Content uploaded by Patrick Nnaji
Author content
All content in this area was uploaded by Patrick Nnaji on Jun 25, 2021
Content may be subject to copyright.
Content uploaded by Valentine Chikaodili Anadebe
Author content
All content in this area was uploaded by Valentine Chikaodili Anadebe on Jun 25, 2021
Content may be subject to copyright.
Dual-purpose optimization of dye-polluted wastewater decontamination
using bio-coagulants from multiple processing techniques via neural
intelligence algorithm and response surface methodology
O.D Onukwuli
a
, P.C Nnaji
b,
*, M.C Menkiti
a
, V.C. Anadebe
c,d,e
, E.O Oke
b
, C.N. Ude
b
, C.J. Ude
b
,
N.A. Okafor
c
a
Department of Chemical Engineering, Nnamdi Azikiwe University, Awka, Nigeria
b
Department of Chemical Engineering, Michael Okpara University, Umudike, Nigeria
c
Department of Chemical Engineering, Alex Ekwueme Federal University Ndufu-Alike, Nigeria
d
CSIR-Central Electrochemical Research Institute Karaikudi, 630003 Tamil Nadu, India
e
Academy of Scientific and Innovative Researh (AcSIR), Ghaziabad 201002, India
ARTICLE INFO
Article History:
Received 22 February 2021
Revised 26 May 2021
Accepted 12 June 2021
Available online xxx
ABSTRACT
Novel Luffa cylindrica seed (LCS) extracts obtained from different processing techniques were employed for
coagulation/flocculation (CF) decontamination of dye-polluted wastewater (DPW). The DPW was simulated
in the laboratory using Cibacron blue dye 3GA (a reactive, azo dye) and distilled water. The bio-coagulants'
proximate and instrumental characterization was performed. The duo: Response Surface Methodology
(RSM) and Artificial Neural Network (ANN) models were proposed to predict color/total suspended particle
(CTSP) and chemical oxygen demand (COD) removal rate using bio-coagulants. Bio-coagulant dosage, waste-
water pH, and stirring time are the input variables. Based on experimental designs, RSM and ANN models
have been generated. Regression coefficient (R
2
) and mean square error (MSE) have been implemented and
correlated to test the adequacy and predictive ability of both models. The fitness of the experimental values
to the expected values established that the Sutherland extract performed better. The model indicator for
Sutherland extract revealed as thus: RSM (R
2
,0.9886 and MSE, 1.4494) for CTSP, and (R
2
, 0.9921 and MSE,
0.9249) for COD; and ANN (R
2
, 0.9999 and MSE, 0.00000057) for CTSP and (R
2
, 0.9999 and MSE,
0.0000000457) for COD. The obtained results revealed that ANN model was preferred for predicting the
removal of CSTP and COD from DPW.
© 2021 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
Keywords:
RSM-ANN
Coagulation/flocculation
Luffa cylindrica
Different processing techniques
Dye-polluted wastewater
1. Introduction
The value of preserving a protected environment for all living
beings cannot be overstated. One of the primary causes of environ-
mental and public health deterioration is water contamination [1,2].
The continuous advances in dye production have resulted in the rapid
increase in wastewater generation especially from the dye-based
industries. This is particularly to meet the textile needs of the ever-
increasing population. Globally, over 10,000 varieties of textile indus-
try dyes are commercially used, with annual production totalling sev-
eral metric tonnes [35]. The pollutants in wastewater of dye origin
usually come from manufacturing chemicals, additives, pigments and
dyes, which increase the biochemical oxygen demand (BOD), COD,
solid content and toxicity [1,6]. In most industries where dye materi-
als are in use, around 2% of the dye is released into the effluent during
production processes, and approximately 10% of the dye is lost during
the colouring phase. The toxicity and little presence of dye colour is
highly visible and harmful to aquatic life [7]. The decay of dye con-
taminants could produce a harmful compound that could cause
genetic mutation and toxicity to aquatic living creatures and cancer,
hemorrhage, fetus cerebral abnormalities, and dermatitis in human
beings [8,9]. Dye and compounds from dye decay are not biodegrad-
able because of its complex molecular structures and large molecular
mass, causes wide pH-value range, dark colourity and high toxicity in
water it contaminates [1013].
Removal of dye pollutants requires multiple treatment techni-
ques, such as membrane separation, aerobic and anaerobic deteriora-
tion, chemical oxidation, adsorption, CF, among others [1418]. Each
of these techniques, whether biological, chemical or physical has its
own limitation which could be high cost, performance efficiency, fea-
sibility and environmental effect [1921]. Given the advancement in
these techniques, especially the application of adsorption techniques
in dye removal from wastewater, the CF approach stands out as the
* Corresponding author.
E-mail address: pc.nnaji@mouau.edu.ng (P.C. Nnaji).
https://doi.org/10.1016/j.jtice.2021.06.030
1876-1070/© 2021 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
Journal of the Taiwan Institute of Chemical Engineers 000 (2021) 115
Contents lists available at ScienceDirect
Journal of the Taiwan Institute of Chemical Engineers
journal homepage: www.elsevier.com/locate/jtice
preferred treatment option. This is because of its easy on-site imple-
mentation, great performance, flexibility and low assembly and oper-
ating costs [20,22].
Coagulation/flocculation process is the application of coagulant to
wastewater to reduce the stability of colour/colloidal dispersion and
afterwards aggregate the individual particles resulting from it. Hence,
the suspended/dissolved particles could be eliminated from the
wastewater [22,23]. CF mechanism that relies on the solution's physi-
cal and chemical properties, the contaminants present and the form
of coagulant include hydrolysis, coagulation, peri-kinetic and orth-
kinetic flocculation [24]. CF is one of the foundation processes at
most water and wastewater treatment plants where they are of
utmost significant in the practice of solid-liquid separation [23,25].
Conventional coagulants like iron salts, aluminium salts etc are
widely applied to extract different contaminants from wastewater
[26]. The browning of equipment when iron salts are applied and the
post contamination associated with aluminium salts are inherent
downside of conventional coagulants [10,26,27]. Alzheimer's disease
in humans, generation of large volume of sludge with its attendant
disposal cost and low performance of aluminium salts in reduced
temperature waters has also been documented [22,28].
