Download as pdf or txt
Download as pdf or txt
You are on page 1of 17

BioEnergy Research (2023) 16:1667–1682

https://doi.org/10.1007/s12155-022-10517-y

Organic Acid Pretreatments and Optimization Techniques for Mixed


Vegetable Waste Biomass Conversion into Biofuel Production
Babu Dharmalingam1 · Prapakorn Tantayotai2 · Elizabeth Jayex Panakkal1 · Kraipat Cheenkachorn3 ·
Suchata Kirdponpattara3 · Marttin Paulraj Gundupalli1 · Yu‑Shen Cheng4 · Malinee Sriariyanun1

Received: 30 June 2022 / Accepted: 20 September 2022 / Published online: 26 September 2022
© The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2022

Abstract
Mixed vegetable and food wastes left from consumption and processing were important uncontrolled sources of greenhouse
gas emissions. Biofuel production from mixed vegetable waste is ineffective due to the recalcitrant structure of lignocellu-
lose. In this work, an artificial neural network (ANN) and response surface methodology (RSM) have been used to optimize
the process parameters for the pretreatment of mixed vegetable waste by using organic acid (acetic, citric, oxalic acid) and
inorganic acid (hydrochloric acid). The study revealed that the maximum reducing sugars yield of 314.1 mg ­g−1 biomass
was obtained from acetic acid-pretreated biomass at a time of 68.32 min, a temperature of 140 °C, and an acid concentration
of 10% w/v, which was 1.66 fold higher than the HCl-pretreated sample. The analysis of biomass composition and FTIR
suggested that both inorganic acid and organic acid pretreatment removed hemicellulose and lignin from biomass, and subse-
quently enhanced the cellulase accessibility to biomass during enzymatic saccharification. It was observed that organic acid
pretreatment is more effective to remove lignin compared to inorganic acid. The pretreated samples were converted to ethanol,
and the maximum yield of 7.60 g L ­ −1 was obtained for the oxalic acid-pretreated sample, which was 2.82 fold higher than
the HCl-pretreated biomass. Also, the maximum biogas yield of 110.72 mL g-VS−1 was obtained for citric acid-pretreated
biomass, which was fivefold higher than the untreated biomass. Hence, organic acid pretreatment is highly recommended
for bioethanol and biogas production from mixed vegetable wastes.

Keywords Bioethanol · Biogas · Biorefinery · Organic acid pretreatment · Mixed vegetable · Optimization

Introduction

Global population, industry development, rapid economic


Babu Dharmalingam and Prapakorn Tantayotai contributed equally growth, urbanization, and energy demand have resulted in
to this work. the rapid depletion of fossil resources. Global warming and
* Malinee Sriariyanun environmental pollution are the major concerns due to the
malinee.s@tggs.kmutnb.ac.th continuous utilization of fossil fuels in automobile indus-
tries, agriculture, and power machinery [1]. Bioenergy from
1
Biorefinery and Process Automation Engineering Center, lignocellulose biomass is considered a suitable and potential
Department of Chemical and Process Engineering, The
Sirindhorn International Thai-German Graduate School replacement for fossil fuels due to lower greenhouse gas
of Engineering, King Mongkut’s University of Technology (GHGs) emissions. Agriculture, forest, and industrial sectors
North Bangkok, Bangkok, Thailand are the main contributors to the availability of lignocellulose
2
Department of Microbiology, Faculty of Science, biomass, and the most promising feedstocks are agricultural
Srinakharinwirot University, Bangkok, Thailand waste and forest residues [2]. In recent years, lignocellulose
3
Department of Chemical Engineering, Faculty biomass has emerged as a potential source to produce biofu-
of Engineering, King Mongkut’s University of Technology els and biochemicals due to its abundance and low cost. Lig-
North Bangkok, Bangkok, Thailand nocellulose materials are composed of three polymer layers,
4
Department of Chemical and Materials Engineering, including cellulose, hemicellulose, and lignin [3]. Commer-
National Yunlin University of Science and Technology, cial production of biofuels and biochemical products from
Douliu, Yunlin, Taiwan

13
Vol.:(0123456789)
1668 BioEnergy Research (2023) 16:1667–1682

lignocellulose materials requires a high cellulose content of rice straw water extractives resulted in a significant drop
and a low lignin content because lignin is composed of com- in biogas yield and a shift of the biogas composition from
plex aromatic polymers that hinder the process efficiency 74% to less than 1% methane. By increasing the ethanol
during biofuel and biochemical production [4]. extractives concentration of rice straw, the biogas yield was
The susceptibility of cellulose to hydrolysis is a key com- increased from 167 to 524 mL ­g−1 [14]. Bittencourt et al.
ponent that is limited due to the hindrance caused by lignin [15] investigated the fractionation of sugarcane bagasse by
and hemicellulose that surround the cellulose microfibrils in hydrothermal and alkaline pretreatments for the production
the plant cell wall [5]. Therefore, pretreatment is important of bioethanol and biogas. The enzymatic hydrolysate was
in improving the process efficiency by solubilizing hemicel- fermented with Saccharomyces cerevisiae, yielding 190.8
lulose and lignin. Post-pretreatment, lignocellulose biomass L ethanol ton-SB−1. Anaerobic digestion of liquid streams
is subjected to saccharification and fermentation to produce from the pretreatment processes produced 27.46 ­NLCH4
biofuel. The pretreatment efficiency is evaluated through kg-SB−1 of biogas. However, furfural and formic acid are
glucose and ethanol yield during the saccharification and formed during chemical pretreatment due to the degrada-
fermentation process [6]. Different conventional pretreat- tion of xylan monomers caused by severe pretreatment
ment techniques are available, such as acid hydrolysis, steam conditions, such as high temperature and longer reaction
explosion, ammonia fiber expansion, wet alkaline oxidation, time [16]. These sugar degradation compounds (furfural,
and hot water pretreatment which exist to liberate cellulose 5-hydroxymethylfurfural, formic acid, acetic acid, and lev-
from the lignin and hemicellulose [7]. Hydrochloric acid, ulinic acid) inhibit the activity and metabolism of enzymes
sulfuric acid, nitric acid, and acetic acid are the most often and microorganisms. Therefore, the selection of the pre-
used chemicals in the pretreatment of lignocellulose bio- treatment method and the optimization of the pretreatment
mass. However, utilizing these acids can lead to equipment conditions are important steps to maximize the enzymatic
corrosion and produce chemical inhibitors as byproducts [8]. saccharification and fermentation efficiency [10].
Physical pretreatment has several disadvantages, including Organic acid pretreatment has been proposed as an alter-
high processing, operating cost, and instrument corrosion. native to conventional acid pretreatment methods due to
Furthermore, the use of alkaline pretreatment can cause lower sugar degradation and less formation of inhibitory
condensation and redistribution of lignin, making the cel- byproducts. Organic acids are non-toxic and environmentally
lulose more resistant than the native cellulose [9]. In addi- friendly compared to mineral acids, such as sulphuric acid
tion, physicochemical pretreatment requires high tempera- ­(H2SO4) and hydrochloric acid (HCl) [17]. Different organic
ture, specialized reactor, and expensive instruments, whereas acids have been explored for various lignocellulose biomass.
organic acid is a weak acid and produces less amount of However, fewer studies exist on the pretreatment of vegeta-
chemical inhibitors as compared to other pretreatments [10]. ble waste using organic acids to produce ethanol and biogas
The effectiveness of pretreatment is dependent upon the con- via fermentation and anaerobic digestion (AD) process.
ditions, such as type of biomass, biomass loading, tempera- Vegetable waste, as a type of food waste, is obtained during
ture, reaction time, catalyst concentration, and mixing [11]. processing and cutting before handling by consumers. Food
Among various methods, chemical pretreatment has been waste is one of the major sources to generate the uncon-
demonstrated and selected for different types of biomasses trolled released greenhouse gas, about 10%, in the form of
due to its simplicity, cost-effectiveness, and efficiency. methane from landfills (Food and Agriculture Organization
Ahmad et al. [12] investigated the effect of sulfuric acid- of the UN). In 2020, it was estimated that only in the USA,
assisted hydrothermal pretreatment on methane produc- 170 million metric tons of carbon dioxide (­ CO2) equivalent
tion from sugarcane bagasse using the RSM approach. The to GHG emissions were produced from food waste (The US
experimental studies revealed that the lowest solid yield Environmental Protection Agency (EPA) report). Therefore,
of 13% was reported at 200 °C, 19 min, and 3% ­H2SO4, it is important to develop a controlled process to convert
whereas maximum methane recovery of 160 NmL (g TVS)−1 vegetable waste to value-added products, such as biogas and
was recorded at 180 °C, 12 min, and 2% ­H2SO4 suggesting bioethanol to reduce the devastation of uncontrolled GHG
the importance of pretreatment optimization. Thulluri et al. emissions to the environment.
[13] examined the effect of acid pretreatment on the cellu- Several studies have focused on employing various
lose of vegetable waste and enzymatic hydrolysis efficiency. approaches to produce biofuel production. However, these
The study revealed that acid pretreatment enhanced the approaches have several disadvantages, including processing
sugar, phenol, and cellulose yields. Another study reported complexity, low accuracy, and higher error rate [1]. The arti-
the enhancement of biogas and bioethanol production from ficial neural network approach has been used in the optimi-
rice straw extractives using water and ethanol. The study zation of biofuel production worldwide due to its higher pre-
observed that the addition of rice straw water extractives cision and low error rate. Pomeroy et al. [18] reviewed the
improved the ethanol yield to 95%, whereas the addition importance of the ANN approach on chemical production

