Energy 242 (2022) 122949

Contents lists available at ScienceDirect

Energy
journal homepage: www.elsevier.com/locate/energy

Combustion and pyrolysis kinetics of Australian lignite coal and

validation by artificial neural networks
SP Sathiya Prabhakaran a, *, Ganapathiraman Swaminathan b, Viraj V. Joshi a
a
Department of Energy and Environment, National Institute of Technology, Tiruchirappalli, 620015, Tamil Nadu, India
b
Department of Civil Engineering, National Institute of Technology, Tiruchirappalli, 620015, Tamil Nadu, India

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

Article history: Lignite (AL), with a calorific value of 5.9 MJ/kg is the most abundant low-rank coal used widely in power
Received 16 February 2021 generation. AL's combustion and pyrolysis characteristics were investigated to provide scientific findings
Received in revised form using thermogravimetric analysis under non-isothermal conditions. Methods utilized in the kinetic
16 December 2021
investigation included Vyazovkin, Flynn-Ozawa-Wall (FOW), Kissinger-Akahira-Sunose (KAS), Freidman,
Accepted 17 December 2021
Available online 27 December 2021
Doyle, Arrhenius, Freeman-Caroll, and Sharp-Wentworth. Multiple heating rate methods delivered the
activation energy (Ea) as 194e211 kJ/mol (combustion) and 450e470 kJ/mol (pyrolysis). Combustion
process followed two dimensional diffusional reaction (2D), volume contracting (R3) solid-state reaction
Keywords:
Combustion
mechanism models and pyrolysis followed volume contracting (R3) as determined by CR (Coats-Red-
Pyrolysis fern), KC (Kennedy-Clark) methods. Master Plot method validated the mechanisms and concluded that it


Combustion indices is of deaccelerating type. Improper combustion at a higher heating rate (50 C/min) was indicated with
Kinetics an increase in burnout Tb, ignition Ti, peak Tp temperatures. Combustion indices (CHCI, IG, IB) reported
Artificial neural network highest values of 4.99 E10 mg2 min2 OC3, 4.19E-05 mg2 min3, 2.65 mg2 min4 at lowest heating
Solid-state reaction model rates. AL's analytical thermal degradation behavior results were validated using artificial neural networks
with best-fit models NNA 7,8. The research study offers a useful guide for spontaneous AL combustion
and pyrolysis prediction on site.
© 2021 Elsevier Ltd. All rights reserved.

1. Introduction structure of the pore, and the specific surface area during coking are
largely affected by this process [5]. The temperature and heating
Combustion is the driving force for power generation in thermal rate of pyrolysis affect the gasification products, and in hydro-
and hybrid industries. Combustion of fuel plays a pivotal role in liquefaction, it affects the cleavage of macromolecular bonding
transportation, industrial, and domestic requirements [1]. Unless into fragments [6]. Pyrolysis is the backbone for clean coal con-
and until the renewable sources of energy on earth are not fully version technologies. It helps to relate the soot generation rate in
exploited in terms of their potential for power generation, the de- coal combustion to tar generation rate and thus predict environ-
pendency of human beings on fossil fuels is seldom ending [2]. mental benefits [7]. Simultaneous series and parallel reactions, and
Because of fossil fuel combustion, global warming has turned into a complex coal structures make it difficult to understand the pyrol-
serious problem. The most significant contributor to global ysis process [8]. When developed economies are concerned, the
warming is CO2 from power generating systems or processes [3]. technological advancement and good quality of fuel help achieve
The power plants which run on coal as fuel adds a share of 33e40% their target of carbon capture and storage. However, for developing
worldwide of carbon emissions, which are anthropogenic [4]. Py- economies of the world, the scenario is different [9]. The present
rolysis is a predominant step in the processes like coking, gasifi- rate of discovery of crude oil is approximately half of that of pro-
cation, liquefaction etc. The reactivity of char and cokes, the duction of fuel, and a downward trend of oil production is supposed
to be observed from 2020 to 2030. This will pave the downward
trend in oil production that will change the economy and invoke
alternative fossil fuels to be used as the primary fuel in the power
* Corresponding author. Department of Energy and Environment, National
Institute of Technology, Tiruchirappalli, Tamilnadu, 620015, India. generation process via the combustion route [10]. Due to the cur-
E-mail addresses: sathiyanitt1491@gmail.com1 (S.S. Prabhakaran), gs@nitt.edu rent rate of greenhouse gas emissions, the temperature of the globe
(G. Swaminathan), virajjoshi91@gmail.com (V.V. Joshi).

