ENVIRONMENTAL TECHNOLOGY, 2015
http://dx.doi.org/10.1080/09593330.2015.1093034
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

hemicelluloses bound a matrix of lignin and cellulose. As
the substrate is made up of a composite form, the main
challenge in this case is removal of lignin and increase in
the accessible surface area along with the reduction of
the degree of crystallinity of the cellulose and increase
in the fraction of amorphous cellulose for its use in the
fermentation process. This can be achieved by the
process of pretreatment, the most suitable form for the
hydrolysis step.[7] Pretreatment is the first step for con-
version of biomass into ethanol and is necessary for effi-
cient conversion.[8] Different types of pretreatment
processes are employed to make cellulose more accessi-
ble which include chemical treatments (alkali or acid),
ammonia fibre explosion, biological treatment, and
steam explosion. Among all the pretreatments, alkaline
pretreatment has many advantages. Compared with
acid pretreatment processes, alkaline treatments are
advantageous as many of the caustic salts can be recov-
ered and/or regenerated, they cause less sugar degra-
dation, and remove lignin efficiently.[9] Further, alkaline
pretreatment is in general more effective on agricultural
residues and herbaceous crops in which a significant
amount of hemicellulose is solublized by the process
delignification.[10] Among the different alkaline pre-
treatments using sodium, potassium, calcium, and
ammonium hydroxides, sodium hydroxide (NaOH) is
the most suitable.[11] It has also been reported in
many experiments that the alkaline pretreatment
process removes acetyl and uronic acid groups present
in hemicelluloses along with some residues [12] and
enhances the enzyme accessibility for degradation of
hemicelluloses.[13] Thus, successful pretreatment of cel-
lulosic material is an essential step for subsequent enzy-
matic hydrolysis of the substrate in order to obtain
glucose that is converted to ethanol by microorgan-
isms,[14] which can be achieved through process optim-
ization using statistical methods.
A statistically based experimental design like the
Taguchi orthogonal array (OA) design is employed to
maximize the collection and sorting of variables to be
taken into consideration, and determination and analysis
of the variables at different parameters with the final
effect of variable error. The Taguchi experimental
design is a quick and effective way of optimization, con-
ferring a remarkable outcome in the simultaneous study
of many factors, making its mark on quality products
supplemented with better process performance, render-
ing high yield and better stability.[15] The Taguchi
design in the present work was used to establish the
importance of statistically aligned experiments in specu-
lating the settings of products (and/or processes) on
various parameters.[14,15] Neural networks are used to
model a system. Artificial Neural Networks (ANNs) have
been used successfully for solving biotechnological
complex problems for modelling and optimization. As a
result ANNs are now one of the most popular artificial
learning tools in biotechnology, with applications
ranging from pattern recognition in chromatographic
spectra, expression profiles, to functional analyses of
genomic and proteomic sequences [16] and even
ethanol production by the fermentation process.[12]
ANN is an information-processing paradigm that is
inspired by the biological nervous systems, such as the
brain which processes the information for the entire
body in the living organisms. Indeed, ANN is a massively
interconnected network structure consisting of many
simple processing elements capable of performing paral-
lel computation for data processing. In ANN the funda-
mental processing element (the artificial neurons or
nodes) simulates the basic functions of biological
neurons.[16] An ANN offers the benefit of modelling a
complex system without the underlying mathematical
descriptions as well as the option to model non-linear
systems easily and could save time and resources when
searching for optimum pretreatment conditions.[16]
The present study emphasizes optimization of different
pretreatment parameters (soaking time, concentration,
and temperature) during pretreatment of lignocellulosic
substrates (Pennisetum purpureum and Pennisetum pedicel-
latum) with alkali NaOH by using the Taguchi robust
method for optimum release of celluloses. The optimum
parameters obtained in Taguchi experiments were
further analysed by High performance liquid chromato-
graphy (HPLC) for hexosans and pentosans. Furthermore,
an ANN model was trained and the optimum conditions
were predicted (using genetic algorithm as the global
optimization procedure) by using the Taguchi results.
Besides, the structural changes in untreated (native) and
pretreated lignocellulosic biomasses were also analysed
by FTIR (Fourier transform infrared spectroscopy) and
SEM (Scanning electron microscopy) analyses.
2. Materials and methods
2.1. Substrate collection
Two grass varieties (Hybrid Napier grass (CO-3) [HNG]
and Denanath grass [DG]) were collected from the Direc-
torate of Research for Women in Agriculture (DRWA),
Bhubaneswar, Odisha, and used as the lignocellulosic
substrates for our experiment. The collected substrates
were brought to the microbiology laboratory, Depart-
ment of Biotechnology, College of Engineering and Tech-
nology, Bhubaneswar. The substrates were properly
washed using tap water to remove unwanted debris,
sundried, and milled into 2 mm powder form in a ball
2S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

mill at the Herbarium section of the Institute of Minerals
and Material Technology, Bhubaneswar, Odisha.
2.2. Estimation of moisture content
The oven drying method [17] was used for the estimation
of the moisture content of the two grass varieties. The
sample was weighed in a glass crucible and then dried
in an oven for 12 h at 105°C. The dried sample was
then cooled to room temperature in desiccators and
weighed. This process was repeated to achieve a con-
stant weight to make the substrate free from moisture
content.