13
BioEnergy Research (2023) 16:1667–1682 1669

via a biorefinery. It was found that machine learning and optimized using the response surface methodology (RSM)-
ANN have emerged as important methods for predicting based Box-Behnken design (BBD) and artificial neural net-
chemical behavior in many of the most well-known biomass work (ANN) to improve the process efficiency of vegetable
valorization processes. RSM is another approach to be used waste conversion. The mechanisms of inorganic and organic
as a cost-effective tool in research and industry, as it saves acids to improve enzymatic saccharification, and the pro-
time and money by reducing the numbers of experiments. It duction of bioethanol and biogas were deciphered based on
has been applied in different scientific research areas, such as analysis of composition analysis and chemical modification
biofuel production, C­ O2 capture, food engineering, chemical of biomass. The optimized process in this work could be fur-
engineering, and the petroleum refinery industry [19]. Das ther applied with the management of mixed vegetable wastes
et al. [20] used both RSM and ANN techniques for optimiz- in controlled processes and facilities to reduce greenhouse
ing the process parameters affecting enzymatic hydrolysis. gas emissions and increase energy security at the same time.
This study observed that fermentation of hydrolysate yielded
a maximum ethanol concentration of 10.44 g L ­ −1 using
−1
Pichia stripes followed by 8.24 and 6.76 g ­L for Candida Materials and Methods
shehata and Saccharomyces cerevisiae, respectively. The
average errors for ANN and RSM were 3% and 4.8%, and the Sample Preparation and Characterization of Mixed
predicted errors for ANN and RSM were 0.95 and 1.41%, Vegetable Wastes
respectively, which highlighted the significance in selecting
the appropriate approach for the process optimization. In this study, vegetable wastes, including Chinese broccoli
The above literature studies exposed that fermentation (CB), cauliflower, and Napa cabbage (NC), obtained after
and AD are the biological processes for the conversion of product processing and cutting were collected from the local
a wide range of biomass feedstock to produce bioethanol market located in the central part of Thailand. The size of
and biogas at a cheaper cost. Even though a variety of pre- each vegetable waste was reduced to about 1-cm size by
treatment procedures have been tested on a variety of feed- manually cutting. In addition to that, the dried sample size
stock, only a few reported techniques achieved high biogas was reduced by grinding, and sieving through a 20-mesh
and bioethanol yield at low enough prices. The successful aluminium sieve (~ 0.841 mm sizing) to achieve consistency
pretreatment should be able to improve the digestibility, in sample size. Then, three types of dried vegetable wastes
avoid product degradation, require low energy, and be cost- were mixed in one big batch with the ratio of 1:1:1 by dried
effective and environmentally friendly. However, the type, weight and stored in a sealed plastic bag until used for all
quality, and property of feedstock have a significant impact experiments in this study. The samples were characterized
on fermentation and AD efficiency. To achieve the feasible for the contents of cellulose, hemicellulose, and lignin by
process for biorefinery of lignocellulose biomass, optimiza- Van Soest and Wine method as used in the previous study
tion experiment and modeling are necessary to maximize the [21]. The volatile solids (VS) and total solids (TS) of the
process efficiency for each type of biomass. Therefore, most mixed vegetables were measured by the American Public
research works and industrial applications, especially in the Health Association (APHA) standard water and wastewater
production of 2nd generation ethanol, have selected and examination methods. The contents of TS and VS in mixed
utilized a single type of lignocellulose for their processes vegetables were 9.11 ± 0.58% and 8.01 ± 0.23%, respectively.
based on the empirical concept in scientific works. Practi- The compositions of cellulose, hemicellulose, and lignin
cally, natural agricultural wastes, including vegetable waste contents of mixed vegetable biomass were 20.67 ± 0.94,
obtained from cutting and processing found in dumpsters 16.97 ± 2.54, and 31.70 ± 4.88, respectively.
or municipal landfills, are mixtures of various types of lig-
nocellulose biomass. Therefore, using mixed lignocellulose Design of Experiment for Process Optimization
biomass as a substrate for biorefining process is more practi- of Pretreatment
cal and it helps to secure the dependency on one single type
of biomass in the situations that our world is facing global Design expert software (version 10.0.3, Stat-Ease, USA) has
disasters, e.g., flooding and severe drought. Hence, in this been chosen in this study for experimental design, modeling,
present work, an attempt has been made to investigate and and optimization. This work used the Box-Behnken design
compare the effect of acid pretreatment on mixed vegeta- to determine the relationship between the response (reduc-
ble lignocellulose biomass using organic acids (i.e., acetic ing sugar yield) and pretreatment parameters (reaction time,
acid, citric acid, oxalic acid) and inorganic acid (i.e., HCl) reaction temperature, and acid concentration). The Box-
to improve the enzymatic saccharification, fermentation, and Behnken design is a three-level incomplete factorial design
anaerobic digestion process. The pretreatment parameters, that is described as a cube, with center points and middle
including time, temperature, and acid concentration, are points of the edges, and necessitates a certain number of

13
1670 BioEnergy Research (2023) 16:1667–1682

experiments [5]. In this work, time (X1, 30–90 min), tem- based on the number of neurons and hidden layers with an
perature (X2, 100–140 °C), and acid concentration (X3, R-value close to unity.
2–10% w/v) were selected as tested parameters because they Backpropagation multilayer feed-forward neural network
have been demonstrated in previous studies to influence the (BPNN) is the most widely used in engineering applications
reducing sugar yield (Y, mg g­ −1 biomass) and they are eas- to predict the output parameters among the other kinds of
ily adjustable parameters during the process operation [21]. ANN architecture [24]. For ANN modeling, the backpropa-
Table 1 depicts the independent variables assigned with gation training technique is the most extensively used. The
coded and un-coded values. Equation 1 shows the generic Levenberg Marquart (LM) uses second-order derivatives,
expression for the second-order model with three independ- which results in better convergent behavior, whereas ordi-
ent variables. nary gradient descent focuses only on first-order derivatives.
3 3 3
In addition, the LM is a close approximation to the Newton’s
∑ ∑ ∑ method [25]. This study used the backpropagation learn-
Y = b0 + bj xj + bij xi xj + bjj xi2 (1)
j=1 ij=1 j=1
ing approach to train the ANN model. To achieve the best
ANN model, the Trainlm training function was used, and
where Y is the expected reducing sugar yield, xi and xj are the transfer function for the hidden layer was the logistic
the parameters, bo is the constant, bj is the linear effect, bij sigmoid. An ANN with several architectures was trained and
is the interaction effect, and bjj is the square effect of the assessed using experimental data to obtain the best predic-
parameters. Multiple analyses was performed to calculate tion. The ANN modeling was done by the MATLAB 10.3
the coefficient of variation, and the resulting equation was neural network toolbox.
used to predict the response parameters. The response was
statistically analyzed using analysis of variance (ANOVA). Organic Acid and Inorganic Acid Pretreatment
of Lignocellulose Biomass
Table 1  Independent variables along with un-coded values
Acetic acid (AA), citric acid (CA), oxalic acid (OA), and
Independent variable Units Uncoded values hydrochloric acid (HCl) pretreatments were performed for
Time X1 min 30 60 90 mixed vegetable wastes according to the condition listed in
Table 2. Five grams of dried vegetable samples were mixed
Temperature X2 °C 100 120 140
with acid solutions (with different concentrations of 2–10%
Organic acid concentration X3 % w/v 2 7 12
w/v) in the screw cap bottle with the volume to prepare the
solid–liquid ratio of 1:10% w/v. Each sample was placed
A point prediction approach was used to predict the optimal in a hot-air oven that was pre-heated to the temperature
pretreatment condition for lignocellulose biomass. Table 2 at the designed pretreatment temperature (100–140 °C)
shows the BBD matrix for three independent variables, with the retention time of the designed pretreatment time
which consisted of 17 experimental runs. (30–90 min) (Table 2). After the pretreatment process, the
sample was allowed to cool in an iced bath to room tempera-
Modeling of an Artificial Neural Network ture to stop the progress of the pretreatment. The solid and
for Pretreatment of Biomass the liquid hydrolysate were separated by using centrifugation
at 8000 × g for 10 min. The residual solids were repeatedly
An artificial neural network (ANN) is a mathematical or washed with deionized water to neutralize the pH and dried
computational model working based on the functional in a hot air oven at 60 °C for 24 h until the dried weight
aspects of a biological neural network. The ANN approach is is constant. The dried solid fraction was subsequently sub-
a nonlinear statistical data modeling tool that is used to rep- jected to enzymatic saccharification.
resent complex input–output relationships. One input layer
and an output layer with one or more hidden layers make Fourier Transform Infrared Spectroscopy Analysis
up the ANN architecture. Each layer is made up of neurons,
which comprise the ANN processing elements [22]. There Fourier transform infrared spectroscopy (FTIR) analysis
is no method for determining the number of neurons in a was performed for treated mixed vegetable lignocellulose
hidden layer or the number of hidden layers [1]. Therefore, biomass to understand the changes in the functional groups
the ANN network is trained by selecting the number of hid- before and after organic acid pretreatment [21]. An FTIR
den layers and neurons randomly. As a result, trial and error spectrometer was used to conduct the analysis (model: Spec-
approaches were used to determine the number of neurons trum 2000; make: Perklin Elmer, USA). The IR spectra were
and hidden layers in it [23]. The optimal network was chosen collected with a resolution of 4 ­cm−1 from 400 to 4000 ­cm−1.