https://doi.org/10.1016/j.energy.2021.122949
0360-5442/© 2021 Elsevier Ltd. All rights reserved.
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

is headed to 3.2
Celsius (pre-industrial levels), as per United Na- cutting mill was employed in the size reduction of the sample (Dry,
tions Environmental Programme report. The target for all the Sieve, Mesh 500 mm). A muffle furnace was used in the ash pro-
countries that are a part of the Paris Climate Change pact is to limit duction of the coal sample. Oven-dried size reduced (known
global warming to 1.5e2
Celsius. Global greenhouse gas emissions amount) sample in a silica crucible was taken and kept in the
should fall by 7.6% to achieve this target, as reported by the United furnace at 1100
C for 4 h. The residual (ash) was then subjected to
Nations Environmental program report 2019 [11]. There are several intensive analytical experimentation. Different methods employed
methods being adopted for the same, but a few dominant ones are in the kinetic investigation are further discussed in section 2.2. The
a) improving the overall fuel utilization rate in the process precision of the instrumentation and the uncertainty measurement
employed in the power plant, b) working on hybrid cycles of results are made available in supplementary section Table S1. Fig. 1
thermodynamics for power generation, c) hydrocarbon fuels to be represents the research procedure adopted for the conventional
replaced by hybrid fuels or renewable energy sources and d) fuel under study.
employing carbon capture technology and its conversion to eco-
friendly fuel, and many other methods available [12]. The carbon- 2.2. Kinetic Studies
dioxide emissions from the power plants can be reduced by
increasing the overall efficiency of the conversion process The determination of kinetic parameters was done in accor-
employed in coal-based power plants [13]. Coal decomposes by dance with recommendations provided by International Confed-
giving out vapors that are organic upon combustion or devolatili- eration for Thermal Analysis and Calorimetry (ICTAC) [19,20]. In
zation or gasification, and one of the best tools to understand this order to discover whether non-Arrhenius temperature dependence
conversion process is thermogravimetric analysis (TGA) [14]. The exists or not, experimentation for a minimum of three different
parameters of the kinetics of the coal conversion process, which are heating rates or more is needed. Rates of heating or temperature
the key players in improving the efficiency of the overall process, (Extreme) dominate the actual accuracy of kinetic parameters as
can be determined from the weight loss versus temperature graphs defined by Arrhenius, and hence their determination at these ex-
of the coal when it is subjected to different heating rates [15]. The tremes is imperative. Thus, integral, differential, iso-conversional
optimum conditions of the reaction and an optimum design of methods were employed in the study of kinetics [19e21]. TGA
combustion equipment can be obtained using these parameters. analysis (experimental explanation) along with different kinetic
Different types of coal exhibit distinct thermal behavior when methods employed and its derived equations are discussed in
subjected to different heating rates under different atmospheres sections 2.2.1 to 2.2.10 [23e46]. The base equation and its deriva-
[16]. Kinetic modeling carried out in the present study was adapted tion can be referred from supplementary section (Kinetic Studies
from petroleum coke's kinetics under non-isothermal conditions base equation).
and was validated using artificial neural networks [17]. The same AL
coal was blended with paint sludge and its co-combustion perfor-
mance was evaluated [18]. The main goal of the study was to 2.2.1. Kissinger-Akahira-Sunose method (KAS)
investigate the kinetics of the AL by free model and based on model The interaction between activation energy and the heating rate
methods for combustion and pyrolysis processes. Thermogravi- at which the experimentation is carried out is proposed by the


metric investigation at 10, 20, 50 C/min of heating rates was car- linear integral for which the system is not at a constant tempera-


ried out. 50 C/min rate of heating was purposely selected to ture is expressed in the form of equation as follows
understand the thermal degradation behavior and kinetics at the
  
b AEa Ea
higher heating rate and to add novelty to the paper. Since, very few ln ¼ ln  (1)
research articles are available for higher heating rate kinetic study. T2 GðaÞ RT


This experimental analysis aids for efficiency improvement of in- b
A plot of ln versus 1 for different heating rate helps us in
dustries employing AL as primary combustion or pyrolysis fuel. Tm 2 T
Research carried out aids in the provision of optimum parameters calculating the activation energy [22]. T is the maximum temper-
for such systems in order to enhance their effective utilization. The ature of the weight loss curve. The slope of the above equation is
interpretation of results with respect to thermodynamic parameter Ea
equal to R .
(Figs. 1d and 2d), Reaction mechanisms (Figs. 4 and 5) and indices
of combustion (Table 5) aids in the provision of optimized solution
for the lignite combustion and pyrolysis efficiency design problems. 2.2.2. Freidman method
Kinetic analysis of a low grade fuel by a variety of kinetic models is This method is a model-free method. It replaces the kinetic
one of the innovation aspect of this research. Analysis of such a low functions which are based on the rate of heating with the functions
grade fuel with respect to performance of combustion via study of based on rates of differential conversion. This equation is obtained
indices, reactivity via TGA study, stabilized combustion behavior by taking the logarithm on both sides of equation (8). The equation
via solid-state reaction models and thermal degradation behavior is expressed as follows
(for both combustion and pyrolysis process) validation by multi-