Then the moisture content was calculated using the
following formula:
% of moisture content =W1−W2
W

∗100,(1)
where
W= Weight of the initial sample (gm),
W₁= Weight of the sample + container before drying
(gm),
W₂= Weight of the sample + container after drying
(gm).
2.3. Taguchi method
The Taguchi method comprises the formulation of differ-
ent experimental situations through OAs and helps to
minimize the effects of noise factors in the process of
optimization, thus leading to a dynamic or robust exper-
imental design.[18,19] The experimental data from the
Taguchi arrays can be analysed by plotting the data
and performing a visual analysis, the one-way analysis
of variance (ANOVA), and Fisher’s exact test to test the
significance of the factor effect.[20] In the Taguchi
method, the term ‘signal’(S) represents the desired
value for the output characteristic, while ‘noise’(N) rep-
resents an undesirable value for the output character-
istic. The S/N ratio is thus the ratio of the signal to the
noise, and the Taguchi method uses this to measure
how a specific quality characteristic deviates from the
desired value.
The performance parameter, signal-to-noise ratio
(S/N), which is the log transformation of the mean-
square deviation of the desired performance, is the
objective function that should be optimized for the
desired output. The S/N ratio can be defined as
S
N=−10 log(MSD),(2)
where MSD is the mean-square deviation for the output
characteristic.[21]
For defining the S/N ratio to optimize the perform-
ance parameter different criteria can be considered
which are detailed below.
Criterion 1: If the goal is to maximize the performance,
the value of S’/N should be high. Criterion 2: If the goal is
to minimize the performance, the value of S”/N should
be low. Criterion 3: If the goal is to target a predetermined
S/N then the value of S”’/N being nominal is better.
2.4. NaOH pretreatment
The effects of the pretreatment parameters such as reac-
tion temperature, acid concentration, and soaking time
of both the substrates HNG and DG on the sugar yields
were examined (in triplicate) by a factorial design devel-
oped using the Taguchi robust design.[22] Ten grams of
substrates (HNG and DG) were treated with 100 ml alka-
line solution of different concentrations (0.5%, 1%, 1.5%,
and 2%) of NaOH in separate conical flasks. Thereafter,
treated samples were soaked for different time intervals
(2, 10, 18 and 26 h) at different incubation temperatures
(30°C, 45°C, 55°C, and 60°C). After soaking, the samples
were filtered and the solid residues were collected. The
collected solid residues were then washed with tap
water until neutral pH was attained. The filtrate obtained
after soaking the sample was dried at 105°C and stored
at room temperature for further analysis of total and
reducing sugars while the solid fraction was used for
the analysis of cellulose, hemicelluloses, and lignin
content.[23]
2.5. Analysis of the solid fraction
2.5.1. Estimation of hemicellulose
The dried biomass (1 gm) of the collected samples was
taken in 250 ml conical flasks and 10 ml of 3% w/v
H
2
SO
4
was added and then the mixture was autoclaved
at 121°C with 15 lb pressure. Thereafter, the autoclaved
samples were cooled to room temperature and the pH
was adjusted to 7.0 followed by addition of distilled
water to make the total volume 100 ml. After this, 5 ml
of p-bromoaniline was added and the sample was incu-
bated for 10 min at 70°C. It was then incubated at
room temperature in the dark for 70 min and the absor-
bance was taken at 540 nm.[24]
2.5.2. Estimation of cellulose
Three millilitres nitric reagent was added and mixed
properly in a vortex mixer to 1 g of the sample in test
tubes and it was placed in a water bath at 100°C for
30 min. Then the tubes were cooled to room tempera-
ture and centrifuged at 5000 rpm for 15–20 min. The
supernatant was discarded and the residue content
ENVIRONMENTAL TECHNOLOGY 3
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

was washed with distilled water. Ten millilitres of 67%
sulphuric acid was added and the sample was allowed
to stand for 1 h. From this solution, 1 ml of the sample
was taken in a 250 ml conical flask and the volume was
made 100 ml using distilled water. Then from this 100
ml solution, 1 ml was taken in a test tube and 10 ml of
anthrone reagent was added and the resulting solution
was incubated in a water bath at 100°C for 10 min.
Then the tubes were cooled to room temperature and
the absorbance was measured at 630 nm.[25]
2.5.3. Estimation of lignin
Three hundred milligrams of dried biomass was weighed
in glass test tubes and 3 ml of 72% H
2
SO
4
was added.
The samples were kept at room temperature for 2 h
with proper mixing at 30 min time intervals in a pressure
tube for proper acid hydrolysis. After the 2 h acid
hydrolysis step, the total volume was brought to 87 ml
and the samples were autoclaved at 121°C for 1 h.
After the second weak acid hydrolysis step, the hydroly-
sates were cooled to room temperature and filtered
through a vacuum pump using a filtering crucible. The
acid soluble lignin (ASL) fraction was determined by
measuring the absorbance at 240 nm.[26]
2.6. Analysis of liquid fraction
The liquid fraction obtained from vacuum filtration was
used for estimation of total sugars [27] using glucose
as the standard and the monomeric sugars were deter-
mined using HPLC. HPLC connected to a refractive
index (RI) detector and Aminex HPX-87H column
(Bio-Rad) (Shimazu Corporation, Japan) was used for
analysis of the sugars released from the grass varieties
after two-stage acid hydrolysis. Then the sugars were
eluted using 5 mM H
2
SO
4
as a mobile phase, at a flow
rate of 0.6 ml/min. The column oven and RI detector
were maintained at 60°C. HPLC grade sugars such as
glucose, xylose, and arabinose were used as standards
for identification and quantification of sugars.