13
Table 2  Design matrix for pretreatment of mixed vegetable wastes
BioEnergy Research (2023) 16:1667–1682

Run Pretreatment parameters Reducing sugar yield ­(mgg−1 biomass)


Time (min) Temp (°C) Acid concentration Acetic acid Citric acid Oxalic acid HCl acid
(% w/v)

1 90 140 6 224.9 155.2 159.6 186.7


2 60 120 6 266.3 128.0 177.4 145.5
3 90 120 10 248.0 105.7 212.7 198.0
4 60 140 10 314.1 128.0 172.4 113.1
5 30 120 2 180.2 161.9 133.1 632.0
6 60 140 2 251.4 160.2 128.0 118.1
7 90 100 6 263.6 163.3 161.3 123.0
8 60 100 10 193.7 129.0 220.9 145.6
9 90 120 2 224.2 192.8 162.6 125.3
10 60 100 2 292.0 167.0 183.6 127.0
11 60 120 6 220.2 117.5 204.6 150.5
12 30 120 10 170.0 211.4 270.3 118.9
13 30 100 6 214.1 133.5 299.5 144.1
14 60 120 6 270.3 118.2 149.0 148.3
15 60 120 6 229.0 111.6 158.9 118.9
16 60 120 6 279.5 135.5 180.2 137.0
17 30 140 6 214.7 110.7 151.1 109.1

13
1671
1672 BioEnergy Research (2023) 16:1667–1682

Enzymatic Saccharification of Lignocellulose a local alcohol factory in Thailand, were courtesy provided
Biomass by the Thailand Institute of Scientific and Technological
Research (TISTR) to be used in the fermentation process.
Based on the reducing sugar yield obtained from enzymatic The fermentation reaction took place in a yeast culture con-
saccharification, the efficiency of the pretreatment and the taining 19 mL of liquid hydrolysate and 1 mL yeast inocu-
optimum settings for the independent variables were cal- lum ­(A600 equal to 1.0). As a carbon and nitrogen source,
culated. The remaining solid sample produced after pre- glucose (1% w/v) and yeast extract (1% w/v) was added to
treatment were enzymatically saccharified with a cellu- the yeast culture to allow acclimatization of yeast culture [5].
lase enzyme mixture containing cellulase (CelluClast1.5L The medium’s pH was adjusted to 5.0, and it was incubated
(Sigma-Aldrich, USA), 20 FPU ­g−1 of dry biomass), and in a temperature-controlled orbital shaker at a temperature
β-glucosidase (Novozyme 188 (Sigma-Aldrich, USA), 100 of 32 °C, a stirrer speed of 100 RPM, and a reaction time of
CBU ­g−1 of dry biomass) [21]. Each saccharification reac- 48 h. Then, the liquid fraction (supernatant) of yeast culture
tion was carried out in a 50-mL centrifuge tube containing was collected by centrifugation at 8000 × g for 10 min. The
0.5 g of residual solid, 20-mL citrate buffer (50 mM; pH ethanol fermentation experiments were conducted in trip-
4.7), and 200-µL sodium azide (2 M) [26]. The tubes were licate. The ethanol concentration was measured with a gas
kept in a shaking incubator for 72 h with the temperature chromatography equipped with a flame ionization detector
and mixing set at 50 °C and 200 rpm. After the incubation (GC-FID) (Model: GC-2010, Make: Shimadzu, Japan). GC-
time, the liquid hydrolysate samples were collected by cen- FID was equipped with a DB-wax column (30 m × 0.25 mm,
trifugation at 8000 × g for 10 min, and the concentrations 0.25 m). An oven program, (a) 40 °C for 4 min, (b) 100 °C
of reducing sugars in hydrolysate samples were analyzed ramp (5 °C ­min−1), and (c) 200 °C ramp (10 °C ­min−1), was
using the 3,5-DNS technique [27]. Furthermore, a hydro- used to analyze the ethanol. The injector volume for ethanol
lysate sample for ethanol fermentation was prepared without analysis in the sample was set at 1 μl in split mode (1:20)
the addition of sodium azide. To limit the variance between [28]. The concentration of ethanol in the sample was calcu-
the experimental data, all of the studies were repeated three lated based on the standard curve of absolute ethanol solu-
times. The reducing sugar yield was calculated by the fol- tion (99.8%). The measurement of ethanol concentration was
lowing equation, conducted in triplicate. The ethanol yield was calculated by
the following equation,
Reducing sugar yield g g−1 = m M−1
( )
Ethanol yield gg−1 = c ∗ V∕M
( )
where m is the mass of reducing sugar (g) in enzymatic
hydrolysis and M is the mass of biomass (g) in enzymatic where c is the concentration of ethanol in grams per liter, V
hydrolysis. is the total volume of fermentation liquid in liter, and M is
the mass of biomass in grams.

Bioethanol Production and Quantification


Biogas Production from Mixed Vegetable
A fermentation reaction was performed in a batch opera- Lignocellulose Biomass
tion for bioethanol production from untreated and pretreated
mixed vegetable samples. Two types of yeast strains (Sac- The biogas is produced from a mixed vegetable sample
charomyces cerevisiae TISTR 5606 and Saccharomyces cer- pretreated with acids (acetic acid, citric acid, oxalic acid,
evisiae TISTR 5596), which were originally obtained from and HCl acid). The anaerobic sludge of seed inoculum was

Table 3  ANOVA for acetic Source Sum of squares df Mean square Fvalue p-value Prob >F
acid pretreated mixed vegetable
lignocellulose biomass Model 189.93 5 37.99 5.43 0.009 Significant
A-Time 41.27 1 41.27 5.90 0.033
B-Temperature 3.41 1 3.41 0.49 0.499
C-Concentration 0.17 1 0.17 0.02 0.880
BC 73.56 1 73.56 10.51 0.007
A2 71.52 1 71.52 10.22 0.008
Residual 76.99 11 7.0 -
Lack of fit 48.66 7 6.95 0.98 0.540 Not significant

13
BioEnergy Research (2023) 16:1667–1682 1673

obtained from anaerobic digesters at the municipal waste- The multi-regression analysis has generated the second-
water treatment plant, located in Bangkok, Thailand. The order polynomial equation for acid (AA, CA, OA, HCl)
seed inoculum was characterized to determine the pH, TS, pretreated lignocellulose biomass as shown in Eqs. 2–5.
and VS and to be found as 7.3 ± 0.05, 65.8 ± 0.66 g L­ −1, and Second-order model significance was tested by using analy-
−1
38.6 ± 0.45 g ­L , respectively. The seed inoculum was kept sis of variance (ANOVA) (p-value < 0.05), lack-of-fit, and
to starve in the incubator at 35 °C and stirred at 100 RPM coefficient of determination (R2). The ANOVA results for the
for 48 h before adding the anaerobic sludge into the anaer- second-order AA, CA, OA, and HCl pretreated samples are
obic digester of the treated biomass. Biogas productions given in Table 3 and Supplement Table 1–3. It was noticed
from untreated and treated mixed vegetable lignocellulose that the quadratic models were significant for all pretreated
solids were carried out in a batch-type sealed Erlenmeyer samples because the p-value was lesser than 0.05. Addi-
flask with a 500-mL working volume containing 3% of TS tionally, the p-value of lack of fit tests in AA, CA, OA, and
(anaerobic sludge), 3% w/v treated biomass, 10 mL of the HCl, were 0.5406, 0.0668, 0.3947, and 0.2865, respectively,
Nutrient Broth (NB, HiMedia, India) media, and 15 mL of which were greater than 0.05 suggesting the confidence of
the 50 mM phosphate buffer (pH 7.0). The pH of the mix- the predicted model. Furthermore, the R2 values obtained
ture was adjusted to 7; then, the reactor was purged with from the models of AA, CA, OA, and HCl were 0.7917,
nitrogen gas to maintain an anaerobic environment. The AD 0.7686, 0.8139, and 0.9368, respectively. The R2 values
process was carried out for 45 days at a temperature of 35 °C greater than 0.75 suggest the reliability of the predicted
[21]. Biogas production was observed by using the water model in several studies [19, 25]. Altogether, the model’s
displacement method. After 45 days, the biogas yield for p-value, lack of fit test, and R2 obtained from this study
each day was determined and calculated for accumulated advocate the confidence of the acid pretreatment models.
biogas yields. The biogas production of each sample was
Yacetic = +25.67 + 2.43 ∗ A + 0.80 ∗ B − 0.13 ∗ C − 1.56 ∗ AB
conducted in triplicates. (2)
+0.60 ∗ AC + 4.56 ∗ BC − 4.63A2 + 1.33 ∗ B2 − 0.72C2