The ability to reduce the aforementioned hazards and growing
need to reduce the eco-system problems associated with the applica-
tion of traditional coagulants in wastewater treatment has made the
use of biomass as an effective substitute relevance. Natural biomass
among others, like Moringa olifera, tannins, Detarium microcarpum,
etc. were reported as been effective in decontaminating wastewater
[22,29]. Wastewater treatment using biomass is an eco-friendly
approach and the generated sludge (sequel to biological conversion)
could serve as soil conditioners [10]. The coagulant used for this
study was obtained from Luffa cylindrica plant. The L. cylindrica plant
is widely grown in Nigeria and has been used as a filter sponge and
support for adsorptive biomass [30], but the seed extract has limited
use in wastewater treatment. This biomass is biodegradable, non-
toxic and safe for both humans and animals. The use of L. cylindrica
seed naturally for the treatment of wastewater with high perfor-
mance and reduced sludge generation has been reported [31]. The CF
process is affected by: temperature, pH, wastewater quality, the
quantity and type of coagulant applied among other factors. The opti-
mization of these variables will greatly improve the performance of
the process. Notwithstanding the time-consuming features of the tra-
ditional method of experiment, it is not possible to secure the precise
option because the influences across variables are not considered.
Existing conceptual methods were implemented to resolve this con-
straint; the statistically dependent method and black box approach
of artificial intelligence.
For statistically dependent approach, the RSM jar test, as proposed
by Box-Behnken using Design Expert 10.0 was adopted. RSM, an
assembly of statistical theorem was proposed for advancing experi-
ment design models, evaluating the influence of variables and having
the most favourable variables’state [10]. Treatment of brewery efflu-
ent using Detarium microcarpum, oyster dried shell [22,27], treatment
of textile wastewater using Cereus peruvianus and FeCl
3
[1,10] are
some of the articles that have reported on the optimization of waste-
water decontamination through coagulation and flocculation using
RSM with good results. The artificial intelligence approach has been
recognized as a valuable tool for simulation and process optimization
in the last two decades, particularly for non-linear multivariate
modelling [32,33]. ANN is a self-adaptive, data-driven approach that
needs least prior assumptions about the process been studied [34].
ANN, a structure-generic mathematical model that emulates the neu-
ral activity of the human brain has the ability to learn from historical
evidence. ANN attempts to measure the response of each input set,
compares it to the output given and correct the deviated response by
altering internal weights. The hit and test progress until the variation
in the outputs and responses is minimized [32]. In several fields, such
as wastewater treatment [32], membrane science and technology
[35], building and construction [36], power technology [37], environ-
mental engineering [38], corrosion [39] etc, artificial intelligence has
been extensively used in modelling, forecasting, fault dictation, and
process regulation.
The key benefit of ANN over RSM is: (i) ANN does not need an
acceptable fitting function to be defined beforehand, and (ii) it has a
high predictive ability, while RSM is only suitable for quadratic esti-
mation. Nevertheless, RSM is more beneficial to gain detailed infor-
mation such as sensitivity analysis and interactive effects of two
components in the system [32]. There are few studies available on
the use of the duo for the optimization of CF modeling for the treat-
ment of DPW using bio-coagulant [40,41]. Nevertheless, the use of
LCS processed using various techniques has not been documented.
The relevance and novelty of the present research, therefore.
The present work has tripartite purposes: (i) optimizing the
decontamination of DPW through coagulation/flocculation method
using LCS processed through different techniques; (ii) determining
the best processing technique; (iii) comparing the efficacy of optimi-
zation techniques based on statistical approach and artificial intelli-
gence. Two responses were evaluated; %CTSP and %COD removal
from DPW. The novelty of the work is the determination of the best
processing technique for LCS through optimization using a statistical
approach and artificial intelligence.
2. Materials and methods
The Cibacron blue (3GA), a reactive dye was obtained in Enugu,
Nigeria. Matured L. Cylindrica sponges were collected from a farm set-
tlement at Amawom, Ikwuano L.G.A, Abia State, Nigeria. DPW was
prepared to get a concentration of 1000 mg/L using distilled water.
2.1. Methodology
2.1.1. Luffa cylindrica seed (LCS)
The collected L. cylindrica sponges were sun-dried and the seeds
were subsequently separated from the sponges. The obtained seeds
were washed, dried and homogenized to obtain LCS bio-coagulant
(LCSC
0
) which was stored in an airtight container. The proximate LCS
analysis was based on standard methods [42].
Part of LCSC
0
obtained above was further processed using water
extraction [43]and modified procedure reported by Sutherland
[44,45] and termed LCSC
1
and LCSC
2
, respectively.
2.1.2. Biomass and DPW characterization
The description of the distinctive nature of DPW has been investi-
gated using standard methods [46]. Mettler Toledo Delta 320 pH
meter, DDS 307 conductivity meter, and UNICO 1100 spectrophotom-
eter were used to determine the solution pH, electrical conductivity
and CTSS. Also, the different coagulants used in the study were char-
acterized with the following instruments: an infrared spectrometer
from Agilent Technologies, equipped with a resolution range of
4000650 cm
-1
and 30 scans at 8 cm
-1
with 16 background scans, FT-
IR spectra were developed. A scanning electron microscope from
Phenon-World, with model number, MVE 016477830 developed the
sample's morphology.
2.2.3. Coagulation/flocculation procedure (Jar Test)
A jar test procedure was applied in each case on 200 mL of DPW
with varying pH of 2, 4, 6, 8 and 10 using 1000, 1200, 1400, 1600 and
1800 mg/L dosages of LCSC, respectively. Five minutes rapid stirring
at 150 rpm followed by 25 min slow stirring at 25 rpm was adopted.
By applying 0.1 M sulphuric acid and 0.1 M sodium hydroxide, sepa-
rately, pH alteration was achieved. The solution was allowed to settle
for 300min. Thereafter, the supernatant from each cylinder was
assessed for CTSP and COD to evaluate the level of their removal.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
2O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
2.3. Empirical modeling and optimization approach
2.3.1. Response surface methodology (RSM)
Numerous approaches of RSM like Central Composite Design
(CCD), Box-Behnken Design (BBD) and Full Factorial Design (FFD) in
the prediction of optimal output were use in many resaerches [5,22].
BBD with a three-level architecture of measured influences with all
factors was applied. The architecture varied over three positions each
of the factors numbered as; a high (+ 1), mid (0) and low (-1). To
achieve the proposed design, 17 experimental runs was required
(Table 1). The reaction factors depicted as Y are the %CTSP(Y
1
) and
%COD(Y
2
) removal. The chosen variables for the study were the LCSC
dosage(X
1
), wastewater pH(X
2
) and stirring time(X
3
). The actual fac-
tors used in the experiment are presented in Tables 1. The arrange-
ment is to agree with a quadratic polynomial equation pattern
[10,24,47]Eq. (1).
y¼boþXk
i¼1biXiþXk
i¼1biiX2
iþXk
iXjbijXiXjð1Þ
Where y is the variable response to be modelled; X
i,
and X
j
are the
independent variables influencing y, b
o,
b
i,
b
ii
and b
ij
the offset terms,
the ith linear coefficient, the iith quadratic coefficient and the ijth
interaction coefficient, respectively.