da Ea
disciplinary tool like neural networks artificial also adds to the ln ¼ lnA þ lnf ðaÞ  (2)
innovation trait. The study helps researchers working technically dt RT
on conventional fossil fuels. It also offers useful guide for sponta-

A plot of ln da versus 1 is used for the determination of acti-
neous AL combustion and pyrolysis prediction on site and off site. dt T
Ea
vation energy from the slope given by R .
2. Materials and methods

2.1. Australian Lignite coal, Ash, instruments 2.2.3. Broido method

This is a very simple method for the calculation of activation
Instrumentation, methods, and models employed in the energy. It takes into consideration a single heating rate at a time
research were reported in Table 1. The cement plant in Ariyallur, [23]. The activation energy at each stage of degradation can be
Tamil Nadu, India, allowed a local collection of the coal sample. A calculated by the following equation
2
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 1. Research procedure.

 2.2.4. Flynn-Ozawa-Wall method (FOW)

1 Ea This method is the one that employs approximation of Doyle for
lnln ¼ þ lnA (3)
Y RT evaluation of energy of activation. The kinetic parameters of the
reaction system does not depend on the rate of heating and a plot of
wo  wt 1 versus lnb is used for the calculation of activation energy [24]. The
Y¼ T
wo  wf expression for the above method is given as


1 1  

A plot of lnln versus gives us an approximation for the AEa Ea
Y T
ln b ¼ ln  5:331  1:052 (4)
Ea GðaÞ RT
broido method which is linear and the slope given by R helps in
the determination of activation energy.
3
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 2. Curves TGA (a), DTG (b), Conversion (c), Energy of activation bars (d) Combustion.

Table 1
Methods, models, instrumentation.

Methods Models Instrumentation

ASTM D3906 XRD rigaku ultima-4 XRD with Cu-Ka radiation

ASTM F1372 Zeiss Evo 18 SEM SEM
ASTM D4464 Horiba LA-960 model Particle size measurement based on light scattering principle (laser based).
ASTM C702 Retsch SM-300 Cutting mill with sieve of mesh size 500 mm.
ASTM D7348 Carbolite AF-1100 Muffle Furnace with process time of 4hr @ 1100 oC.
ASTM E1131 Perkin-Elmer TGA 8000 TGA with O2 atmosphere and 20 ml/min gas flowrate in the temperature range of 30
e900
C along with 10, 20, 50 oC/min heating rates.
USEPA 2009, SW-846, 3051 A, Thermos-Fischer 7000 ICP-OES, Milestone Acid digestion @ 200
C and 600 psi for 24 h followed by ICPOES.
6010C EPHOS one Microwave digester
ASTM D5291 & APHA 22nd Elementar Vario EL cube Ultimate Analyzer & Ultimate & Proximate Analyzers
Edition, 2012, 2540 G Leco 501 Proximate Analyzer
ASTM D 5468e02 (2007) IKA C 6000 Bomb Calorimeter Bomb calorimeter for calorific value analysis.
ASTM E168, UATR FTIR spectrum 2 Perkin-Elmer FTIR analyzer

2.2.5. Coats-Redfern method (CR) compute activation energy, and by using the energy of activation,
The coats-Redfern method helps us in the determination of ki- the pre-exponential factor is calculated [23].
netic parameters and for proper analysis of thermal degradation
behavior of the material with rest to reaction mechanism. The 2.2.6. Kennedy-Clark method (KC)
equation is as follows The method is proposed by Kennedy and Clark and can be
described below as follows

  
GðaÞ AR Ea

ln ¼ ln  (5) b*GðaÞ Ea
T2 b*Ea RT ln ¼ ln½A  (6)
T  To RT


GðaÞ
The slope of the plot between ln T2
and 1/T is used to The heating rate under consideration allows us to select To. This
4
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

is the temperature at which reaction occurs and is usually consid- supplementary section (Vyazovkin Method). To compute Ea


following expression is required to be minimized
ered to be 20
C. The plot of ln bTTo
*GðaÞ
vs 1=T will yield a straight
  
line with slope equal to (-Ea/R) and intercept being equal to lnA X
n Xn I Ea Ta;i *bj
∅ðEa Þ ¼    (11)
[25,26].
i isj
I Ea Ta;j *bi