Statistical Analysis
Ycitric = +11.27 + 0.18 ∗ A − 0.31 ∗ B − 1.79 ∗ C + 0.36 ∗ AB
(3)
The statistical significance for the changes in the experimen- −2.80 ∗ AC + 0.14 ∗ BC + 1.88A2 + 0.31 ∗ B2 + 2.58C2
tal data for yields of reducing sugar, ethanol, and accumu-
lated biogas produced from untreated and pretreated samples Yoxalic = +17.41 − 1.97 ∗ A − 3.18 ∗ B + 3.36 ∗ C + 3.67 ∗ AB
(4)
was performed through ANOVA and Tukey’s post hoc test −2.18 ∗ AC + 0.18 ∗ BC + 1.86A2 + 0.018 ∗ B2 + 2.68C2
(p < 0.05) by using the SPSS software (version 28.0.0).
Yhcl = +14.21 + 2.72 ∗ A + 1.52 ∗ B + 3.20 ∗ C + 2.47 ∗ AB
(5)
+0.42 ∗ AC − 3.45 ∗ BC + 1.38A2 − 0.1.51 ∗ B2 + 2.95C2
Results and discussion
After the analysis of model significance, the significance
Influence of Acid Pretreatment on Enzymatic of each term model representing the pretreatment parameter
Saccharification of Mixed Vegetable Waste was analyzed to identify the influence of individual pretreat-
ment parameters and the interacting effect of two parameters
The mixed vegetable waste samples were pretreated with at a time. The significant term models (p-value < 0.1) were
organic acids (acetic acid, citric acid, oxalic acid, and inor- selected to be used to generate the contour plots that pre-
ganic acid (HCl) based on conditions designed and listed sented the influence of each pretreatment factor and interact-
in Table 2. After pretreatment, the solid fraction was sepa- ing factors on reducing sugar yields. The interaction effect of
rated and enzymatically hydrolyzed by a mixed cellulase time, temperature, and acid concentration on reducing sugar
enzyme. Reducing sugar concentration in the biomass yield for organic acid pretreated biomass is shown in Fig. 1.
hydrolysate was measured as the indicator to determine the Based on the three-dimensional contour plot, it was
pretreatment efficiency. The reducing sugar yield obtained observed that the RS yield decreased with an increase in
from organic acid pretreated samples (AA, CA, OA, HCl) temperature from 100 to 140 °C, and with an increase in acid
varied from 170 to 314.1 mg ­g−1, 105.7–192.8 mg ­g−1, concentration from 2 to 10% w/v for acetic acid pretreated
128–299.5 mg ­g−1, and 63.2–198 mg ­g−1 biomass, respec- samples (Fig. 1a). This sugar loss during acid pretreatment
tively, whereas the sugar yield released from the untreated could be due to the degradation of sugar during the intense
sample was 99.2 mg ­g−1 biomass. The difference between pretreatment condition. For the citric acid pretreatment, it
the maximum and minimum yields obtained from each set was observed that the RS yield increased as the acid con-
of acid pretreatment suggested the importance of pretreat- centration and time increased from 2 to 10% w/v and 30 to
ment optimization. 90 min, respectively (Fig. 1b). Increased acid concentration

13
1674 BioEnergy Research (2023) 16:1667–1682

Fig. 1  Contour plot of reducing sugar yield (a acetic acid; b citric acid; c, d oxalic acid; e, f HCl acid against time, temperature, and acid con-
centration)

Table 4  Optimum conditions for pretreated lignocellulose biomass using different acids
Pretreatment Time (X1, min) Temperature Acidconcentra- Predicted RS yield Observed RS yield % error
(X2, °C) tion (X3,%) ­(mgg−1 biomass) (mg ­g−1 biomass)*

Acetic acid 68.32 140 10 309.7 314.1 ± 1.27a 1.40


Citric acid (CA) 89.8 106.13 2.11 211.5 220 ± 0.08b 4.26
Oxalic acid (OA) 31 101 8.52 300.9 288.2 ± 1.46c 4.16
Hydrochloric acid (HA) 89 137 6 200.2 189.2 ± 1.06d 5.82

*The small letters presented that the data of each sample was statistically different at p-value < 0.05

13
BioEnergy Research (2023) 16:1667–1682 1675

solubilizes hemicellulose and partially changes lignin in model prediction and experiments were 1.40–5.39%, sug-
plant material. The removal of the hemicellulose and lignin gesting the high validity of the model.
layers allows cellulase enzymes easier access to the cellu- The reducing sugar yields of mixed vegetable samples,
lose layer, which leads to an increase in the RS yield [29]. which were pretreated with CA, OA, AA, and HA, changed
After oxalic acid pretreatment, the RS yield decreased when significantly during the saccharification when compared to
the reaction temperature was increased from 100 to 140 °C. the untreated sample. Among all acid pretreated samples,
However, the RS yield increased as an increased the acid AA-pretreated samples released the highest sugar yields at
concentration from 2 to 10% w/v (Fig. 1c, d). In the HCl- 314.1 mg ­g−1 biomass, which was higher than the untreated
pretreated sample (Fig. 1e, f), the RS yield increased with sample (at 99.2 mg ­g−1 biomass) by 3.16 fold. After the
an increase in temperature from 100 to 120 °C. This is due AA-pretreated sample, OA and CA pretreatments were the
to an increase in reaction temperature improves the solubi- second and third acids to release the highest sugar yields,
lizes of hemicellulose and partially changes lignin in plant respectively. Likewise, the HCl-pretreated sample released
material. On the other hand, the RS yield decreased with an a higher reducing sugar yield than the untreated sample
increase in reaction time from 30 to 60 min, but the RS yield for 1.91 fold. In this study, it was observed that the sugar
increased when further increased the reaction time from yields obtained from organic acid pretreated samples were
60 to 90 min. Furthermore, it was found that the RS yield higher than from inorganic acid pretreated samples (HCl).
increased with an increase in reaction temperature from 100 This observation is in agreement with the previous study on
to 140 °C and with an increase in acid concentration. At organic acid pretreatment of rice straw that the saccharifica-
lower temperatures and acid concentrations, the RS yield tion efficiency of the OA-pretreated sample was higher than
was minimum due to the low efficiency of pretreatment. A the HCl-pretreated sample [21]. However, among three types
point prediction approach was employed to predict the pre- of organic acids in previous works, OA was the best acid for
treatment optimum conditions to obtain maximum RS yield. improving enzymatic saccharification in rice straw [21] and
Based on the generated RSM models, the predicted oil palm trunk [30]. Altogether, this pretreatment experiment
yields of reducing sugars obtained from pretreated samples and previous works revealed the importance of optimization
by AA, CA, OA, and HCl were 309.7, 211.5, 300.9, and work for each type of biomass to maximize the efficiency of
200.2 mg ­g−1 biomass, respectively (Table 4). The predicted the biorefining process.
optimum conditions were validated by performing the pre-
treatment experiments for lignocellulose biomass using dif- Influence of Acid Pretreatments on Biomass
ferent acids and the sugar yields obtained from the experi- Composition and Functional Chemical Group
ment were measured and compared with the predicted yields. Rearrangements
The experimental results showed that RS yields obtained
from pretreated samples by AA, CA, OA, and HCl were Acid pretreatment has previously been reported to enhance
314.1, 220, 288.2, and 189.2 mg ­g−1 biomass, respectively hemicellulose solubilization leading to increased glucose
(Table 4). The differences in sugar yields obtained from the yield. The composition of the organic acid pretreated bio-
mass is represented in Table 5. It is observed that the organic
acid pretreatment helped in hemicellulose removal from the
biomass when compared to the untreated sample. Addition-
Table 5  Compositional analysis of mixed vegetable biomass after
organic acid pretreatment
ally, the lignin content of the pretreated sample has also
reduced in comparison with the untreated sample. On the
Sample Biomass composition (%wt)* other hand, the cellulose contents in the acid pretreated bio-
Cellulose Hemicellulose Lignin mass are enriched, except for the biomass pretreated with
citric acid and hydrochloric acid. The presence of cellulose
Untreated biomass 20.67 ± 0.94b 16.97 ± 2.54a 31.70 ± 4.88a
content in CA-pretreated and HCl-pretreated samples can be
AA-pretreated 22.12 ± 2.16b N.A 25.60 ± 2.03b
biomass correlated with reduced sugar yields (Table 4). Chen et al.
CA-pretreated 18.58 ± 0.16c 6.46 ± 0.49b 17.14 ± 2.38d 2020 selected 4 types of organic acids for the pretreatment
biomass of eucalyptus biomass for Kraft pulp productions and found
OA-pretreated 28.06 ± 1.25a N.A 12.98 ± 0.84e that all acid pretreatments showed high efficiency in the pro-
biomass motion of reducing sugars yields and hemicellulose removal.
HCl-pretreated 20.666 ± 2.96b N.A 22.71 ± 1.79c However, lignin was dissolved by acetic acid pretreatment,
biomass while other acid pretreatments unexpectedly cause severe
*The small letters presented that the data of each sample was statisti- cellulose degradation and reduction in sugar yields [31]. The
cally different within the same column at p-value < 0.05 pretreatment could have disrupted the hydrogen bonds in
N.A. not available the cellulose layer in addition to hemicellulose and lignin