2.6.1. Physicochemical characterization of the
feedstock
FTIR spectra and SEM analyses were employed in order
to investigate the structural changes in the physical
and chemical features of the untreated and pretreated
lignocellulosic biomasses (HNG and DG).
2.6.2. FTIR analysis
For FTIR analyses, the spectra of the untreated and pre-
treated lignocellulosic biomasses in the KBr phase were
recorded in a Perklin FTIR spectrophotometer, averaging
44 scans to improve signal-to-noise ratio and nominal
resolution of 2 cm
−1
.
2.6.3. SEM–EDX analysis
An SEM–EDX analysis was used to investigate the struc-
tural transformations of native and alkali pretreated
grass varieties. A SEM–EDX (Hitachi, Model: S-3400 N;
EDX –Peltier cooled X-ray head from Thermo, USA)
was used for the analyses. The detectors used for SEM
and EDX were Secondary Electron; Semiconductor
Back-scattered Electrons (Quad type)* and Cathodolumi-
nescence detector, respectively.
2.7. Neural network model of pretreatment
In this work, networks used were simple as well as robust
Multiple Layer Perceptrons (MLPs) for the model build-
ing. MLP is a feed-forward back-propagation neural
network containing three layers, namely input layer,
hidden layer, and output layer.[28,29] It has an adaptive
learning ability and does not have the capacity to make
any assumptions regarding the underlying probability
density functions or any other probabilistic information
regarding the pattern classes under consideration.
However, the data have to be carefully selected for train-
ing the ANN and also to provide a wide range and good
distribution.
In the present work the data are selected using the OA
table in the statistical design of experiments proposed by
Taguchi.[15] However, it is necessary to train an artificial
neural network before using it for a particular appli-
cation. The accuracy of the ANN after training was
assessed using the coefficient of determination (R
2
)
value and root mean-square error (RMSE) for each of
the outputs. A feed-forward back-propagation neural
network was generated using MATLAB version
7.6.0.324 R2008a which modelled the pretreatment.
The back-propagation training algorithm trail was used
to train the neural networks in MATLAB. Three inputs,
the pretreatment conditions temperature (°C), acid con-
centration (% w/w), and soaking time (h), were
mapped for four outputs, that is, total sugar, cellulose,
hemicelluloses, and ASL, from the pretreatment. The
data obtained from the experiment were randomly
divided out of which 80% of the data were used for train-
ing the network, 20% of the data were used for testing
the network, and 20% of the data were used for vali-
dation. All of the data input into the neural network
was subject to normalization of means and standard
deviations. The neural network was composed of one
hidden layer with six neurons and a sigmoid transfer
function and one output layer with one neuron and a
linear transfer function. To ensure that the six hidden
4S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

layer neurons were the best choice, the network was run
100 times over a range of 4–10 neurons to see the effect
on the R
2
and RMSE. In this experiment (data not shown),
six neurons minimized the R
2
and RMSE values, support-
ing this choice.
2.8. Results and discussion
2.8.1. Statistical optimization using Taguchi
analysis
In the present experiment, maximum delignification and
cellulose release was attained from NaOH-treated grass
samples by studying factors such as temperature (30°C,
45°C, 55°C, and 60°C), alkali concentration (0.5%, 1%,
1.5%, and 2%), and soaking time (2, 10, 18, and 26 h) in
four levels (level 1 to level 4) in the pretreatment exper-
iment in a shake flask condition. Batch-wise hydrolysis
reactions of cellulose, hemicellulose, and lignin for both
the grasses were investigated according to the L16 OA
of the Taguchi design. In this study, alkali concentration
(X1), temperature (X2), and the soaking time (X3) were
chosen as the factors that affect exposure of cellulose
and lignin degradation of HNG and DG and the four
levels were considered for each factor. The performance
parameters considered in this were: (i) mg cellulose per
gm of biomass (Y1) and (ii) % lignin per gm of biomass
(Y2). The first three parameters are used to create an
L16 OA with 16 combinations. The 16 combinations are
repeated for each sample to create the design database
for the analysis (Table 1 and 2). The influence of the
selected parameters on the total sugar, cellulose, hemi-
celluloses, and lignin yields was analysed to obtain the
optimum condition using the S/N ratio criterion of
larger is better.
The criterion considered here was criterion 1, that is, a
larger S/N ratio is best for both the performance par-
ameters (Y1 and Y2) as our goal is to enhance the cellu-
lose yield and decrease the lignin yield. The mean effect
plot of the three aforementioned factors on the cellulose
and hemicellulose yield defined as mg of cellulose
formed per gram of pretreated biomass and % of
lignin in pretreated biomass is shown in Figure 1. For
all the three factors at each level, the S/N ratio has
been tabulated in Tables 3(a) and 3(b) for HNG and DG,
Table 1. List of experiments and results according to the Taguchi L16 array for DG.