2.3.2. Artificial neural network
By following certain defined patterns, the ANN, simulates human
neural network to tackle critical challenges. For data processing, ANN
performs computational tasks (using neurons and linkages comput-
ing network) without obeying any specified rules. The traditional
methods learn via a defined scheme, whereas ANN, learns via trends
[5,32,35]
An individual computational processor that functions with (i)
summing junction and (ii) transfer function [35] is an artificial neu-
ron. The relation is made up of weights, w, and biases, b, with neu-
rons that address data. A discrete neuron's summing junction
operator sums up the weight and bias into an input node, S, known
as the logic to be presented:
S¼Xn
i¼1xiwiþbð2Þ
Where x
i
is the input parameter. The transfer function takes the
logic and generates a single neuron's scalar output. Purelin, logsig,
and tansig are the popular transfer functions for solving linear and
non-linear regression problems. The logsig transfer function can be
written as:
logsig SðÞ¼ 1
1þexp SðÞ ð3Þ
The linkage between the inputs and outputs of the neurons is
known as the architecture (Fig. 1). The neurons of such a network are
split into several bands, called layers. The input layer of three neurons
(coagulant dosage, wastewater pH, and stirring time), output layer
(CTSP and COD removal), and hidden layer make up a multi-layer
neural network. The multi-layer feed-forward neural network, often
Table 1
BB-design and response generated after investigating CTSP and COD removal from DPW with bio-coagulant obtained
via multiple processing techniques.
Runs Dosage (mg/L) pH Stirring time (min) CTSP COD
LCSC
0
LCSC
1
LCSC
2
LCSC
0
LCSC
1
LCSC
2
1. 1400 6 20 55.93 69.78 63.02 89.89 72.8 72.09
2. 1000 6 10 49.62 77.95 78.49 87.1 85.48 84.02
3. 1800 2 20 88.53 99.64 98.81 98.59 96.98 99.62
4. 1400 2 10 86.18 98.1 99.55 98.1 97.97 99.82
5. 1400 10 10 41.98 49.7 62.11 83.3 63.21 65.08
6. 1000 6 30 50.28 68.88 73.98 85.34 87.49 80.11
7. 1000 10 20 36.33 55.69 68.51 84.01 63.12 72.08
8. 1800 6 10 55.45 79.98 75.01 87.8 87.16 86.38
9. 1000 2 20 83.16 95.09 95.55 96.59 96.94 96.22
10. 1400 6 20 52.04 67.11 67.02 89.44 75.8 75.04
11. 1400 6 20 52.26 67.06 64.02 88.89 76.8 77.2
12. 1400 6 20 56.7 72.81 69.02 89.01 74.8 75.4
13. 1400 6 20 55.43 70.83 68.02 89.44 72.8 74.04
14. 1400 10 30 55.39 64.71 67.21 85.3 73.21 71.11
15. 1400 2 30 97.83 98.9 99.96 97.07 99.6 99.84
16. 1800 10 20 45.4 60.24 57 88.93 70.04 68.42
17. 1800 6 30 64.18 89.79 85.13 90.8 90.93 91.54
X1
X2
X3
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ logsig
∑ purelin
Inputs
Hidden layers
Output
Fig. 1. Artificial neural network framework.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 3
Fig. 2. a. FTIR spectra of (LCS) (a), LCS Extract (b) b. SEM graphics (a) LCS, LCS Extract (b). c. The XRD spectra of (a) LCS, (b) LCS extract.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
4O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
identified as the multi-layer perceptron, MLP, is the regular algorithm
used to resolve non-linear regression problems. The most prevalent
feed-forward neural network training algorithm is called back-propa-
gation (BP). ANN training by virtue of the backpropagation algorithm
is an iterative method of maximization where the output function is
reduced by properly adjusting the weights. The experimental data
sets are needed to train the ANN model for forecasting the output
after choosing the architecture. A total of 100 experiments from DoE
used in RSM were applied randomly in the current investigation
using the same variables and responses to achieve convergence. It is
often beneficial to scale the input and target data prior to training to
compel them to fall in the same range. After scaling, these 100 experi-
mental data points were fed into the ANN structure where samples
were sub-divided into training (70), validation (15), and testing (15)
subgroups, respectively. From the research data collection, the sam-
ples for validation and testing were arbitrarily picked to evaluate the
adequacy of the model. The performance function generally used is
MSE, defined as:
MSE ¼1
HM X
m
i¼1
ðypyaÞ2ð4Þ
M is the number of trends used in the training dataset, where H
represents the number of output nodes, y
p
and ya are the goals
(experimental) and (predicted) of the output node, respectively.
MATLAB software's neural network tool box has been used for scien-
tific programming and ANN model generation.
3. Results and discussion
3.1. Characterization of LCS and DPW
The proximate composition of LCS indicated 9.3% moisture, 2.16%
ash, 6.8% fat, 21.88% crude protein, 60.46% carbohydrate and 8.7%
crude fibre. The results indicated reasonable content of crude protein
Fig. 2. Continued.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 5
Fig. 3. Impact on CTSP (a) & (b) and COD (c) & (d) removal by bio-coagulant dosage and wastewater pH. e. Impact on CTSP removal at optimum bio-coagulant dosage and varying
wastewater pH and settling time for Sutherland processing technique.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
6O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
when compared to those reported previously. Documented reports
indicate defatted LCS crude protein content; 42.1770.65% [48],
45.0650.06% [49], non-defatted LCS; 28.45% of unrefined protein
[50]. It has been established that unrefined protein is the major work-
ing component aiding coagulation/flocculation. The results of DPW
characterization, indicate high solids/dissolved particles (SDP);
248.68mg/L, BOD
5
(39.63mg/L), COD (303.02mg/L) and pH (6.85) and
pH (6.85) in addition to heavy metals. These are pollution indicators;
hence, the wastewater requires treatment to remove these pollu-
tants.
3.1.1. Instrumental characterization
Fourier transform infrared spectroscopy: To be able to exploit this
bio-coagulant, it is essential to comprehend the nature of the mate-
rial [10].Fig. 2a showed the identification of functional group of the
polymeric-OH and N-H stretching of amide groups usually found in
protein structure at 3276.3 cm
1
, an asymmetric group of methyl C-
H and methylene C-H of the CH
2
group found in the fatty acids
extending between 2951.1 and 2914.8 cm
1
, respectively, in LCS and
its extract. Also, 1707.1 cm
1
could be assigned to carboxylic ketone
extension [25,51]. For the attachment of colloidal particles and cer-
tain ions in solution, the existence of OH, carboxyl extension and
other groups provides active sites [10,52].