13
1676 BioEnergy Research (2023) 16:1667–1682

Fig. 2  FTIR analysis for


untreated and treated lignocel-
lulose samples

removal, resulting in cellulose solubilization and reduced 2918 ­cm−1, and 3422 ­cm−1. In pretreated samples, there was
sugar yield [32]. However, the increased sugar yield a difference in the transmittance strength of these peaks as
obtained from the AA-pretreated and OA-pretreated sam- compared to untreated samples. The transmittance intensity
ples were the results of the lignin and hemicellulose removal of peak 1049 ­cm−1 decreased in acid pretreated samples,
through the cleavage of acetyl groups in the hemicellulose, especially CA, OA, and HCl when compared to untreated
thereby exposing cellulose to enzymatic hydrolysis. Gener- samples. The peak of 1049 ­cm−1 in hemicellulose was attrib-
ally, the solubilized sugars and the degradation products of uted to the mid-strong stretching of C–O, C–C bonds, and
cellulose, hemicellulose, and lignin, which formed during the mid-strong bending of C–OH bonds [34]. This observa-
the pretreatment, are mixed and solubilized in the pretreat- tion suggested that both organic acids and inorganic acids
ment hydrolysate. Therefore, an additional step of detoxifi- remove hemicellulose or cause hemicellulose hydrolysis
cation of hydrolysate is recommended to remove the inhibi- during the pretreatment process [33]. In comparison to the
tory compounds, such as furfural, 5-hydroxymethylfurfural, untreated, AA, CA, and HCl pretreated biomass, the peak
and phenolic chemicals, which may impair or inhibit the intensity of 1319 ­cm−1 attributed to the C-O, C–C, and ­CH2
enzymatic saccharification and microbial activity leading stretching vibrations in cellulose in the OA-pretreated sam-
to reduced sugar and bioethanol yield [32]. Similar find- ple was increased in OA pretreated biomass. The high cel-
ings about the mechanism of organic acid pretreatment were lulose ratio left in the OA-pretreated sample obtained from
revealed by Luo et al. 2021 [33]. When eucalyptus biomass the composition analysis (Table 5) supported the results of
was pretreated by glycolic acid at the optimal condition (acid the FTIR analysis. The phenolic chemicals found in lignin
concentration 5.40%, temperature 140 °C, time 3.0 h), hemi- are responsible for the peak of 1638 ­cm−1, and the inten-
cellulose was hydrolyzed to xylose at the yield of about 56%, sity of this peak was reduced in pretreated samples with CA
whereas the repolymerization of lignin was inhibited and and HCl when compared to untreated samples suggesting
cellulose content was preserved and recovered. the removal of lignin during acid pretreatment [21]. On the
To understand the mechanism of acid pretreatment for other hand, oxalic acid showed a drop in peak intensity of
improvement of enzymatic saccharification, FTIR analysis 1725 ­cm−1 due to the C = O bond in hemicellulose followed
for treated and untreated samples was conducted to observe by AA, CA, HCl, and untreated samples implying the loss
the alteration and rearrangement of chemical bonding within in hemicellulose during acid pretreatment. The peak inten-
lignocellulose biomass (Fig. 2). The primary peaks were sity of 2918 ­cm−1 corresponding to CH stretching vibration
observed for lignocellulose samples from the FTIR analy- in the aromatic methoxyl, methyl, and methylene groups in
sis, such as 1049 ­cm−1, 1319 ­cm−1, 1638 ­cm−1, 1725 ­cm−1, lignin was reduced in the acid pretreated samples, especially

13
BioEnergy Research (2023) 16:1667–1682 1677

Fig. 3  Ethanol yield for treated


and untreated lignocellulose
samples. Different small alpha-
bet and capital alphabet repre-
sent the statistical difference at
p-value < 0.05

CA and AA pretreatment [5], which were well correlated CA pretreatment on the oil palm trunk. The highest sugar
with the composition analysis (Table 5). Also, the broad yield was obtained from OA-pretreated palm trunk, while
absorption peak of 3422 ­cm−1 representing the OH bond the highest ethanol yield was obtained from CA-pretreated
stretching vibration in cellulose [35] was increased in AA sample. This anomaly observation, in which the OA-pre-
pretreated samples when compared to untreated samples treated sample was not the method promoting the highest
suggesting the enrichment of cellulose and it was subse- sugar yields from enzymatic saccharification (Table 4) but
quently hydrolyzed to high yields of reducing sugars. producing the highest ethanol yields from fermentation,
could be hypothesized to be the results of the consump-
Bioethanol Production from Mixed Vegetable tion of remaining oxalic acid in pretreated biomass by yeast
Lignocellulose Biomass for ethanol production via TCA cycle. Zhang et al. applied
oxalic acid pretreatment on wheat straw and corn stover to
Figure 3 depicts the bioethanol production from mixed improve ethanol production up to 9.9% and 9.5%, v/v. This
vegetable biomass using the fermentation process with S. study proposed the explanation that oxalic acid functions as
cerevisiae TISTR5606 and S. cerevisiae TISTR5596. Using the catalyst for lignocellulose hydrolysis and simultaneously
various acid pretreatment, the ethanol yield produced from degrade the inhibitory byproducts and finally oxalic acid was
each pretreated sample was enhanced when compared to consumed by microbes at low pH condition [36].
untreated biomass. Using S. cerevisiae TISTR5606, the Similar trends in the effect of each acid on ethanol pro-
highest ethanol yield at 7.60 g ­L−1 was obtained from the duction from S. cerevisiae TISTR5606 and S. cerevisiae
OA-pretreated sample, which was equivalent to 86.36% of TISTR5596 were observed, although the ethanol yield
theoretical yields, followed by CA, AA, HCl, and untreated obtained from S. cerevisiae TISTR5596 was less than from
samples, respectively. In this study, pretreatment of mixed S. cerevisiae TISTR5606. The efficiency of ethanol pro-
vegetable waste with OA improved ethanol yield by 2.75- duction of different yeast strains could be due to the char-
fold time compared to the untreated sample. It was noted acteristics of yeast and biomass. Although S. cerevisiae
that AA pretreatment provided the highest reducing sugar TISTR5606 and S. cerevisiae TISTR5596 were obtained
yield, while OA pretreatment was the best option for eth- from alcohol factories using sugarcane molasses as sub-
anol production. This finding was different from the pre- strate, they were further applied for ethanol production in
vious study [30] that reported the effect of AA, OA, and the fermentation processes of fruit waste [37] and sorghum

13
1678 BioEnergy Research (2023) 16:1667–1682

straw [38], respectively. Therefore, the results reported in higher ethanol yields than AA-pretreated biomass. For
our work could be used as a guideline to select the yeast ethanol production, wild yeast species, e.g., S. cerevisiae,
strain to improve ethanol production and suggested that depend on C6 carbohydrates, because native yeast species
both strains have the potential for 2nd generation ethanol are incapable of converting xylose to ethanol during the
production. fermentation process. However, the commercialization of
Several biomass byproducts were identified and charac- genetically modified yeast capable of fermenting C5 car-
terized as inhibitors to hydrolysis enzymes and microbial bohydrates to ethanol has yet to occur [42]. Alternatively,
activities. The effects of remaining inhibitory byproducts co-culturing of yeast strains, S. cerevisiae and Pichia stip-
generated during acid pretreatment on bioethanol produc- ites, which can ferment C6 and C5 sugars, respectively, is
tion were demonstrated previously [39]. The study reported proposed to be the solution to fully utilize lignocellulose
that unwashed samples obtained the minimum reducing biomass for the production of ethanol. Similarly, organic
sugar yield and ethanol after the saccharification and fer- acid pretreatment of cotton gin by using lactic acid, oxalic
mentation process. This is due to the inhibition caused acid, citric acid, and maleic acid is optimized to maximize
by furfural and 5-hydroxymethylfurfural bound to the ethanol production from the co-culturing of S. cerevisiae
biomass. Also, the furfural and 5- hydroxymethylfurfural and Pichia stipites. It was demonstrated that lignin in
exhibited a stronger inhibitory effect on enzymes than maleic acid-pretreated biomass was removed up to 88%,
formic acid, acetic acid, and phenolic acid compounds 686.13 g/g saccharification yield was obtained, and the
produced during the pretreatment. Furthermore, CA-pre- highest ethanol yield at 18.74 g/l was achieved when using
treatment on rice straw was demonstrated to generate a an optimum pretreatment condition [43]. Comparatively,
low concentration of furan and phenolic compounds, which based on the results of our work (using mixed vegetable
subsequently improved ethanol production when compared waste) and other works (using a single type of biomass) on
to untreated rice straw [40]. The difference between acid organic acid pretreatment [21, 30, 36, 40, 41, 43], it was
pretreatment on sugar yield and ethanol yield of various demonstrated that the optimization experiment of pretreat-
types of biomass highlight the significance of optimization ment parameters, including type of organic acid, pretreat-
and tailor-made process for each feedstock to establish a ment time, pretreatment temperature, and acid concentra-
highly efficient process. A similar finding was observed tion was necessary to select the suitable method with high
by Marsden et al. [41] that CA-pretreated biomass had efficiency for each type of biomass. Also, a single acid for