Sl. no. Concentration (%) Temperature (°C) Soaking Time (h) Cellulose (mg/g) Hemicellulose (mg/g) ASL (%)
1 0.5 30 2 302.21 107.4 3.51
2 0.5 45 10 200.07 175.3 3.22
3 0.5 55 18 365.39 173.6 3.43
4 0.5 60 26 183.22 168.9 2.66
5 1 30 10 263.25 191 3.23
6 1 45 2 287.46 133 2.45
7 1 55 26 533.87 169.7 1.03
8 1 60 18 477.09 155 1.24
9 1.5 30 18 251.01 153 1.51
10 1.5 45 26 130.57 127.6 1.68
11 1.5 55 2 124.25 158.5 2.59
12 1.5 60 10 347.63 153 1.53
13 2 30 26 247.45 134.5 1.64
14 2 45 18 262.19 145.3 1.98
15 2 55 10 242.19 140.8 1.99
16 2 60 2 272.72 137.9 2.66
Table 2. List of experiments and results according to the Taguchi L16 array for HNG.
Sl.No Concentration (%) Temperature (°C) Soaking Time (h) Cellulose (mg/g) Hemicellulose (mg/g) ASL (%)
1 0.5 30 2 141.1 160.5 2.99
2 0.5 45 10 208.49 179.5 1.65
3 0.5 55 18 296.94 156.6 1.32
4 0.5 60 26 164.26 151.6 1.55
5 1 30 10 475.95 121.8 1.99
6 1 45 2 288.52 133 1.5
7 1 55 26 326.43 122.5 1.34
8 1 60 18 593.89 178.5 1.04
9 1.5 30 18 446.47 188 1.99
10 1.5 45 26 163.18 120.4 1.98
11 1.5 55 2 197.96 229 2.45
12 1.5 60 10 365.39 174 1.66
13 2 30 26 491 135.1 1.34
14 2 45 18 546.61 193.4 1.55
15 2 55 10 481.22 135.2 1.23
16 2 60 2 336.96 136.4 2.5
ENVIRONMENTAL TECHNOLOGY 5
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

respectively, and the range was calculated as the differ-
ence of the S/N values between the highest and lowest
levels of the factor. From Table 3(a) and Figure 1(a) it is
evident that all the factors showed a positive effect on
Y1 (cellulose yield), confirming that an increase in
factor level resulted in increased performance for DG.
The range values suggest that alkali concentration (1%)
was the factor that influenced the yield of cellulose
most (533.87 mg/g) in DG and temperature (55°C) was
the factor that had the least effect. The same trend was
also observed in the case of Y1 for HNG, liberating
593.89 mg/g of cellulose at optimized parameters of con-
centration (1%), temperature (60°C), and incubation time
(18h) as given in Table 3(b) and Figure 1(d). Similarly for
Y2 (delignification), alkali concentration (1%) was the
major factor that influenced the reduction of ASL
(70.90%) for DG whereas temperature was the factor
that least affected the reduction in lignin as shown in
Table 3(a) and Figure 1(c). In case of HNG, soaking time
was the major factor that affected the delignification
(84.76%) while temperature had the least effect on Y2
(Table 3(b) and Figure 1(f)). The effect of different
Figure 1. Taguchi graphs for DG (a, b, and c) and HNG (d, e, and f).
6S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

parameters on hemicellulose content of the DG and HNG
is given in Figure 1(b) and (e), respectively. It is evident
from the figure that soaking time and alkali concentration
are the two factors that have maximum effect on the
reduction of hemicelluloses (107.4 and 121.8 mg/g) in
DG and HNG, respectively, whereas temperature has the
least effect. Effectiveness of alkaline pretreatment
depends on numerous factors but it is generally more
effective on agricultural residues and herbaceous crops
than on woods. Alkaline pretreatment techniques are
basically delignification processes in which a significant
amount of hemicellulose is solubilized.[10]
2.8.2. HPLC analysis of pretreated grass varieties
The carbohydrates released from the two NaOH-pre-
treated grass varieties (HNG and DG) before and after
the two-stage acid hydrolysis were analysed using an
HPLC (Schimazu Corporation, Japan) system. In this
process, the alkaline pretreated samples (optimized in
L16 orthogonal experiments) were hydrolysed with
72% H
2
SO
4
and then vacuum filtered following the
National Renewable Energy Laboratory procedure. The
filtrate obtained was sent for HPLC analysis, the results
of which are as given in Table 4. The major carbohydrate
contents in untreated HNG and DG were glucan (32.94%
and 37.34%), xylan (13.44% and 19.22%), arabinan
(2.44% and 2.41%), and mannan/galactan (1.75% and
1.97%) whereas for pretreated HNG and DG the carbo-
hydrate contents were glucan (46.50% and 47.94%),
xylan (7.95% and 9.35%), arabinan (2.22% and 2.15%),
and mannan/galactan (1.52% and 1.44%). After pretreat-
ment with NaOH, the glucan content increased up to
21.4% and 20.08% for HNG and DG, respectively. HNG
showed a significant increase in glucan content after pre-
treatment in comparison to DG. The xylan content
decreased up to 30.43% and 58.63% for HNG and DG,
respectively. The study conducted by Pandian et al.
[30] on pretreatment of Parthenium sp with 1% NaOH
showed an increase in the proportion of cellulose and
pentosans content by 30.5% and 22%, respectively,
whereas lignin content decreased by 16.6% in the
biomass. In the present study, the galactose and
mannose contents of the grass varieties are present in
very low quantities, whose values are less than 1.95%
of the total biomass for grasses. Similar results have
also been observed by Garlock et al. [31] which reveals
that for grasses the galactose and mannose contents
tend to be very low, in sum less than 1.5% of the total
biomass. The ASL contents of untreated HNG and DG
were 2.95% and 2.79% which decreased up to 64.7%
and 59.8%, respectively, after treatment. Generally, in
various grass species the lignin content concentration
ranges from 4.4% to 5.9%.[32] In another study Daud
et al. [33] ascertained that Napier grass contains 1.7%
of ASL, which reveals a similar low ASL content as the
other grass varieties.