Scanning electron microscopy: The bio-coagulant surface structure
was studied by applying SEM. The fibrous nature of LCS and its
extract [30] with some fissures and pores [53] was shown in Fig. 2b.
This is an indication of the existence of a macroporous structure that
introduces CF and adsorption active sites [54].
XRD outcomes of LCS and its extracts: X-ray diffraction was used to
study the material's crystalline structure. Fig. 2c displays the X-ray
diffraction spectra for both LCS and its extract. The Figures show
some high peaks, numerous halos with amorphous humps and lots of
noise between 10 <2u<70
o
strip. This indicates characteristic pat-
terns of a partially crystalline nature where scattered bands were
formed by poorly organized molecules. If material studied indicates
well defined peaks, it is said to be crystalline, whereas non-crystalline
or amorphous material displays halos or humps [37,55].
3.2. Individual review of parameters
Firstly, the individual impact of the process variables was
explored. By keeping one other parameter constant, the impact of
each operating variable was analyzed. The choice of ranges was
basedondocumentedreports[1,10,22,29]. The result obtained at
optimum conditions are displayed in Fig. 3.Thefigure showed
the impact of LCS dosage and DPW pH at constant slow stirring
time of 25mins. While Fig. 3a and b illustrates the impact of these
variables for CTSP removal, Fig. 3c and d illustrates the effect for
COD removal. The removal efficiency decreased after the opti-
mum pH of 2 was reached, as shown in the figures. Charge rever-
sal may be responsible for this. The removal efficiency increased
at higher pH, which could be due to the fact that dye particles
maintained their anionic charge at higher pH, making it easier for
positively charged coagulants to combine with the particles [1].
The dosage effect followed a similar pattern, with the exception
that higher dosages after the optimum dosage did not signifi-
cantly boost performance. The findings are consistent with pub-
lished reports [1,22,29].
With the best bio-coagulant extract from Sutherland processing
techniques for the removal of CTSP, the individual effect of settling
time was investigated at optimum pH and dosage (Fig. 3e). Despite
the 300 min settling period, >90% removal efficiency was achieved in
less than 30 min.
3.3. Generated response models from BBD
The generated pattern of second-order regression for CTSP (Y
1
), as
obtained from the experiment are exhibited in Eqs. (5)(7), while,
that of COD removal (Y
2
) are displayed in Eqs. (8)(10).
Y1LCSC0¼54:47 þ4:27X122:08X2þ4:31X3þ0:93X1X2
þ2:02X1X3þ0:44X2X33:29X12þ12:17X22
þ3:70X32ð5Þ
Table 2
ANOVA findings for three bio-coagulants for CTSP and COD removal.
Response Source SS DF MS F-value Prob>FR
2
Adj R
2
CTSP
LCSC
0
Model 4941.24 9 549.03 70.67 <0.0001 0.9871 0.9751
Residual 54.38 7 7.77
Lack of fit 35.57 3 11.86 2.52 0.1966
Pure error 8.82 4 0.70
LCSC
1
Model 3828.74 9 425.42 41.71 <0.0001 0.9817 0.9582
Residual 71.29 7 10.20
Lack of fit 52.61 3 17.54 3.74 0.1177
Pure error 8.78 4 0.69
LCSC
2
Model 3037.62 9 337.51 67.56 <0.0001 0.9886 0.9740
Residual 34.97 7 5.00
Lack of fit 26.84 3 8.95 4.40 0.0931
Pure error 2.13 4 0.03
COD
LCSC
0
Model 372.21 9 41.36 94.65 <0.0001 0.9920 0.9824
Residual 3.06 7 0.44
Lack of fit 2.42 3 0.81 5.10 0.0747
Pure error 0.63 4 0.016
LCSC
1
Model 2419.35 9 264.82 97.78 <0.0001 0.9859 0.9679
Residual 19.25 7 2.75
Lack of fit 6.45 3 2.15 0.67 0.3128
Pure error 2.80 4 0.20
LCSC
2
Model 2152.09 9 239.12 54.56 <0.0001 0.9921 0.9820
Residual 30.68 7 4.38
Lack of fit 21.05 3 7.02 2.92 0.1640
Pure error 3.63 4 0.41
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 7
Fig. 4. Pareto CTSP and COD graph for LCSC
2.
Fig. 5. Presentation of 3D-surface plots for CTSP elimination using (a) LCSC
0
(b) LCSC
1
, (c) LCSC
2.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
8O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
Y1LCSC1¼69:85 þ4:05X120:22X2þ2:07X30:087X1X2
þ4:72X1X3þ3:55X2X3þ4:60X12þ3:30X22
þ4:70X32ð6Þ
Y1LCSC2¼67:14 þ1:30X116:72X2þ0:83X31:44X1X2
þ2:66X1X3þ1:30X2X3þ6:30X12þ9:28X22
þ5:71X32ð7Þ
Y2LCSC0¼89:33 þ1:63X16:10X2þ0:28X3þ0:73X1X2
þ1:19X1X3þ0:76X2X30:24X12þ2:94X22
1:33X32ð8Þ
Y2LCSC1¼74:60 þþ1:76X115:49X2þ2:18X3þ1:22X1X2
þ0:44X1X3þ2:09X2X3þ5:97X12þ1:70X22
þ7:20X32ð9Þ
Y2LCSC2¼73:75 þ1:94X114:69X2þ1:33X31:27X1X2
0:23X1X3þ2:18X2X3þ5:60X12þ4:23X22
þ5:16X32ð10Þ
The constant associated with individual factors indicate the influ-
ence of that particular variable, while that connected to multiple fac-
tors and squared factors indicate interaction among variables and the
quadratic effect. A plus sign before the constant shows agreement,
whereas, a minus sign is an indication of disagreement [10,47].
Sequel to getting rid of non-relevant interaction terms for both CTSP
and COD elimination, Eqs. (11)(16) have been produced.
Fig. 6. Presentation of 3D-surface for COD elimination using (a) LCSC
0
(b) LCSC
1
, (c) LCSC
2.
Table 3
Comparison of ANN and RSM predictive strength.