2.2.7. Freeman-Carroll method (FC) where functional of integral isoconversional method is given by
A method that can be used without any priori knowledge of ∅ðEa Þ, Ea is the energy of activation for each a and Ta;i being the
reaction order was suggested by Freeman-Carroll. The equation corresponding temperature for ith heating rate, n is the number of
describing the method is written below as follows heating rate. The pre exponential factor is then calculated by the
2   3 2 0
 13 following equation
6
Dln ddta 7 6 B
D 1T C7
6 7 ¼ 6  Ea*B C7 bE
@R*Dlnð1  aÞA5 þ n (7) E
4Dlnð1  aÞ5 4 A¼ *eTmax (12)
2 *f 0
RTmax amax

The plot of LHS of above equation versus RHS yields a straight

line with slope equal to Ea. This method is used for accurate pre-
2.2.12. Ozawa-flynn-wall (OFW) method
diction of activation energy [27,28].
This method is the one that employs approximation of Doyle for
evaluation of energy of activation. It is an iso-conversional model-
2.2.8. Sharp-Wentworth method (SW)
free analysis method [38,39]. The expression for the above method
This method is a superior form of the method suggested by
is given as (at a constant heating rate b)
Freeman-Carroll. The equation is as follows
 

2 3 AEa Ea
da , ln b ¼ ln  lnðfxÞ  5:331 þ 1:052 (13)
  R RT
6 dt 7
ln6 7 ¼ ln A  Ea ðR*TÞ (8)
41  a 5 b After kinetic investigation through the above-discussed
methods, the combustion process was scrutinized using the study
of indices of combustion in section 2.3.
a is the fraction of material decomposed and A, b and Ea have
their usual meanings respectively [29,30].
2.3. Parameters of combustion
2.2.9. Master plots method (MP)
Loss in weight (Wm, mg/min, maximum rate), Loss in weight
This method is used to validate the reaction mechanism as that
(Wa, mg/min, average rate), Temperature (Pt,oC Peak, It,oC Ignition,
obtained from CR and KC method. The equation representing the
Bt,oC Burnout) were the parameters used to understand the prop-
method is as follows:
erties for the process of combustion. Index of comprehensive
GðaÞ pðuÞ combustibility (CHCI), index of ignition (IG), index of burnout (IB)
¼ (9) were determined to evaluate combustion process performance.
Gð0:5Þ pðu0:5Þ
ð  Wm *  Wa Þ
the possible model can then be obtained by plotting theoretical CHCI ¼ (14)
GðaÞ It2 *Bt
Gð0:5Þ versus a for sixteen models as reported for CR and KC methods
pðuÞ
respectively against the pðu0:5Þ versus a for experimental results and ð  Wm Þ
IG ¼ (15)
then comparing as per equation (9). For a certain value of a, both ti *tp
sides of equation (9) will converge. They may not converge if
models are inappropriate with respect to G(a) in either of the case ð  Wm Þ
respectively [31e33]. IB ¼ (16)
tp *tb *t1=2

2.2.10. Doyle method The intersection point of the TGA curve (Horizontal line) and
This is an approximate integral method which does not take into DTG curve (Tangent Line) is It, whereas temperature (Tb, 98% loss in
consideration f(a) function and thus avoids deviation resulting weight) and time with respect to temperature were represented by
from selected reaction mechanism [34]. It helps for testing energy ti,tb, tp [25,40]. The results of thermogravimetric investigation and
of activation obtained by hypothesized reaction mechanisms indices of combustion acted as base data for the application of NNA,
[35,36]. The equation is as follows: which was further discussed in section 2.4.
 

AEa Ea Ea
ln b ¼ ln  2ln  (10) 2.4. Neural networks Artificial (NNA)
R*f ðaÞ RT RT
An equivalent of human brain function along with complex
a plot of 1T versus lnb is used for the calculation of activation energy pattern recognition quality is called NNA [41]. Agriculture, finance,
by this method. security, management, medical science, engineering, science, edu-
cation, commodity, trading, and art, along with difficulties in the
2.2.11. Vyazovkin Method area of manufacturing, marketing transportation, computer secu-
This is one of the most widely used modern iso-conversional rity, insurance, energy, and many more find applications of NNA
methods. It allows computing Ea of a reaction system without [42]. NNA domain challenges revolve around 1) capability
knowing its model [25,37]. The derivation can be referred from improvement (model robustness increment), b) incrementing
5
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