2.8.3. FTIR analysis
Results of FTIR analysis of untreated and pretreated HNG
and DG are shown in Figure 2, which depict the distinc-
tive bands in the fingerprint region and the components
to which these peaks are attributable. The result indi-
cates that alkaline treatment (NaOH) causes degradation
of the fibrillar structure of cellulose and lignin to a great
extent. The absorbances (3420, 2935, 1651, 1431, 1268,
1048, and 890 cm
−1
) and (3476, 2920, 1659, 1380, 1078,
and 897 cm
−1
) are associated with untreated and pre-
treated HNG and absorbances (3412, 2939, 1670, 1248,
1130, 1067, and 949 cm
−1
) and (3400, 2925, 1642, 1168,
and 1052 cm
−1
) are associated with untreated and
treated DG, respectively. The bands at 3476 and
3400 cm
−1
depict the stretching of the hydroxyl group
in the treated HNG and DG, respectively. The absor-
bances at 2920 and 2925 cm
−1
in treated HNG and DG
Table 3. (a) S/N ratio at each level of factors and rank of factors
(DG); (b) S/N ratio at each level of factors and rank of factors
(HNG).
S/N ratio at each level of factors and rank of factors (DG)
Level
S/N ratio corresponding to Y1 S/N ratio corresponding to Y2
X1 X2 X3 X1 X2 X3
(a)
1 48.68 50.04 49.331 10.067 7.242 8.863
2 50.88 50.04 50.13 5.023 7.095 7.503
3 50.55 50.31 50.84 5.011 6.301 5.522
4 50.75 50.46 50.55 6.176 5.639 4.389
Delta 2.20 0.42 1.51 5.055 1.602 4.474
Rank 1 3 2 1 3 2
(b)
1 49.96 49.74 49.72 5.020 6.002 7.194
2 51.07 50.01 50.06 3.095 4.403 4.132
3 50.16 53.09 0.84 6.024 3.634 3.134
4 1.77 0.13 50.53 4.026 4.127 3.706
Delta 2.20 0.42 1.51 2.929 2.369 4.060
Rank 1 3 2 2 3 1
Table 4. Carbohydrate and ASL content analyses of HNG and DG.
Carbohydrates and ASL ( % dry basis in biomass)
Substrates Glucan Xylan Arabinan Mannan/galactan ASL (%) Total carbohydrate
Untreated DG 32.94 19.22 2.44 1.75 2.79 56.35
Untreated HNG 37.34 13.44 2.41 1.97 2.95 55.16
Pretreated DG 46.50 7.95 2.22 1.52 1.12 59.31
Pretreated HNG 47.94 9.35 2.15 1.44 1.04 60.88
ENVIRONMENTAL TECHNOLOGY 7
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

arise from C–H stretching in the methyl and methylene
groups. The absorbances at 1659 and 1642 cm
−1
of
treated HNG and DG correspond to the C = O stretching
in unconjugated ketones, carbonyls, and ester groups
frequently of carbohydrate origin. The absorbances at
1598, 1426, 1372.9, and 1270 cm
−1
correspond to the
aromatic skeleton vibration, ring breathing in the C–O
stretching in lignin, CH deformation in cellulose and
hemicelluloses, and C-O linkages in guaiacyl aromatic
methoxy groups, respectively.[34] Peaks at 1659, 1431,
1380, and 1268 cm
−1
in HNG and 1642 cm
−1
in DG in
the present study showed a co-relation with the results
of the above study. The findings of the present FT-IR
study were also similar to the results obtained by Irfan
et al. [35] on sugarcane baggase. The bands observed
at 1048 cm
−1
for HNG and 1052 cm
−1
for DG are typically
related to the structural characteristics of cellulose and
hemicelluloses. Hsu et al. [36] also showed similar FTIR
results with pretreated rice straw. The bands at 897
and 994 cm
−1
in HNG and DG are attributed to absorp-
tion of β-glycosidic linkages, respectively. The character-
istic absorption peaks of lignin and hemicellulose in
NaOH-pretreated grass samples of both the varieties
were significantly weakened compared to those of the
untreated grass samples, suggesting that lignin and
hemicellulose were partially degraded during the
process of pretreatment.
2.8.4. SEM–EDX analysis
The effects of NaOH pretreatment on morphological
changes in the lignocellulosic biomass of HNG and DG
were observed using SEM–EDX. The SEM images of
untreated and NaOH-treated (1.5% and 1% NaOH) HNG
and DG are shown in Figure 3(a–h). A comparison of
the structures of untreated and pretreated biomass
was performed by SEM which showed a clear, highly
ordered, and smooth surface for untreated substrate/
biomass,(Figure 3(a) and 3(e)) whereas the biomass
treated with NaOH appeared to be disrupted and
rough in structure (Figure 3(b–d) and 2(f–h)). Moreover,
specifically the untreated substrate of HNG showed a
linear array of biomass. On the other hand, the pre-
treated surface of the grass biomass showed that some
Figure 3. SEM analysis of untreated DG and HNG (a and e) and 1.5% NaOH-treated DG(b, c, and d) and 1% NaOH-treated HNG (f, g and h).