Bio-coagulant/Responses ANN RSM
MSE R
2
MSE R
2
LCSC
0
- CTSP 1.4341e-5 0.999999 2.7192 0.9871
COD 1.80499e-7 0.999999 0.2687 0.9920
LCSC
1
- CTSP 4.43033 0.979379 2.0563 0.9817
COD 1.41284 0.990015 1.4215 0.9679
LCSC
2
- CTSP 5.71924e-7 0.999999 1.4494 0.9886
COD 4.57391e-8 0.999999 0.9249 0.9921
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 9
Y1LCSC0¼54:47 þ4:27X122:08X2þ4:31X33:29X12
þ12:17X22þ3:70X32ð11Þ
Y1LCSC1¼69:85 þ4:05X120:22X2þ4:72X1X3þ4:60X12
þ4:70X32ð12Þ
Y1LCSC2¼67:14 16:72X2þ2:66X1X3þ6:30X12þ9:28X22
þ5:71X32ð13Þ
Y2LCSC0¼89:33 þ1:63X16:10X2þ1:19X1X3þ2:94X22
1:33X32ð14Þ
Y2LCSC1¼74:60 þþ1:76X115:49X2þ2:18X3þ2:09X2X3
þ5:97X12þ7:20X32ð15Þ
Y2LCSC2¼73:75 þ1:94X114:69X2þ5:60X12þ4:23X22
þ5:16X32ð16Þ
3.3.1. Suitability of the pattern
Variance analysis (ANOVA), presented in Table 2, was evaluated to
obtain the suitability of the pattern. Table 2 showed the R
2
and adjR
2
fo LCSC
0
r,LCSC
1
and LCSC
2
, respectively, in terms of CTSP and COD
removal for the choosen model. From the table, the recorded R
2
implies that the independent variables explained the variations in
the respective responses up to 90% for both CTSP and COD using vari-
ous bio-coagulants. The high R
2
values is an indication of the model’s
good fit to the respective responses [24] The observation shows that
the information obtained from experiment were reasonably in agree-
ment due to the nearness of R
2
to the adjR
2
[24,47].
3.3.2. Analysis of processes
The pronounced effect shown by the dosage and pH could be due
to the initial concentration of the reactant and solution pH effect on
Fig. 7. Training, testing, validation, and overall plots of the CTSP removal using ANN model (a) LCSC
0
(b) LCSC
1
, (c) LCSC
2.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
10 O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
the CF led reactions. Since the amino groups in proteins are normally
cationic, the negatively charged dye particles are easily destabilized.
It was also discovered that the dye particles maintained their anionic
charge at high pH, enhancing the cationic coagulant's ability to com-
bine with particles. This is consistent with the documented reports
[1,51,56,57,58]. Due to the high f-values obtained for both responses
in the case of LCSC
0
LCSC
1
and LCSC
2
, the quadratic regression model
was significant. Similarly, the models’p-values which provide indica-
tions of the relevance of the pattern as it relates to f-value were
<0.05. Hence, with a confidence level of 95% the pattern was statisti-
cally relevant.
Despite the existence of non-significant terms, the magnitude of
the lack of fitf-values (Table 2) implies that the lack of fits is not sig-
nificant relative to the pure errors. It also, revealed that there are
19.66 %, 11.77 % and 9.31 %, respectively, chance that a lack of fitf-
values this large could be due to noise. Similarly, for COD removal, it
also implies that there are 7.47%, 31.28% and 16.40% chance that the
lack of fit could be due to noise. The lack of fit values were good,
hence, the model fit.
3.3.3. The studied impact of variables on the elimination of % CTSP and
%COD
The influence of individual variables and their interactions, the
Pareto study was carried out based on Eq. (17) [47,59] and Fig. 4 dis-
played the graph for the Sutherland process bio-coagulant.
Pi¼bi2
Pbi2
!
X100 i6¼ 0ðÞ ð17Þ
Fig. 4a and b, represented the Pareto plots for %CTSP and %COD
elimination. Fig. 4a representing %CTSP removal indicated pH as the
most dominant factor with linear(X
2
) effect of 99.15% while the qua-
dratic pH effect X
22
was 54.33%. Even though the dosage effects were
low, the interaction between the dosage and stirring time (X
1
X
3
)
recorded high effects of 65.24%. Similarly, for %COD removal (Fig. 4b),
the pH linear effect(X
2
) of 97.5% and the quadratic dosage effect X
12
of 41.39% was observed.
Fig. 5 showed the 3D-surface plots of the %CTSP removal based on
the BB-design. Fig. 5a, indicating LCSC
0
revealed reasonable curvature
Fig. 7. Continued.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 11
which is consistent with quadratic models and an indication of high
interactive effect. Similar curvature was observed for LCSC
1
(Fig. 5b)
and LCSC
2
(Fig. 5c), showing the interdependence of these three fac-
tors [60]. The orange shade indicates the area of high interaction. The
LCSC
1
and LCSC
2
showed better performance with >99% removal effi-
ciency.
Fig. 6 indicates the 3D-surface plots of the %COD removal using
the three bio-coagulants. Fig. 6ac indicated similar result as CTSP
removal plots. The orange shades were more pronounced in the X
1
vs
X
2
, and X
2
vs X
3
plots. However, like in CTSP removal, the degree of
pronouncement of the shades differs. Similarly, the LCSC
1
and LCSC
2
recorded removal efficiency >99% indicating them as better method
for processing the biomass
3.4. The building of ANN and training
For ANN, the same data was added for three independent varia-
bles used for RSM. The responses (outputs) are also, the same. For the
ANN method, selecting the number of hidden layers of neurons is of
critical interest. Training data errors are decreased as hidden neurons
are increased, but the complexity of the network is increased. We
evaluated a network comprising two to fifteen neurons. The best gen-
eralisation and model complexity were suggested by a ten-neuron
hidden layer.
The architecture of the network is presented as 3-6-4-1, a four-
layer feed-forward backdrop network consisting of three input neu-
rons, ten hidden layers, and one output layer neuron [35]. More than
ten training functions were tested, such as Levenberg-Marquardt BP,
Scale conjugate gradient BP, Gradient descent with adaptive learning
rate BP, etc., and the best result was found through Levenberg-Mar-
quardt BP. For the hidden neurons, the log-sigmoid transfer function
(logsig) was chosen, while the output layer neuron was selected for
the linear transfer function (purelin).
Training was then conducted separately to achieve the best-
trained network. After 11 iterations for both CTSP and COD removal
using LCSC
0
, the best performance was achieved. The best result was
achieved for LCSC
1
after 15 and 13 iterations for CTSP and COD
removal, respectively. However, the best trained network after 12
and 11 iterations, respectively, was achieved for the removal of CTSP
and COD using LCSC
2
. For CTSP and COD elimination, the R
2
reported
was >0.9. Table 3 summarized the reliability metrics. The perfor-
mance and regression plots are shown in Figs. 7 and 8.