Fig. 4  Cumulative biogas


production from treated and
untreated lignocellulose sam-
ples. Different small alphabets
represent the statistical differ-
ence at p-value < 0.05

13
BioEnergy Research (2023) 16:1667–1682 1679

pretreatment is not recommended for all types of biomass optimized to improve methane production. It was demonstrated
before preliminary experiments or optimization experi- that each pretreatment chemical needs optimization to obtain
ments are conducted. Furthermore, when using one single appropriate pretreatment severity to improve methane produc-
type of biomass, it was demonstrated that different types tion because each chemical has a specific mechanism for pre-
of organic acids were specifically suitable for improvement treatment. For example, the moderate severity of citric acid and
of enzymatic saccharification, ethanol fermentation, and ­H2O2 pretreatment conditions was used; the methane yield from
biogas production suggesting the crucial role of optimiza- winery waste was maximized up to 163.7 and 190.2 mL/g-VS,
tion experiment of each type of biomass. respectively. On the other hand, using ethanol pretreatment pre-
ferred a more severe condition to be a highly effective pretreat-
Biogas Production from Mixed Vegetable ment [44]. Although in this study, the digestate of the biogas
Lignocellulose Biomass reactor was not further analyzed for its qualitative values, it has
been proposed in several works that the digestate obtained after
The influence of organic acid pretreatment on biogas produc- pretreatment and anaerobic digestion were applied as feedstocks
tion from mixed vegetable biomass using an anaerobic pro- for microalgae cultivation [45] and as fertilizers of plants [46].
cess was investigated. The anaerobic process was carried out The sustainable processes to utilize the anaerobic digestate of
for the pretreated samples, and the pH value of the samples lignocellulose biomass for the production of algal biomass for
was adjusted to 7.0 using 1 N NaOH. The pH condition plays further biorefining process of other value-added products were
an important role in biogas production and can cause the reviewed recently by Javed et al. 2022 [47]. The main bottle-
system to fail during the reactor startup stage. As a result, neck of the utilization of lignocellulose digestate was described
pH neutralization to 7.0 is required for biogas production to be the lacking of an efficient pretreatment method to improve
during anaerobic digestion of mixed vegetable lignocellulose the conversion of lignocellulose biomass [47].
biomass. The cumulative biogas yield for untreated, acetic,
citric, oxalic, and hydrochloric acid pretreated solids varied ANN Results in Comparison with Experimental
as 276.5 mL, 634 mL, 1394 mL, 1098 mL, and 469.5 mL, Values
respectively (Fig. 4). After 35 days of the AD process, the
biogas yield of acetic, citric, oxalic, and hydrochloric acid In this study, the ANN models were additionally used to predict
pretreated biomass increased by 2.29, 5.04, 3.97, and 1.69 the reducing sugar yield of pretreated lignocellulose biomass
times, respectively, as compared to the untreated sample. with time, temperature, and acid concentration as input vari-
The highest biogas yield (equivalent to 110.72 mL/g- ables. The tansigmoid transfer function was adopted for both
VS) was found in the following order: citric acid > oxalic hidden and output layer neurons in the selected feed-forward
acid > acetic acid > HCl acid > untreated sample (equivalent network [1]. The transfer function gives neural network non-
to 21.96 mL/g-VS). This is because of the maximum sugar linearity, which makes it more efficient than the linear trans-
content remaining in the biomass during pretreatment when formations and the performance function to be optimized by
compared to other pretreated samples. Further, the degree of mean square error. The numbers of neurons in the hidden layer
sugar solubilization is dependent on the pretreatment operat- were increased from two to twenty-nine for each optimal itera-
ing parameters. tion. The MSE was used as a criterion throughout the topology
In previous reports, the CA pretreatment on rice straw and research. The neural network was built and trained by using the
palm oil trunk was demonstrated to be a more effective method data given in Table 2. The ANN is built using 29 data points in
to enhance biogas production than other acid pretreatments total. The network is trained with 70% of the total experimental
(AA, OA, and HCl) [21, 30]. In comparison to the other treated data, while the rest of the data was used for model testing and
and untreated samples, CA pretreatment caused chemical and validation. Prior to the training session, the input and target val-
compositional changes in the lignin, hemicelluloses, and cel- ues are normalized. The goal of the normalization is to prevent
lulose of lignocellulose to be more susceptible to enzymatic overflow situations. The neural network training is forced to
hydrolysis [14]. Unlike ethanol production, when using wild stop when the network error (MSE) decreases to a low number
yeast strain for fermentation, only C6 sugars are consumed and (MSE ≤ Eo). In the present study, the training came to end after
converted to ethanol; the AD process overcomes the disadvan- 9 iterations (epochs) [20].
tages of fermentation by utilizing C5, C6 sugars, and VFA as The optimal architecture for reducing sugar yield with acetic
substrates by the functions of microbial consortium in AD acid pretreated biomass was found to be three neurons in the
reactor. The efficiency of acid pretreatment is governed by the input layer, sixteen neurons in the hidden layer, and two neu-
severity of pretreatment conditions, which are depending on rons in the output layer (3–16-1), while the optimal architecture
the type of chemical and type of biomass. The pretreatments for RS yield pretreated with citric acid was found to be three
of winery waste, cotton gin waste, olive pomace, and juice neurons in the input layer, eighteen neurons in the hidden layer,
industry waste by using eight different types of chemicals were and two neurons in the output layer (3–18-1). For oxalic acid

13
1680 BioEnergy Research (2023) 16:1667–1682

pretreated biomass, the optimum architecture for RS was three Conclusion


neurons in the input layer, fourteen in the hidden layer, and one
in the output layer (3–17-1), whereas the optimum architecture The effect of organic acid pretreatment on mixed vegetable
for RS yield pretreated with HCl was found to be three neurons waste conversion into bioethanol and biogas was modeled and
in the input layers, nineteen neurons in the hidden layer, and one experimentally investigated based on RSM and ANN tech-
neuron in the output layer (3–19-1). The optimum architecture niques. The predicted reducing sugar yields obtained from
for the performance model was found to be (4–16-3), whereas RSM are relatively nearer to the experimental results, and
the optimal architecture for the emission model was found to low error percentage, e.g., 1.4% error in prediction of acetic
be (4–14-6) [1]. acid pretreated sample. The Levenberg–Marquardt algorithms
The generated ANN model is significant and may be along with transigmoid transfer function have accurately pre-
utilized to predict the best topology, as evidenced by the dicted the RS yield, which was in good agreement with experi-
strong correlation coefficient (R2 = 0.997) and adjusted cor- mental results. In this study, the highest conversion yield of
relation coefficient (R2 = 0.987) is in reasonable agreement reducing sugar at 314.1 ± 1.27 mg/g biomass was obtained
with each other. The degree of agreement between experi- from acetic acid-pretreated sample, whereas the highest yields
mental and model-predicted values shows that the major- of ethanol (7.60 g/L, equivalent to 86.36% of theoretical yields)
ity of the experimental results lie on the 45-degree line, and biogas (110.72 mL/g-VS) were obtained in oxalic acid and
indicating that the predicted results are extremely similar to citric acid pretreated samples, respectively. In addition, organic
the experimental data. The generalization potentiality of the acid pretreatment has the advantage of breakdown the inhibi-
developed ANN model was increased by using the follow- tory chemicals, which obstruct enzyme and microbial action.
ing techniques such as normalization, the ideal numbers of Also, organic acids could be additionally consumed by micro-
neurons, and separating the target data into training, valida- bial fermentation to increase the yields of the targeted product.
tion, and test data subsets [22]. The result reveals that the Therefore, organic acid has been suggested for the pretreatment
R and RMSE values of RS yield pretreated with different of mixed vegetable wastes as a greener chemical for biofuel
acids are in the range of 0.921–0.986 and 0.0123–0.0166, production to reduce the dependency on using a single type of
respectively. Hence, the ANN model can be utilized as an biomass that is often used in current R&D activities.
effective tool for modeling and predicting RS yield from
lignocellulose vegetable waste pretreated with organic and Supplementary Information The online version contains supplemen-
inorganic acids. Das et al. [20] optimized the enzymatic tary material available at https://​doi.​org/​10.​1007/​s12155-​022-​10517-y.
saccharification of water hyacinth biomass for bioethanol
Funding This work was financially supported by King Mongkut’s
production using the ANN and RSM approach. The study University of Technology North Bangkok (Grant Contract Nos.
revealed that the correlation coefficient values were in the KMUTNB-Post-65–09, KMUTNB-66-KNOW-03) and Srinakharin-
range of 0.98 to 0.99. Similarly, Neto et al. [48] observed wirot University (Grant Contract No. 410/2565-SWU).
correlation coefficient values between 0.68 and 0.99 in their
research work. If the R value is close to 1, it seems that the Declarations
ANN model has a lower error percentage.
The overall correlation coefficient of the developed network Conflict of interest The authors declare no competing interests.
for RS yield pretreated with acetic acids were analyzed and
showed that the R values in training, testing, and validation were
0.97, 0.97, and 0.98, respectively, with an overall R value of References
0.97. A similar finding was observed from the previous research
works. Kumari et al. [49] optimized the pretreatment process for 1. Kshirsagar CM, Anand R (2017) Artificial neural network
Leucaena leucocephala using the ANN approach. It found that applied forecast on a parametric study of Calophyllum ino-
phyllum methyl ester-diesel engine out responses. Appl Energy
the LM neural network model has better prediction with MSE 189:555–567. https://​doi.​org/​10.​1016/j.​apene​rgy.​2016.​12.​045
and R values of 0.07 and 0.99 respectively. Similarly, Esfand- 2. Saravanan A, Senthil Kumar P, Jeevanantham S, Karishma S, Dai-
yari et al. [50] used the BPFFNN network for the prediction of Viet N (2022) Recent advances and sustainable development of
Fisher-Tropsch synthesis of natural gas with carbon monoxide biofuels production from lignocellulosic biomass. Bioresour Tech-
nol 344:126203. https://​doi.​org/​10.​1016/j.​biort​ech.​2021.​126203
using an aluminium oxide catalyst. The study revealed that (4–7- 3. Saba N, Jawaid M, Hakeem KR, Paridah MT, Khalina A, Alothman
9–3) was the optimal architecture with a correlation coefficient OY (2015) Potential of bioenergy production from industrial kenaf
and an RMSE value of 0.93 and 0.66, respectively. Another (Hibiscus cannabinus L.) based on Malaysian perspective. Renew Sust
study reported the R values in training, testing, and validation Energ Rev 42:446–459. https://​doi.​org/​10.​1016/j.​rser.​2014.​10.​029
4. Das A, Ghosh P, Paul T, Ghosh U, Pati BR, Mondal KC (2016)
of ANN emission were 0.99, 0.98, and 0.98 respectively, with an Production of bioethanol as useful biofuel through the bioconver-
overall R value of 0.99. The ANN models were in good agree- sion of water hyacinth (Eichhornia crassipes). 3 Biotech 6:1–9.
ment with statistical error measurement. https://​doi.​org/​10.​1007/​s13205-​016-​0385-y