Figure 2. FTIR analysis of untreated and treated DG and HNG.
8S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

parts of the outer surface were missing and unfolded.
These findings indicate that alkali pretreatment
prompted the removal of some external fibres.[37] The
removal of these external fibres will lead to an increase
in the external surface area and porosity of the pre-
treated grass biomass, ultimately leading to an improve-
ment of the enzymatic hydrolysis efficiency.[37] Barman
et al. [38] conducted SEM analysis of pretreated wheat
straw and reported that 2% NaOH-treated straw
exposed more cellulose fibres. Identical observations
were reported through SEM analysis of bamboo pre-
treated with cold NaOH/urea [39] and also with sugarcane
tops pretreated with NaOH.[40] Similar structural changes
due to NaOH pretreatment have been reported by Bak
et al. [41]Koetal.[42] and Remli et al. [37]intheirexper-
iments on rice straw and Asghar et al. [43] on wheat straw.
Furthermore, the structure of the lignocellulosic biomass
was opened up and more sponge-like structures were
observed after NaOH pretreatment, which can provide
higher surface area for subsequent enzymatic reactions.
[44]
Further analysis with EDX (Figure 4) along with SEM
clearly showed an increase in the carbon compound
and decrease in the potassium component in the grass
biomass of both the varieties after pretreatment with
alkali (NaOH). But the EDX results of DG are compara-
tively less conspicuous than those of HNG. Reduction
in potassium contents of the grass feedstock grown for
fuel purposes is important to achieve combustion
efficiencies.[45] When the grass materials are pulped,
the entire plant is used and hence the undesirable
elements like potassium (K), Nitrogen (N), Magnesium
Figure 4. EDEX graphs for untreated and pretreated DG (I) and HNG (II).
ENVIRONMENTAL TECHNOLOGY 9
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

(Mg), Calcium (Ca), Phosphorous (P), Chlorine (Cl), etc. are
carried along with it.[46] In the present study, these
undesirable elements were mostly absent for the
NaOH-treated grass samples of the two varieties.
Specifically HNG showed more absence of the undesir-
able elements like Mg, P, N, and Ca than DG. Therefore,
pretreatment plays an important role in the elimination
of these elements for better productivity of bioethanol.
2.8.5. ANN analysis
The results, shown in Table 6(a) for DG and Table 5(b) for
HNG, show that the standard deviations of the R
2
statistic
are reasonable compared to the average values,
suggesting that the ANN is consistently learning similar
relationships between inputs and outputs independent
of how the data are chosen. However, the standard devi-
ation of RMSE is large in comparison to the average value,
suggesting alternatively that some values predicted by
the ANN can be affected by how the training data are
chosen. In these cases, there is more chance of a few
Table 6. Average ANN R
2
and RMSE.
Output R
2
RMSE
Total sugar 0.98 0.027
Cellulose 0.96 0.08
Hemicellulose 0.89 0.205
Lignin 0.92 0.33
Table 5. (a) ANOVA for DG; (b) ANOVA for HNG.
DF Seq SS Adj SS Adj MS FP
(a)
Factor for Y1
X1 3 24048 24048 8015.9 8.38 .014
X2 3 1888 1888 629.4 0.66 .607
X3 3 8369 8369 2789.8 2.92 .122
Residual error 6 5737 5737 956.2
Factor for Y2
X1 3 33.281 33.281 11.094 5.34 .039
X2 3 3.361 3.361 1.120 0.54 .672
X3 3 108.323 108.323 36.108 17.40 .002
Residual error 6 12.453 12.453 2.076
(b)
Factor for Y1
X1 3 16447.1 16447.1 5482.4 5.41 .038
X2 3 781.0 781.0 260.3 0.26 .854
X3 3 7412.4 7412.4 2470.8 2.44 .163
Residual error 6 6081.1 6081.1 1013.5
Factor for Y2
X1 3 54.086 54.086 18.029 4.72 .051
X2 3 8.747 8.747 2.916 0.76 .554
X3 3 117.999 117.999 39.333 10.31 .009
Residual error 6 22.897 22.897 3.816
Figure 5. ANN graphs showing the actual and predicted values for DG and HNG.
10 S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

large outliers despite a seemingly good fit otherwise.
Figure 5 (a–d) shows plots of average predicted versus
actual values for cellulose, hemicellulose, total sugars,
and lignin of pretreated samples of DG and HNG, respect-
ively, from the ANN which was used. As seen in the plots,
the ANN predicts the majority of values very closely to the
experimental data with few outlying values when predict-
ing total sugars, cellulose, and lignin from the pretreated
biomass. Hemicellulose values are predicted less accu-
rately with many values being either higher or lower
than the actual value by a relatively large amount. In an
investigation by Ezhumalai et al. (2010) [47] ANN was
used for optimization of incubation temperature (25–
45°C), pH (5–7), and fermentation time (24–120 h) for pre-
treatment and ethanol production of sugarcane baggase
into ethanol. Table 6 contains the R
2
and RMSE values for
the ANN plots as shown in Figure 5.
3. Conclusion
Pretreatment with NaOH of two grass cultivars (HNG and
DG) resulted in proper delignification (73.6% and 70.4%)
and maximum exposure of cellulose (50.6% and 47.2%)
for HNG and DG, respectively, as evidenced by SEM–EDX
and FTIR analysis for further accessibility of enzymes for
cellulose hydrolysis essential for bioethanol production.