3.4.1. RSM and ANN forecasting ability compared
For the DoE, ANN and RSM model were compared. For both mod-
els, the data obtained from DoE has been trained. Based on parame-
ters such as MSE and R
2
, the predictive capacity of ANN and RSM has
been examined. These parameter values are provided in Table 3. For
all the output, the ANN model was fitted with higher precision to the
experimental data as shown in Figs. 7 and 8for the two targets. As
evidenced by MSE, the RSM model demonstrates considerable devia-
tion while ANN exhibits a substantially higher generalization capabil-
ity. The greater predictive accuracy of ANN evidenced by higher R
2
can be due to its ability to approximate all types of structures.
Fig. 8. Training, testing, validation, and overall plots of the COD removal using ANN model(a) LCSC
0
(b) LCSC
1
, (c) LCSC
2.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
12 O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
4. Conclusion
Three (3) separate bio-coagulants have been successfully
extracted from LCS (LCSC
0
,LCSC
1
,andLCSC
2
). In the CF treatment
of DPW, the synthesized bio-coagulants behaved optimally, how-
ever, the extract from the Sutherland method (LCSC
2
) performed
better than the extract from the other two methods. RSM and
ANN methods were applied for modeling the prediction of effi-
cient removal of CTSP and COD from DPW using the synthesized
bio-coagulants. The effect of the operating factors was investi-
gated by response surface analysis.
Although RSM was well known for insight analysis of variable
interactive effects, the ANN approach provided a better model
describing the process. This is because ANN is capable of handling all
manner of non-linear process, but, RSM is restricted to quadratic pol-
ynomials.
Comparative analysis on the predictive strength using R
2
and MSE
applied to test the suitability of the RSM equation and ANN model
established the strong potential of the bio-coagulants for decontami-
nation of DPW with an extract from the Sutherland method giving
the best performance. This is evidenced by the following indicators:
R
2
>0.9999 and MSE of 5.72e-7 and 4.57e-8 for CTSP and COD
removal, respectively.
Declaration of Competing Interest
The authors declare that they have no known competing financial
interests or personal relationships that could have appeared to influ-
ence the work reported in this paper.
Acknowledgment
V.C. Anadebe is grateful to CSIR, India and TWAS Italy for the Post-
graduate Fellowship (Award No. 22/FF/CSIR-TWAS/2019) to purse
research programme in CSIR-CECRI, India. In addition, Alex Ekwueme
Federal University Ndufu-Alike Ebonyi State, Nigeria, is acknowl-
edged for the Research Leave to visit CECRI, India
References
[1] Karam A, Bakhoum ES, Zaher K. Coagulation/flocculation process for textile mill
effluent treatment: experimental and numerical perspectives. Int J Sustain Eng
2020;00:1–13. doi: 10.1080/19397038.2020.1842547.
[2] Chu Z, Chen K, Xiao C, Ji D, Ling H, Li M, et al. Improving pressure durability and
fractionation property via reinforced PES loose nanofiltration hollow fiber mem-
branes for textile wastewater treatment. J Taiwan Inst Chem Eng 2020;108:71–
81. doi: 10.1016/j.jtice.2019.12.009.
Fig. 8. Continued.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 13
[3] Naraian R, Kumari S, Gautam RL. Biodecolorization of brilliant green carpet indus-
try dye using three distinct Pleurotus spp. Environ Sustain 2018;1:141–8. doi:
10.1007/s42398-018-0012-4.
[4] Samsami S, Mohamadi M, Sarrafzadeh MH, Rene ER, Firoozbahr M. Recent advan-
ces in the treatment of dye-containing wastewater from textile industries: Over-
view and perspectives. Process Saf Environ Prot 2020;143:138–63. doi: 10.1016/j.
psep.2020.05.034.
[5] Thirunavukkarasu A, Nithya R. Adsorption of acid orange 7 using green synthe-
sized CaO /CeO
2
composite: an insight into kinetics, equilibrium, thermodynam-
ics, mass transfer and statistical models. J Taiwan Inst Chem Eng 2020;111:44–
62. doi: 10.1016/j.jtice.2020.04.007.
[6] Anastopoulos I, Pashalidis I. Environmental applications of Luffa cylindrica-based
adsorbents. J Mol Liq 2020;319:114127. doi: 10.1016/j.molliq.2020.114127.
[7] Azmi NH, Ali UFM, Muhammad Ridwan F, Isa KM, Zulkurnai NZ, Aroua MK. Prepa-
ration of activated carbon using sea mango (Cerbera odollam) with microwave-
assisted technique for the removal of methyl orange from textile wastewater.
Desalin Water Treat 2016;57:29143–52. doi: 10.1080/19443994.2016.1168134.
[8] Karam A, Zaher K, Mahmoud AS. Comparative studies of using nano zerovalent
iron, activated carbon, and green synthesized nano zerovalent iron for textile
wastewater color removal using artificial intelligence, regression analysis,
adsorption isotherm, and kinetic studies. Air Soil Water Res 2020;13. doi:
10.1177/1178622120908273.
[9] da Rocha HD, Reis ES, Ratkovski GP, da Silva RJ, Gorza FDS, Pedro GC, et al. Use of
PMMA/(rice husk ash)/polypyrrole membranes for the removal of dyes and heavy
metal ions. J Taiwan Inst Chem Eng 2020;110:8–20. doi: 10.1016/j.jtice.2020.03.003.
[10] de Souza MTF, de Almeida CA, Ambrosio E, Santos LB, Freitas TKF de S, Manholer
DD, et al. Extraction and use of Cereus peruvianus cactus mucilage in the treat-
ment of textile effluents. J Taiwan Inst Chem Eng 2016;67:174–83. doi: 10.1016/j.
jtice.2016.07.009.
[11] Hadi SM, Al-Mashhadani MKH, Eisa MY. Optimization of dye adsorption process
for Albizia lebbeck pods as a biomass using central composite rotatable design
model. Chem Ind Chem Eng Q 2019;25:39–46. doi: 10.2298/CICEQ180210021H.
[12] Konstantinovic SS, Kodric MG, Nicic R, Djordjevic DM. Decolorization of model
wastewater by adsorbent obtained from waste hemp fibers. Chem Ind Chem Eng
Q 2019;25:11–9. doi: 10.2298/CICEQ170720013K.
[13] Xu MY, Jiang HL, Xie ZW, Li ZT, Xu D, He FA. Highly efficient selective adsorption
of anionic dyes by modified b-cyclodextrin polymers. J Taiwan Inst Chem Eng
2020;108:114–28. doi: 10.1016/j.jtice.2020.01.005.
[14] Olugbenga SB, Bukola ML, Olamide JA, Vanain E. Scavenging rhodamine B dye
using Moringa olifera seed pod. Chem Speciat Bioavailab 2017;29:120–34.