model transparency, extraction of knowledge enablement from nitrogen and Sulphur content help in the reduction of GHG's. The
trained NNA's, c) incrementing extrapolation ability, d) adaptation high fixed carbon.
of uncertainty in models. NNA can be classified as follows a) Content of the coal showed that it had a greater gasification
Feedforward NNA (FFNNA), b) Feedbackward NNA (FBNNA) [43]. tendency. The high ash generation of this coal might be a disad-
Single-layer perceptron, multilayer perceptron, radial basis func- vantage, but after ICP-OES analysis, it was confirmed that the ash
tion network forms the other type of Feedforward networks. generated is highly beneficial in co-processing technology for
Bayesian regularized NN, Kohonen's self-organizing network, cement manufacturing [46]. After retrospection, it was found that
Hopfield networks, competitive networks, art networks form the C, H, N, S, O content respectively was within the tolerance limit
feedback type. Each unit in a layer in FFNNA is connected with all for Australian lignite coal [47]. The minerals like silica oxides,
other units in other layers. The weight associated decides the aluminum oxides, ferrous oxides were found in high concentration
knowledge associated within the layer. These units are called from ICP-OES analysis of the sample (Table 3). The ratio of fixed
nodes. Coordination within nodes produces sequential graphs in carbon content to volatile matter content is called the fuel ratio. The
FBNNA. NNA is a tool that helps the researchers in a virtual simu- fuel ratio value for the coal was low and suggested better com-
lation of experiments, provided initial sets of the experiment are bustion properties with respect to other lignite coals [48]. Thus,
carried out [44]. 1) TRANSIG, 2) LOGSIG, 3) PURELIN (mathematical Australian lignite coal was a potential fuel with a distinctive com-
functions) allow the processing of input data (Fig. 8). The trans- bination of minerals [49,50]. Acid digestion for the ash of the
sigmoidal function accepts infinity values (-,þ) and pours out in the Australian lignite coal sample was carried out (200
C and 600 psi
range of 1 to þ1. Faster networks prefer it. The output of sigmoidal pressure @ 24 h). Watmann filter paper was used for filtering the
(logistic) is in the limit of 0e1. Linear is the 3rd, and the output layer sample, and the ICP-OES system was injected with the filtrate.
seldom has its application. NNA predictions were carried out by Table 3 reported the evidence of the availability of significant metal
MATLAB R2019b [44,45]. oxides in lignite (after ashing). Materials with catalytic properties
Validation of the TGA experiments conducted was confirmed by like nickel, copper, lead, manganese, Barium, Chromium were
the application of NNA. The input data for the NNA was the TGA present in high proportions (Table 4) [7]. Ash carried a high amount
data for different heating rates along with temperature and gas of oxides of silica, alumina, ferrous, and other materials. The char-
atmosphere, respectively, for combustion and pyrolysis. The output acteristics of any material decide its thermal degradation behavior
from NNA was the mass loss data. It validated the effect of input along with its kinetics. This was further elaborated in section 3.2.
parameters on output parameters. Table 6 reported the outputs
obtained for derived single layer perception and multiple layer
models. For models like SLP (Single layer perception), the topology 3.2. Lignite combustion (thermal degradation behavior)
of the network (3*1*1) indicated the presence of three neurons in
the first (input) layer followed by a single neuron in the 2nd and From the TGA curve (Fig. 2a), the thermo-oxidative degradation
output layer. For MLP (Multiple layer perception), two different of Australian coal had an induction period wherein oxygen satu-
models were tested, namely MLP1 and MLP2. MLP1 (3*8*4*1) rates the sample. This further boosted activity of autocatalysis along
consisted of three, eight, four, one neuron in the input, first, second, with concentration rise for peroxy groups in the sample. After the
and output layer (2 layer model). In contrast, MLP2 (3*8*4*1*1) saturation point of concentration was reached, the termination
consisted of three, eight, four, one, one neuron in the input, first, stage was activated and accompanied by the termination of peroxy
second, third, and output layer (3 layer model) [17]. The perfor- groups. A temperature rise facilitated oxygen absorption. The sur-
mance and transfer functions deployed in the process were MSE face of the coal sample, along with the coefficient of diffusion,