The hemicelluloses for HNG and DG were reduced by
22.0% and 22.35%, respectively. Statistical methods like
the one by Taguchi were helpful to determine
maximum delignification and cellulose release for NaOH-
treated grass samples by choosing appropriate par-
ameters for pretreatment experiments. The optimized par-
ameters were alkali concentration (1%), temperature (60°C
and 55°C), and soaking time (18 and 26 h) for HNG and
DG, respectively. ANN was effective for accurately model-
ling all the experimental data of total sugars, cellulose,
hemicelluloses, and lignin from pretreated biomass and
showed very close similarity to experimental data.
Disclosure statement
No potential conflict of interest was reported by the authors.
References
[1] Olsson L, Hahn-Hägerdal B. Fermentative performance of
bacteria and yeasts in lignocellulose hydrolysates. Process
Biochem. 1993;28:249–257.
[2] Aristidou A, Pentlila M. Metabolic engineering appli-
cations to renewable resource utilization. Curr Opin
Biotechnol. 2000;11:187–198.
[3] Bullock GE. Ethanol from sugarcane. Australia: Sugar
Research Institute; 2002.
[4] Zhang M, Eddy C, Deanda K, Finkelstein M, Picataggio S.
Metabolic engineering of a pentose metabolism
pathway in ethanologenic Zymomonas mobilis. Science.
1995;267:240–243.
[5] Mukhtar M, Ishii Y, Tudsri S, Idota S, Sonoda T. Dry
matter productivity and overwintering ability in the
dwarf and normal napier grasses as affected by the plant-
ing density and cutting frequency. Plant Prod Sci.
2003;6:65–73.
[6] Hamelinck CN, Hooijdonk GV, Faaij APC. Ethanol from lig-
nocellulosic biomass: techno-economic performance in
short-, middle- and long-term. Biomass Bioenergy.
2005;28:384–410.
[7] Palmqvist E, Hahn-Hagerdal B. Fermentation of lignocellu-
losic hydrolysates.I: inhibition and detoxification: review.
Bioresource Technol. 2000;74:17–24.
[8] Pandey A, Soccol CR, Mitchell W. New development in
solid state fermentation: I-Bioprocess and products.
Process Biochem. 2000;35:1153–1169.
[9] Seng B, Khanal SK, Visvanathan C. Anaerobic digestion of
combined ultrasound and chemical pretreated waste acti-
vated sludge. Environ Technol. 2010;3:257–265.
[10] Chang VS, Burr B, Holtzapple MT. Lime pretreatment of
switchgrass. Appl Biochem Biotechnol. 1997;63–65:3–19.
[11] Khanal SK, Rasmussen M, Shrestha P, Leeuwen HJV,
Visvanathan C, Liu H. Bioenergy and biofuel production
from wastes/residues of emerging biofuel industries.
Wat Envrion Res. 2008;80:1625–1647.
[12] Sun Y, Cheng JJ. Hydrolysis of lignocellulosic materials for
ethanol production: a review. Bioresource Technol.
2002;83:1–11.
[13] Chang VS, Holtzapple MT. Fundamental factors affecting
biomass enzymatic reactivity. Appl Biochem Biotechnol.
2000;84–86:5–38.
[14] Lynd LR, Weimer PJ, Zyl WHV, Pretorius IS. Microbial cellu-
lose utilization: fundamentals and biotechnology.
Microbiol Molecul Biol Review. 2002;66:506–577.
[15] Taguchi G. Introduction to quality engineering. Designing
quality into products and processes. Tokyo, Asian
Productivity Organization; 1986.
[16] Dwyer JO, Zhu L, Granda CB, Chang VS, Holtzapple MT.
Neural network prediction of biomass digestibility
based on structural features. Biotechnol Progr.
2008;24:283–292.
[17] Ehrman T. Standard method for determination of total
solids in biomass. NREL Ethanol Project Laboratory
Analytical Procedure #001. NREL, Golden, CO, 1994.
[18] Kiani A, Shahbazi M, Asempour H. Hydrogel
membranes based on gum tragacanth with tunable struc-
ture and properties. I. Preparation method using
Taguchi experimental design. J Appl Polym Sci.
2012;124:99–108.
[19] Wang J, Wan W. Effect of temperature on fermentative
hydrogen production by mixed cultures. Int J Hydrogen
Energy. 2008;33:5392–5397.
[20] Radhakumaria M, Ballb A, Bhargavac SK, Satyavathia B.
Optimization of glucose formation in karanja biomass
hydrolysis using Taguchi robust method. 2014. In press.
[21] Chen GL, Chen GB, Li YH, Wu WT. A study of thermal pyrol-
ysis for astor meal using the Taguchi method. Energy.
2014;71:62–70.
[22] Wang LJ, Xiu S, Shahbazi A. Pretreatment and fraction-
ation of wheat straw with acetic acid solution. ASABE
Annual Meeting; 2009 June 21–25; Reno, Nevada.
ENVIRONMENTAL TECHNOLOGY 11
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015

[23] Xu J, Cheng JJ, Shivappa RRS, Burns JC. Sodium hydroxide
pretreatment of switch grass for ethanol production.
Energy Fuels. 2010;24:2113–2119.