[15] Irfan M, Butt T, Imtiaz N, Abbas N, Ahmad R, Shafique A. The removal of COD, TSS
and colour of black liquor by coagulation flocculation process at optimized pH,
settling and dosing rate. Arab J Chem 2017;10:S2307–18. doi: 10.1016/j.
arabjc.2013.08.007.
[16] Mdlovu NV, Lin KS, Chen ZW, Liu YJ, Mdlovu NB. Treatment of simulated chro-
mium-contaminated wastewater using polyethylenimine-modified zero-valent
iron nanoparticles. J Taiwan Inst Chem Eng 2020;108:92–101. doi: 10.1016/j.
jtice.2019.12.011.
[17] Mohamad Yusof MS, Othman MHD, Abdul Wahab R, Abu Samah R, Kurniawan TA,
Mustafa A, et al. Effects of pre and post-ozonation on POFA hollow fibre ceramic
adsorptive membrane for arsenic removal in water. J Taiwan Inst Chem Eng
2020;110:100–11. doi: 10.1016/j.jtice.2020.02.014.
[18] Nguyen CH, Tran HN, Fu CC, Lu YT, Juang RS. Roles of adsorption and photocataly-
sis in removing organic pollutants from water by activated carbonsupported
titania composites: kinetic aspects. J Taiwan Inst Chem Eng 2020;109:51–61. doi:
10.1016/j.jtice.2020.02.019.
[19] Badawi AK, Bakhoum ES, Zaher K. Sustainable evaluation of using nano zero-val-
ent iron and activated carbon for real textile effluent remediation. Arab J Sci Eng
2021. doi: 10.1007/s13369-021-05349-5.
[20] Badawi AK, Zaher K. Hybrid treatment system for real textile wastewater remedi-
ation based on coagulation/flocculation, adsorption and filtration processes: per-
formance and economic evaluation. J Water Process Eng 2021;40:101963. doi:
10.1016/j.jwpe.2021.101963.
[21] Samsami S, Mohamadi M, Sarrafzadeh MH, Rene ER, Firoozbahr M. Recent advan-
ces in the treatment of dye-containing wastewater from textile industries: over-
view and perspectives. Process Saf Environ Prot 2020;143:138–63. doi: 10.1016/j.
psep.2020.05.034.
[22] Okolo BI, Nnaji PC, Onukwuli OD. Nephelometric approach to study coagulation-
flocculation of brewery effluent medium using Detarium microcarpum seed pow-
der by response surface methodology. J Environ Chem Eng 2016;4:992–1001. doi:
10.1016/j.jece.2015.12.037.
[23] ¸Cırak M, Atay HY. Coagulation/flocculation process for marble processing plant
effluent: modelling and optimization through response surface methodology.
Asia Pac J Chem Eng 2019;14:1–13. doi: 10.1002/apj.2371.
[24] Howe KJ, Hand DW, Crittenden JC, Trussell RR, Tchobanoglous G. Principles of
Water Treatment. John Wiley and Sons, Inc; 2012. p. 139–42.
[25] Al-sameraiy M. A new approach using coagulation rate constant for evaluation of
turbidity removal. Appl Water Sci 2017;7:1439–48. doi: 10.1007/s13201-015-
0341-8.
[26] Wong PW, Teng TT, Nik Norulaini NAR. Efficiency of the coagulation-flocculation
method for the treatment of dye mixtures containing disperse and reactive dye.
Water Qual Res J Canada 2007;42:54–62. doi: 10.2166/wqrj.2007.008.
[27] Okolo BI, Nnaji PC, Oke EO, Adekunle KF, Ume CS, Onukwuli OD. Optimizing bio-
coagulants for brewery wastewater treatment using response surface methodol-
ogy. Niger J Technol 2018;36:1104. doi: 10.4314/njt.v36i4.16.
[28] Aniagor CO, Menkiti MC. Kinetics and mechanistic description of adsorptive
uptake of crystal violet dye by lignified elephant grass complexed isolate. J Envi-
ron Chem Eng 2018;6:2105–18. doi: 10.1016/j.jece.2018.01.070.
[29] Ani JU, Nnaji NJN, Onukwuli OD, Okoye COB. Nephelometric and functional
parameters response of coagulation for the purification of industrial wastewater
using Detarium microcarpum. J Hazard Mater 2012;243:59–66. doi: 10.1016/j.
jhazmat.2012.09.069.
[30] Maroneze MM, Zepka LQ, Vieira JG, Queiroz MI, Jacob-Lopes E. A tecnologia de
remo¸c~
ao de f
osforo: gerenciamento do elemento em resíduos industriais. Rev
Ambient e Agua 2014;9:445–58. doi: 10.4136/1980-993X.
[31] Nnaji P, Anadebe C, Onukwuli OD. Application of experimental design methodol-
ogy to optimize dye removal by mucuna sloanei induced coagulation of dye-
based wastewater. Desalin Water Treat 2020;198:396–406. doi: 10.5004/
dwt.2020.26017.
[32] Joshi S, Bajpai S, Jana S. Application of ANN and RSM on fluoride removal using
chemically activated D. sissoo sawdust. Environ Sci Pollut Res 2020;27:17717–29.
doi: 10.1007/s11356-020-08153-0.
[33] Desai KM, Survase SA, Saudagar PS, Lele SS, Singhal RS. Comparison of artificial
neural network (ANN) and response surface methodology (RSM) in fermentation
media optimization: Case study of fermentative production of scleroglucan. Bio-
chem Eng J 2008;41:266–73. doi: 10.1016/j.bej.2008.05.009.
[34] YoussefiS, Emam-Djomeh Z, Mousavi SM. Comparison of artificial neural network
(ANN) and response surface methodology (RSM) in the prediction of quality
parameters of spray-dried pomegranate juice. Dry Technol 2009;27:910–7. doi:
10.1080/07373930902988247.
[35] Khayet M, Cojocaru C, Essalhi M. Artificial neural network modeling and response
surface methodology of desalination by reverse osmosis. J Memb Sci
2011;368:202–14. doi: 10.1016/j.memsci.2010.11.030.
[36] Hammoudi A, Moussaceb K, Belebchouche C, Dahmoune F. Comparison of artifi-
cial neural network (ANN) and response surface methodology (RSM) prediction
in compressive strength of recycled concrete aggregates. Constr Build Mater
2019;209:425–36. doi: 10.1016/j.conbuildmat.2019.03.119.
[37] Elsayed K, Lacor C. Modeling, analysis and optimization of aircyclones using artifi-
cial neural network, response surface methodology and CFD simulation
approaches. Powder Technol 2011;212:115–33. doi: 10.1016/j.
powtec.2011.05.002.
[38] Gupta KN, Kumar R. Fixed bed utilization for the isolation of xylene vapor: Kinet-
ics and optimization using response surface methodology and artificial neural
network. Environ Eng Res 2020;26. doi: 10.4491/eer.2020.105.