13
BioEnergy Research (2023) 16:1667–1682 1681

5. Gundupalli MP, Cheng YS, Chuetor S, Bhattacharyya D, Sriariyanun 20. Das S, Bhattacharya A, Haldar S, Ganguly A, Ting YP, Chat-
M (2021) Effect of dewaxing on saccharification and ethanol pro- terjee PK (2015) Optimization of enzymatic saccharification of
duction from different lignocellulosic biomass. Bioresour Technol water hyacinth biomass for bio-ethanol: comparison between
339:125596. https://​doi.​org/​10.​1016/j.​biort​ech.​2021.​125596 artificial neural network and response surface methodology.
6. Khan MU, Usman M, Ashraf MA, Dutta N, Luo G, Zhang S Sustain Mater Technol 3:17–28. https://​doi.​org/​10.​1016/j.​sus-
(2022) A review of recent advancements in pretreatment tech- mat.​2015.​01.​001
niques of lignocellulosic materials for biogas production: oppor- 21. Amnuaycheewa P, Hengaroonprasan R, Rattanaporn K, Kird-
tunities and limitations. Chem Eng J Adv 10:100263. https://​doi.​ ponpattara S, Cheenkachorn K, Sriariyanun M (2016) Enhanc-
org/​10.​1016/j.​ceja.​2022.​100263 ing enzymatic hydrolysis and biogas production from rice straw
7. Mood SH, Golfeshan AH, Tabatabaei M, Jouzani GS, Najafi by pretreatment with organic acids. Ind Crops Prod 87:247–254.
GH, Gholami M, Ardjmand M (2013) Lignocellulosic biomass https://​doi.​org/​10.​1016/j.​indcr​op.​2016.​04.​069
to bioethanol, a comprehensive review with a focus on pretreat- 22. Kannan GR, Balasubramanian KR, Anand R (2013) Artificial
ment. Renew Sustain Energy Rev 27:77–93. https://​doi.​org/​10.​ neural network approach to study the effect of injection pressure
1016/j.​rser.​2013.​06.​033 and timing on diesel engine performance fueled with biodiesel.
8. Woiciechowski AL, Neto CJ, de Souza Vandenberghe LP, de Car- Int J Automot Technol 14:507–519. https://​doi.o​ rg/1​ 0.1​ 007/s​ 1223​
valho Neto DP, Sydney AC, Letti LA, Karp SG, Torres LA, Soc- 9−013−0055−6
col CR (2020) Lignocellulosic biomass: acid and alkaline pretreat- 23. Sarkar B, Sengupta A, De S, DasGupta S (2009) Prediction of
ments and their effects on biomass recalcitrance — conventional permeate flux during electric field enhanced cross-flow ultrafiltra-
processing and recent advances. Bioresour Technol 304:122848. tion-A neural network approach. Sep Purif Technol 65:260–268.
https://​doi.​org/​10.​1016/j.​biort​ech.​2020.​122848 https://​doi.​org/​10.​1016/j.​seppur.​2008.​10.​032
9. Roy R, Rahman MS, Raynie DE (2020) Recent advances of greener 24. Rao KP, Babu VT, Anuradha G, Rao BVA (2017) IDI diesel
pretreatment technologies of lignocellulose. Curr Res Green Sustain engine performance and exhaust emission analysis using biodiesel
Chem 3:100035. https://​doi.​org/​10.​1016/j.​crgsc.​2020.​100035 with an artificial neural network (ANN). Egypt J Pet 26:593–600.
10. Jamaldheen SB, Kurade MB, Basak B, Yoo CG, Oh KK, Jeon BH, Kim https://​doi.​org/​10.​1016/j.​ejpe.​2016.​08.​006
TH (2020) A review on physico-chemical delignification as a pretreat- 25. Baranitharan P, Ramesh K, Sakthivel R (2019) Measurement of
ment of lignocellulosic biomass for enhanced bioconversion. Bioresour performance and emission distinctiveness of Aegle marmelos
Technol 346:126591. https://​doi.​org/​10.​1016/j.​biort​ech.​2021.​126591 seed cake pyrolysis oil/diesel/TBHQ opus powered in a DI diesel
11. Areepak C, Jiradechakorn T, Chuetor S, Phalakornkule C, Sriari- engine using ANN and RSM. Measurement 144:366–380. https://​
yanun M, Raita M (2022) Improvement of lignocellulosic pretreat- doi.​org/​10.​1016/j.​measu​rement.​2019.​05.​037
ment efficiency by combined chemo — mechanical pretreatment for 26. Cheng YS, Wu ZY, Sriariyanun M (2019) Evaluation of Maca-
energy consumption reduction and biofuel production. Renew Energy ranga tanarius as a biomass feedstock for fermentable sugars pro-
182:1094–1102. https://​doi.​org/​10.​1016/j.​renene.​2021.​11.​002 duction. Bioresour Technol 294:122195. https://d​ oi.o​ rg/1​ 0.1​ 016/j.​
12. Ahmad F, Silva V, Sakamoto IK, Sun J, Silva EL, Varesche MB biort​ech.​2019.​122195
(2022) Bioprospecting sulfuric acid assisted hydrothermal pre- 27. Miller GL (1959) Use of dinitrosalicylic acid reagent for determi-
treatment of sugarcane bagasse and microbial community struc- nation of reducing sugar. Anal Chem 31:426–428. https://​doi.​org/​
ture for methane production. Bioenergy Res 15:631–649. https://​ 10.​1021/​ac601​47a030
doi.​org/​10.​1007/​s12155-​021-​10268-2 28. Gundupalli MP, Chuetor S, Cheenkachorn K, Rattanaporn K,
13. Thulluri C, Goluguri BR, Konakalla R, Shetty PR, Addepally U Show PL, Cheng YS, Sriariyanun M (2021) Interferences of
(2013) The effect of assorted pretreatments on cellulose of selected waxes on enzymatic saccharification and ethanol production from
vegetable waste and enzymatic hydrolysis. Biomass Bioenerg lignocellulose biomass. Bioengineering 8(11):171. https://d​ oi.o​ rg/​
49:205–213. https://​doi.​org/​10.​1016/j.​biomb​ioe.​2012.​12.​022 10.​3390/​bioen​ginee​ring8​110171
14. Tajmirriahi M, Momayez F, Karimi K (2021) The critical impact of 29. Siqueira G, Arantes V, Saddler JN, Ferraz A, Milagres AMF
rice straw extractives on biogas and bioethanol production. Bioresour (2017) Limitation of cellulose accessibility and unproduc-
Technol 319:124167. https://​doi.​org/​10.​1016/j.​biort​ech.​2020.​124167 tive binding of cellulases by pretreated sugarcane bagasse
15. Bittencourt GA, da Silva BE, Brandão RL, Baêta BE, Gurgel LV lignin. Biotechnol Biofuels 176:4058. https://​doi.​org/​10.​1186/​
(2019) Fractionation of sugarcane bagasse using hydrothermal s13068-​017-​0860-7
and advanced oxidative pretreatments for bioethanol and biogas 30. Rattanaporn K, Tantayotai P, Phusantisampan T, Pornwongthong
production in lignocellulose biorefineries. Bioresour Technol P, Sriariyanun M (2018) Organic acid pretreatment of oil palm
292:121963. https://​doi.​org/​10.​1016/j.​biort​ech.​2019.​121963 trunk: effect on enzymatic saccharification and ethanol produc-
16. Nitsos CK, Matis KA, Triantafyllidis KS (2013) Optimization tion. Bioprocess Biosyst Eng 41:467–477. https://d​ oi.o​ rg/1​ 0.1​ 007/​
of hydrothermal pretreatment of lignocellulosic biomass in s00449-​017-​1881-0
the bioethanol production process. Chemsuschem 6:110–122. 31. Chen Y, Yan Z, Liang L, Ran M, Wu T, Wang B, Zou X, Zhao M,
https://​doi.​org/​10.​1002/​cssc.​20120​0546 Fang G, Shen K (2020) Comparative evaluation of organic acid pre-
17. Peinado S, Mateo S, Sánchez S, Moy AJ (2019) Effectiveness of treatment of eucalyptus for kraft dissolving pulp production. Mater.
sodium borohydride treatment on acid hydrolyzates from olive- Lett (Basel) 12;13(2):361. https://​doi.​org/​10.​3390/​ma130​20361
tree pruning biomass for bioethanol production. Bioenergy Res 32. Chotirotsukon C, Raita M, Yamada M, Nishimura H, Watanabe
12:302–311. https://​doi.​org/​10.​1007/​s12155-​019-​09979-4 T, Laosiripojana N (2021) Sequential fractionation of sugarcane
18. Pomeroy B, Grilc M, Likozar B (2022) Artificial neural net- bagasse using liquid hot water and formic acid-catalyzed glycerol-
works for bio-based chemical production or biorefining: A based organosolv with solvent recycling. Bioenergy Res 14:135–
review. Renew Sust Energ Rev 153:111748. https://​doi.​org/​10.​ 152. https://​doi.​org/​10.​1007/​s12155-​020-​10181-0
1016/j.​rser.​2021.​111748 33. Luo Y, Li Y, Cao L, Zhu J, Deng B, Hou Y, Liang C, Huang C, Qin
19. Singh NK, Singh Y, Sharma A (2022) Optimization of biodiesel C, Yao S (2021) High efficiency and clean separation of eucalyp-
synthesis from Jojoba oil via supercritical methanol: a response tus components by glycolic acid pretreatment. Bioresour Technol
surface methodology approach coupled with genetic algorithm. 341:125757. https://​doi.​org/​10.​1016/j.​biort​ech.​2021.​125757
Biomass Bioenerg 156:12–30. https://​doi.​org/​10.​1016/j.​biomb​ 34. Thulluri C, Goluguri BR, Konakalla R, Reddy SP, Addepally U
ioe.​2021.​106332 (2012) The effect of assorted pretreatments on cellulose of selected