[24] Roe JH, Rice EW. A photometric method for the determi-
nation of free pentoses in animal tissues. J Biol Chem.
1948;173:507–512.
[25] Updegraff DM. Semimicro determination of cellulose in
biological materials. Anal Biochem. 1969;32:420–424.
[26] Sluiter A, Hames B, Ruiz R, et al. Determination of structural
carbohydrates and lignin in biomass. NREL/TP-510-4261.
2008.
[27] Dubois M, Gilles KA, Hamilton JK. Colorimetric method for
determination sugars and related substances. Analytical
Chem. 1956;28:350–356.
[28] Hu Z, Wang Y, Wen Z. Alkali (NaOH) pretreatment of
switch grass by radio frequency-based dielectric heating.
Appl Biochem Biotechnol. 2008;148:71–81.
[29] Pandiyan K, Tiwari R, Singh S, et al. Optimization of enzymatic
saccharification of alkali pretreated Parthenium sp.using
response surface methodology. Enzyme Research. 2014:1–8.
[30] Sanjari M, Taheri AK, Movahedi MR. An optimization
method for radial forging process using ANN and
Taguchi method. Int J Adv Manuf Tech. 2009;40:776–784.
[31] Garlock RJ, Balan V, Dale BE, et al. Comparative material
balances around pretreatment technologies for the con-
version of switchgrass to soluble sugars. Bioresour
Technol. 2011;102:11063–11071.
[32] Rafay A, Majumdar S, Prakash B. Exploring potassium-
dependent GTP hydrolysis in TEES family GTPases. FEBS
Open Bio. 2012;2:173–177.
[33] Daud Z, Hatta MZM, Kassim ASM, Awang H, Aripin AM.
Exploring of agro waste (pineapple leaf, corn stalk and
napier grass) by chemical composition and morphological
study. Bioresource. 2014;9:872–880.
[34] He Y, Pang Y, Liu Y, Li X, Wang K. Physiochemical charac-
terization of rice straw pretreated with sodium hydroxide
in the solid state for enhancing biogas production. Energy
Fuels. 2008;22:2775–2781.
[35] Irfan M, Gulsher M, Abbas S, Syed Q, Nadeem M, Baig S. Effect
of various pretreatment conditions on enzymatic saccharifica-
tion. Songklanakarin J Sci Technol. 2011;33:397–404.
[36] Hsu TC, Guo GL, Chen WH, Hwang WS. Effect of dilute acid
pretreatment of rice straw on structural properties and enzy-
matic hydrolysis. Bioresour Technol. 2010;101:4907–4913.
[37] Remli NAM, Shah UKM, Mohamad R, Aziz SA. Effects of
chemical and thermal pretreatments on the enzymatic
saccharification of rice straw for sugars production. Bio
Resources. 2014;9:510–522.
[38] Barman DN, Haque MA, Kang TH, et al. Alkali pretreatment
of wheat straw (Triticum aestivum) at boiling temperature
for producing a bioethanol precursor. Biosci Biotechnol
Biochem. 2012;76:2201–2207.
[39] Li X, Kim TH, Nghiem NP. Bioethanol production from corn
stover using aqueous ammonia pretreatment and two-
phase simultaneous saccharification and fermentation
(TPSSF). Bioresour Technol. 2010;101:5910–5916.
[40] Sindhu R, Kuttiraja M, Preeti VE, Vani S, Sukumaran RK,
Binod P. A novel surfactant-assisted ultrasound pretreat-
ment of sugar cane tops for improved enzymatic release
of sugars. Bioresour Technol. 2013;135:67–72.
[41] Bak JS, Ko JK, Choi IG, Park YC, Seo JH, Kim KH. Fungal pre-
treatment of lignocellulose by Phanerochaete chrysospor-
ium to produce ethanol from rice straw. Biotechnol
Bioeng. 2009;104:471–482.
[42] Ko JK, Bak JS, Jung MW, et al. Ethanol production from
rice straw using optimized aqueous-ammonia soaking
pretreatment and simultaneous saccharification and fer-
mentation processes. Bioresour Technol. 2009;100:4374–
4380.
[43] Asghar U, Irfan M, Iram M, et al. Effect of alkaline pretreat-
ment on delignification of wheat straw. Nat Prod Res.
2015;29:125–131.
[44] Goshadrou A, Karimi K, Taherzadeh MJ. Bioethanol pro-
duction from sweet sorghum bagasse by Mucor hiemalis.
Ind Crops Prod. 2011;34:1219–1225.
[45] Roger S, Bano M. Strategies to reduce the ash content in
perennial grasses. Bioenergy 98. Expanding Bioenergy
Partnerships. 1998. http://www.reap-canada.com/online_
library/feedstock_biomass/24%20Strategies%20to.pdf
[46] Feasibility of Improving Biomass Combustion through
Extraction of Nutrients, Prepared For: Ontario Federation of
Agriculture, June 17, 2011. http://www.ofa.on.ca/uploads/
userfiles/files/cennatek%20ofa%20report-feasibility%20of%
20improving%20biomass%20combustion%20through%
20extraction%20of%20nutrients.pdf.
[47] Ezhumalai S, Thangavelu V. Kinetic and optimization
studies on the bioconversion of lignocellulosic material
into ethanol. Bioresources. 2010;5:1879–189.
12 S. MOHAPTRA ET AL.
Downloaded by [Hrudayanath thatoi] at 00:49 24 November 2015