[39] Anadebe VC, Onukwuli OD, Abeng FE, Okafor NA, Ezeugo JO, Okoye CC. Electro-
chemical-kinetics, MD-simulation and multi-input single-output (MISO) model-
ing using adaptive neuro-fuzzy inference system (ANFIS) prediction for
dexamethasone drug as eco-friendly corrosion inhibitor for mild steel in 2 M HCl
electrolyte. J Taiwan Inst Chem Eng 2020;115:251–65. doi: 10.1016/j.
jtice.2020.10.004.
[40] Zhao Z, Sun W, Ray MB, Ray AK, Huang T, Chen J. Optimization and modeling of
coagulation-flocculation to remove algae and organic matter from surface water
by response surface methodology. Front Environ Sci Eng 2019;13. doi: 10.1007/
s11783-019-1159-7.
[41] Zin KM, Effendi Halmi MI, Abd Gani SS, Zaidan UH, Samsuri AW, Abd Shukor MY.
Microbial decolorization of triazo dye, direct blue 71: an optimization approach
using response surface methodology (RSM) and artificial neural network (ANN).
Biomed Res Int 2020;2020. doi: 10.1155/2020/2734135.
[42] AOAC. Official methods of analysis. 16th Edition Association of Official Analytical
Chemist, Gaitherburg; 2005.
[43] Sutherland J.P., Process for preparing coagulants for water treatment. United
States Patent, 6890565
[44] Menkiti M, Ezemagu I, Okolo B. Perikinetics and sludge study for the decontami-
nation of petroleum produced water (PW) using novel mucuna seed extract Colli-
sion factor for Brownian transport. Pet Sci 2016;13:328–39. doi: 10.1007/s12182-
016-0082-9.
[45] Menkiti MC, Okoani AO, Ejimofor MI. Adsorptive study of coagulation treatment
of paint wastewater using novel Brachystegia eurycoma extract. Appl Water Sci
2018;8:1–15. doi: 10.1007/s13201-018-0836-1.
[46] AWWA. APHA, WEF. 22ndEdition New York: Standard Method for the Examina-
tion of Water and Wastewater; 2012.
[47] Kim SC. Application of response surface method as an experimental design to
optimize coagulation-flocculation process for pre-treating paper wastewater. J
Ind Eng Chem 2016;38:93–102. doi: 10.1016/j.jiec.2016.04.010.
[48] Abitogun AS, Ashogbon AO, Polytechnic RG, State O. Nutritional assessment and
chemical composition of raw and defatted Luffa cylindrica seed flour. Ethnobot
Leafl2010;14:225–35.
[49] AJ A, OM J. Effect of two extractants on the chemical composition of the defatted
seed of luffa cylindrica. Int J Innov Sci Eng Technol 2017;4:356–68.
[50] Salem RH. Functional characterization of luffa (Luffa cylindrica) seeds powder and
their utilization to improve stabilized emulsions. Middle East J Appl Sci
2017:613–25.
[51] Dalvand A, Gholibegloo E, Ganjali MR, Golchinpoor N, Khazaei M, Kamani H, et al.
Comparison of Moringa stenopetala seed extract as a clean coagulant with alum
and moringa stenopetala-alum hybrid coagulant to remove direct dye from tex-
tile wastewater. Environ Sci Pollut Res 2016;23:16396–405. doi: 10.1007/
s11356-016-6708-z.
[52] Adewuyi A, Vargas F. Underutilized Luffa cylindrica sponge : a local bio-adsorbent
for the removal of Pb (II) pollutant from water system. Beni-Suef Univ J Basic Appl
Sci 2017;6:118–26. doi: 10.1016/j.bjbas.2017.02.001.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
14 O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115
[53] Oboh I, Aluyor E, Audu T. Kinetic modelling for zinc (II) ions biosorption onto Luffa
cylindrica. AIP Conf Proc 2015;1653. doi: 10.1063/1.4914270.
[54] Ahmad R, Haseeb S. Competitive adsorption of Cu
2+
and Ni
2+
on Luffa acutangula
modified Tetraethoxysilane (LAP-TS) from the aqueous solution: thermodynamic
and isotherm studies. Groundw Sustain Dev 2015;1:146–54. doi: 10.1016/j.
gsd.2016.03.001.
[55] Zhao B, Xiao W, Shang Y, Zhu H, Han R. Adsorption of light green anionic dye
using cationic surfactant-modified peanut husk in batch mode. Arab J Chem
2017;10:S3595–602. doi: 10.1016/j.arabjc.2014.03.010.
[56] Baharlouei A, Jalilnejad E, Sirousazar M . Fixed-bed column pe rformanc e of
methylene blue biosorption by Luffa cylindrica: statistical and mathematical
modeling. Chem Eng Commun 2018;205:1537–54. doi: 10.1080/
00986445.2018.1460364.
[57] Sangal VK, Kumar V, Mishra IM. Optimization of structural and operational varia-
bles for the energy efficiency of a divided wall distillation column. Comput Chem
Eng 2012;40:33–40. doi: 10.1016/j.compchemeng.2012.01.015.
[58] Zhang W, Wei Q, Xiao J, Liu Y, Yan C, Liu J, et al. The key factors and removal
mechanisms of sulfadimethoxazole and oxytetracycline by coagulation. Environ
Sci Pollut Res 2020;27:16167–76. doi: 10.1007/s11356-019-06884-3.
[59] Ahmadi M, Ghanbari F, Madihi-bidgoli S. Photoperoxi-coagulation using activated
carbon fiber cathode as an efficient method for benzotriazole removal from aque-
ous solutions: modeling, optimization and mechanism. J Photochem Photobiol A
Chem 2016;322323:85–94. doi: 10.1016/j.jphotochem.2016.02.025.
[60] Imen F, Lamia K, Asma T, Neacute ji G, Radhouane G. Optimization of coagulation-
flocculation process for printing ink industrial wastewater treatment using response
surface methodology. Afr J Biotechnol 2013;12:4819–26. doi: 10.5897/ajb12.1900.
ARTICLE IN PRESS
JID: JTICE [m5G;June 25, 2021;2:30]
Please cite this article as: O.D. Onukwuli et al., Dual-purpose optimization of dye-polluted wastewater decontamination using bio-coagulants
from multiple processing techniques via neural intelligence algorithm and response surface methodology, Journal of the Taiwan Institute of
Chemical Engineers (2021), https://doi.org/10.1016/j.jtice.2021.06.030
O.D. Onukwuli et al. / Journal of the Taiwan Institute of Chemical Engineers 00 (2021) 115 15