and TRANSIG. The goal of the performance is kept at 107 along decides this rate. At the 10 C/min rate of heating, from the TGA
with thousand of iterations for each model. The output layer had curve, it was observed that the loss in weight for the temperature
linear neurons, while activation neurons (non-linear) were present range of 30e105
C is 4.23%, which was evidence of the first stage of
in the input, hidden layer(s). NNA7 proved to be the best fit NNA degradation of coal representing the moisture loss of coal. The
model for the process under consideration (Table 6). The input layer second stage of degradation was evident in the temperature range
and hidden layer(s) had non-linear activation neurons, and the of 105e380
C, wherein the low volatile organic carbon matter was
output layer had linear neurons. 11 different models were tested. lost, the weight loss observed was 7.68%. The third stage of
The first nine models included testing of SLP, MLP models, with degradation, which was peak degradation, was observed in the
each being subjected to MSE, MSEREG, SSE transfer functions. temperature zone of 380e600
C, wherein all the volatile matter
(organic) was lost, and the weight loss was 82.64%. The last stage of
degradation was observed in the temperature range of 600e900
C,
3. Results and discussion and the weight loss observed was 5.44%. Suggesting thermal sta-
bility was achieved. For the same heating rate, from the DTG curve
3.1. Characteristics of Australian lignite coal (Fig. 2b), various characteristic weight loss peaks were observed.
This represented the conversion of various components present in
Table 2 reported the values for analysis (Ultimate, Proximate) of the coal to gaseous form. The components evolved consisted of
the coal. The high carbon and hydrogen content of the coal sug- aromatic compounds, nitrogen oxides, carbon-di-oxide, sulphur
gested that it is a fuel with potential calorific value. Whereas low oxides, ammonia, carbon mono-oxide etc. [51]. The first

Table 2
Analysis of Australian lignite.

Analysis Proximate Ultimate Gross Calorific Value

Details

Sample M VM % Ash % FC % FR C H N S O (GCV) kJ/kg

Details % % % % % %

AL 3.75 28.61 25.45 42.19 1.47 64.72 2.83 1.2 0.593 30.657 5999.05

6
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Table 3
ICP-OES Oxide analysis.