13
1682 BioEnergy Research (2023) 16:1667–1682

vegetable waste and enzymatic hydrolysis. Biomass Bioenerg 44. Pellera FM, Gidarakos E (2018) Chemical pretreatment of lignocel-
49:205–213. https://​doi.​org/​10.​1016/j.​biomb​ioe.​2012.​12.​022 lulosic agroindustrial waste for methane production. J Waste Manag
35. Zhang Y, Huang M, Su J, Hu H, Yang M, Huang Z (2019) Over- 71:689–703. https://​doi.​org/​10.​1016/j.​wasman.​2017.​04.​038
coming biomass recalcitrance by synergistic pretreatment of 45. Nagarajan D, Lee DJ, Chang JS (2019) Integration of anaerobic
mechanical activation and metal salt for enhancing enzymatic con- digestion and microalgal cultivation for digestate bioremediation
version of lignocellulose. Biotechnol Biofuels 12:13513. https://​ and biogas upgrading. Bioresour Technol 290:121804. https://d​ oi.​
doi.​org/​10.​1186/​s13068-​019-​1354-6 org/​10.​1016/j.​biort​ech.​2019.​121804
36. Zhang B, Khushik FA, Zhan B, Bao J (2021) Transformation 46. Chang R, Guo Q, Pandey P, Li Y, Chen Q, Sun Y (2021) Pre-
of lignocellulose to starch-like carbohydrates by organic acid- treatment by composting increased the utilization proportion of
catalyzed pretreatment and biological detoxification. Biotechnol pig manure biogas digestate and improved the seedling substrate
Bioeng 118(10):4105–4118. https://​doi.​org/​10.​1002/​bit.​27887 quality. J Waste Manag 15(129):47–53. https://​doi.​org/​10.​1016/j.​
37. Chamchoi N (2019) Utilization of fruit waste for bioethanol pro- wasman.​2021.​05.​010
duction by co-cultures of Aspergillus niger and Saccharomyces 47. Javed MU, Mukhtar H, Hayat MT, Rashid U, Mumtaz MW,
cerevisiae. Appl Environ Res 41:63–72. https://​doi.​org/​10.​35762/​ Ngamcharussrivichai C (2022) Sustainable processing of algal
AER.​2019.​41.2.6 biomass for a comprehensive biorefinery. J Biotechnol 352:47–58.
38. Jutakridsada P, Ladadok C, Kamwilaisak K (2014) Acid hydroly- https://​doi.​org/​10.​1016/j.​jbiot​ec.​2022.​05.​009
sis sorghum straw for bioethanol fermentation of Saccharomyces 48. Neto JG, Ozorio LV, de Abreu TC, Dos Santos BF, Pradelle F
cerevisiae TISTR 5596. Adv Mater Res 931:188–193. https://​doi.​ (2021) Modeling of biogas production from food, fruits and
org/​10.​4028/​www.​scien​tific.​net/​AMR.​931-​932.​188 vegetables wastes using artificial neural network (ANN). Fuel
39. Jönsson LJ, Martín C (2016) Pretreatment of lignocellulose: for- 285:119081. https://​doi.​org/​10.​1016/j.​fuel.​2020.​119081
mation of inhibitory by-products and strategies for minimizing 49. Kumari N, Garg S, Singhal A, Kumar M, Bhattacharya M, Jha PK
their effects. Bioresour Technol 199:103–112. https://​doi.​org/​10.​ (2019) Optimizing pretreatment of Leucaena leucocephala using
1016/j.​biort​ech.​2015.​10.​009 artificial neural networks (ANNs). Bioresour Technol Reports
40. Lee C, Zheng Y, VanderGheynst JS (2015) Effects of pretreatment 7:100289. https://​doi.​org/​10.​1016/j.​biteb.​2019.​100289
conditions and post-pretreatment washing on ethanol production 50. Esfandyari M, Amiri M, Koolivand-Salooki M (2015) Neural
from dilute acid pretreated rice straw. Biosyst Eng 137:36–42. network prediction of the Fischer-Tropsch synthesis of natural
https://​doi.​org/​10.​1016/j.​biosy​stems​eng.​2015.​07.​001 gas with Co(iii)/Al2O3 catalyst. Chem Eng Res Bull 17:25–33.
41. Marsden W, Gray PP, Nippard GJ, Quinlan MR (1982) Evaluation https://​doi.​org/​10.​3329/​cerb.​v17i1.​22915
of the DNS method for analysing lignocellulosic hydrolysates. J
Chem Technol Biotechnol 32:1016–1022. https://d​ oi.o​ rg/1​ 0.1​ 002/​ Publisher's Note Springer Nature remains neutral with regard to
jctb.​50303​20744 jurisdictional claims in published maps and institutional affiliations.
42. Harner NK, Wen X, Bajwa PK, Austin GD, Ho CY, Habash MB
(2015) Genetic improvement of native xylose-fermenting yeasts Springer Nature or its licensor holds exclusive rights to this article under
for ethanol production. J Ind Microbiol Biotechnol 42:1–20. a publishing agreement with the author(s) or other rightsholder(s);
https://​doi.​org/​10.​1007/​s10295-​014-​1535-z author self-archiving of the accepted manuscript version of this article
43. Sahu S, Pramanik K (2018) Evaluation and optimization of is solely governed by the terms of such publishing agreement and
organic acid pretreatment of cotton gin waste for enzymatic applicable law.
hydrolysis and bioethanol production. Appl Biochem Biotechnol
186(4):1047–1060. https://​doi.​org/​10.​1007/​s12010-​018-​2790-7

13
BioEnergy Research is a copyright of Springer, 2023. All Rights Reserved.

You might also like