Analysis ICP-OES Analysis

Details

Sample SiO2 Al2O3 Fe2O3 CaO MgO SO3 K2O Na2O TiO2 Mn2O3 P2O5
Details

AL 56.16 21.85 7.54 6.15 0.77 2.3 0.2 0.2 1.20 0.05 0.84

Table 4 resulting in a dual process. DTG curve (Fig. 3b) for pyrolysis clearly
ICP-OES Heavy Metal analysis. indicates that there is only a single zone of degradation for
Analysis ICP-OES Analysis respective heating rates employed. A vertical shift is observed in the
Details characteristics peaks as the rate of heating decreases. Thus the
Metals Cu Mn Ni Pb Se As Ba Be Co Cr As lower heating rate is suitable for slow pyrolysis and clean coal
conversion process. Fig. 3c shows the variation in conversion ob-
Unit (mg/kg) 38 234 95 91 0.01 0.01 434 2.01 25 327 0.01
tained for the pyrolysis process. With a lower heating rate, con-
version takes place at a lower rate, and maximum conversion is
obtained in the temperature zone of 400e700
C. At a higher
characteristic derivative weight loss peak was observed in the
heating rate, the conversion rate increases. Thus, it was inferred
temperature zone of 50e100
C, followed by the second peak in the
that the coal pyrolysis and combustion process took place in almost
range of 200e300
C and the third and the most prominent peak in
the same degradation zone. Pyrolysis decreased conversion rate as
the range of 450e600
C. At 20
C/min, the TGA curve shifts up
most of the reactive components became non-reactive due to the
vertically, and the ash generation jumped off to 18.5%, and the zone
inert atmosphere. But, pyrolysis favored the utilization of coal's
of degradation was in the temperature range of 380e620
C. From
volatile matter, which has potential calorific value and resulted in
the DTG curve, the temperature zone for the first and second
high char generation compared to combustion. This is a specific
characteristic derivative weight loss peaks remained the same as
case with respect to low-grade coal. It varies with the rank of the
for the 10
C/min heating rate, but the third one expanded in the
coal and the volatile content present. Thermal degradation char-
zone of 380e620
C. At 50
C/min, the TGA curve shifts up vertically
acteristics of any material are driven as a result of its kinetic pa-
above the curve for 20
C/min, and the ash generation was up to
rameters, which were discussed in detail in section 3.3.
27.2%. The rest zones of degradation remained the same as that for
the 10
C/min rate of heating. From the DTG curve for 50
C/min
heating rate, the first characteristic peak of degradation was in the 3.3. Kinetic study (combustion and pyrolysis)
zone of 50e180
C, and the second peak was in the temperature
zone of 325e650
C. Upon combustion at variable heating rates, The analysis of the thermo-oxidative and pyrolysis degradation
Australian lignite coal exhibited characteristic thermal degradation behavior of the Australian lignite coal kinetics was highly beneficial
behavior [52]. The predominant catalytic reactions in the Australian for optimizing all the processes and the equipment employed in the
lignite coal during combustion were desulfurization (Hydro), process [55,58]. The slopes required for activation energy calcula-
denitrogenation (Hydro), aromatic liquid production via hydro- tion were computed from plots represented in Fig. 4. The assess-
fining, cracking, hydrogenation. Methane production via Fischer- ment of kinetic parameters (Ea, K, n) for the coal combustion at
Tropsch reactions, liquids of hydrocarbon, CO/H2 process of syn- different heating rates required selecting a proper mathematical
thesis in the production of chemicals and alcohols constituted other model. It also aids in the differentiation between methods (integral,
reactions. Shift reaction (water-gas) and hydrogenolysis, isomeri- differential, iso-conversional) with respect to thermodynamic pa-
zation, dehydrogenation, dehydrocyclization, hydrogenation via rameters. For computing kinetics through FOW, KAS, Freidman, and
hydroforming, reforming, and cracking were other predominant Doyle models, conversion of mass was required for different heat-
reactions in the process. The end result was the formation of ash, ing rates and temperatures (Fig. 2c).
which was the mineral content of coal. It is a combination of By the FOW method (Fig. 2d), the change in activation energy
minerals (sulfate, silicate, carbonate, alumino-silicates, sulfide, and with respect to conversion indicated that from 0.3 to 0.6 conver-
others) [53e57]. In the same temperature zone (400e600
C), the sion, the activation energy required for the coal remained almost
length of characteristic peaks decreased with an increase in heating constant. The peak activation energy requirement was at 0.7 con-
rate. This acted as evidence for the improper combustion behavior version and was equal to 245.18 kJ/mol. At 0.2 and 0.8 values of
of coal at higher heating rates. During pyrolysis, the breaking of conversion, a rise and fall in the energy of activation were visible.
covalent bonds results in the devolatilization of fragments, which is The same kinetic phenomena were exhibited by the KAS method.
collected as mass loss in the TGA experiment. From the pyrolysis The peak activation energy requirement is reported at 0.7 conver-
TGA curve (Fig. 3a), it is inferred that the coal char generation is sion value and equal to 243.62 kJ/mol. By the Freidman method
directly proportional to the heating rate employed. Lower the (Fig. 2d), it was visible that low values of conversion required high
heating rate lower is the char generation. The highest char gener- energy of activation. Highest value (0.2, 259.81 kJ/mol) was re-
ation of 74.80% is observed for a 50
C/min rate of heating followed ported. After 0.2 conversion, the energy requirement remained the
by 20,10
C/min. The peak degradation occurred in the temperature same for the rest of the conversion.
limit of 400e550
C for 10
C/min and 20
C/min rate of heating, Friedman showed a distinct behavior in consideration for values
but for 50
C/min, it extends above 550
C. At higher heating rates, of conversion and accordingly energy of activation because of
diffused oxygen in the sample participates in the degradation, inducement of differential conversion rate function. Since the FOW

7
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 3. Curves (TGA (a), DTG (b), Conversion (c), Energy of activation bars (d) Pyrolysis).

method was derived from Doyle, the variation of activation energy heterogeneous composition of Australian lignite coal and depicts
was very similar in nature for Doyle. For the values of conversion that collision reactions involved were highly complex in nature.
where high Ea was visible endothermic reaction behavior was fol- Activation energy calculations were also done by four single heat-
lowed. The conversion values where low Ea was noted favored ing rate methods, namely Arrhenius (modified form), Broido,
exothermic behavior. The average values of activation energy ob- Freeman-Carroll, Sharp-Wentworth, and the values were reported
tained by FOW, KAS, Freidman, Doyle methods were 194.46, 191.78, in Tables S6 and S7 (supplementary data). The best fit single heating
196.94, 211.12 kJ/mol, respectively, and the closeness of values by rate method was FC because the values of energy of activation were
four methods indicated the accuracy of the computed kinetics. For in alignment with that obtained from multiple heating rate
pyrolysis, the energy of activation showed sinusoidal behavior methods. The values obtained by Broido were imprecise. The
(Fig. 3d). The highest values were obtained at 0.3, 0.6 conversion reason was that it proposes that the combustion reaction of
values. This tendency followed respectively for FOW, KAS, and Australian lignite coal is a single step reaction dominantly, but
Doyle methods. Freidman showed exactly the opposite behavior actually, it's a highly complex and multi-step process. Broido
and reported a linear rise in activation energy requirement with an method was crude and also presumed the whole process was linear
increase in conversion. The respective average values of activation in nature. In addition to the above-mentioned methods, the two
energy obtained by FOW, KAS, Freidman, Doyle methods were most widely used modern iso-conversional methods, namely
447.81, 473.51, 408.04, 471.15 kJ/mol. The slopes required for Ea ‘Vyazovkin and OFW,’ were employed in the determination of Ea.
calculation were computed from plots represented in Fig. 7. Com- For the combustion process, the highest value of the energy of
plex series-parallel competitive reaction phenomena were activation was obtained at 0.6 conversion value (152.06 kJ/mol
responsible for the erratic behavior of energy of activation under Vyazovkin, 177.54 kJ/mol OFW), whereas for the pyrolysis process,
nitrogen and oxygen atmosphere for combustion and pyrolysis it was at 0.3 conversion value (290.82 kJ/mol Vyazovkin, 314.52 kJ/
process, respectively. The factors (K) calculated by FOW, KAS, mol OFW). For the combustion process, both methods showed an
Freidman, Doyle methods was in the range of 1010 to 1014 min1 increase in energy of activation up to 0.6 value and then showed a
(combustion), 1011 to 1051min1 (Pyrolysis), and variation in the decrease in the value. Also, for the process of pyrolysis, the energy
factors (K) obtained by the four methods was due to highly of activation decreased with an increase in conversion. The average

8
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 4. Calculation of Activation Energy (a,b,c), VZ and OFW model (d) (Combustion).

energy of activation by these methods was 139.77 kJ/mol Vya- reaction mechanisms followed in the process were of R3, D2
zovkin, 164.41 kJ/mol OFW (combustion) and 211.16 kJ/mol Vya- models. R3 model (contracting volume reaction mechanism) was
zovkin, 236.44 kJ/mol OFW (Pyrolysis). The difference between the followed in the first half of the conversion range, while the D2
energy of activation computed by these iso-conversional methods model (two-dimensional diffusional reaction mechanism) takes it
and other methods was due to limitations associated with each up from the second half (Fig. 5). This suggested that the combustion
respective method. The accuracy of methods (FC, SW, Arrhenius) in of lignite coal follows a multi-step reaction mechanism. After
comparison to methods (FOW, KAS Friedman, Vyazovkin, OFW) retrospection, it was evident that Ea values by FOW, KAS, Freidman,
was less for the determination of kinetic parameters. In the case of and Doyle methods were close to Ea given by the P2/3 model (Power
model-based methods, the calculation of Ea was filled with errors as law reaction mechanism). For the 10
C/min rate of heating, the Ea
the regression line didn't fit data points accurately. A regression line value by the P2/3 model was 175.96 kJ/mol and respectively at 20
(precise, accurate) that could match maximum data points was and 50
C/min (188.26 and 203.07 kJ/mol), which mirrored
obtained for methods (Multiple rates of heating). Which reaction respective values by FOW, KAS methods, and so was identified as an
mechanism is predominant in the lignite coal combustion and optimal mechanism. The porous nature extent of coal is seldom
pyrolysis was identified using CR, KC, and also validated by Master related to reaction mechanisms (solid-state). For the pyrolysis
plot method (Figs. 5 and 6). The reaction mechanism models (solid- process, the entire range of conversion followed R3 solid-state re-
state) were reported in Table S2 of supplementary data, and the action mechanism (Fig. 6).
values for these models were reported in Tables S3 and S4 by CR, KC This showed that pyrolysis of coal takes place by contracting the
methods in the supplementary data, respectively. By CR, KC, and volume reaction mechanism. Supplementary data (Tables S9 and
master plot method, it was observed that despite subjecting the S10) reported the values for different reaction models. Three ma-
coal at different heating rates for the combustion process, the prime jor classifications of solid-state reaction mechanisms are 1)

9
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 5. CR graphs (a,b,c) KC graphs (d,e,f) (Combustion).

Fig. 6. CR graphs (a,b,c) KC graphs (d,e,f) (Pyrolysis).

10
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 7. Calculation of Activation Energy (a,b,c), VZ and OFW model (d) (Pyrolysis).

Fig. 8. Performance functions used in NNA models.

Sigmodial, 2) Deaccelerating, 3) Accelerating. If the activation en- Australian lignite coal for both of these processes increases with an
ergy of a model reaches its peak value at the intermediate stage, it is increase in conversion as the energy of activation decreases with
sigmoidal. If the same decreases with conversion, it is deacceler- conversion and shift in the reaction mechanism is observed (R3, D2
ating, and the mechanism showing the opposite behavior of the & R3) for change in process (combustion & pyrolysis) respectively
deaccelerating type is called an accelerating mechanism. In the case [59,60]. The values computed by different kinetic methods are re-
of combustion and pyrolysis processes, respectively, all the models ported in Table (S5eS8) supplementary data.
fall into the deaccelerating type mechanism. Thus, the reactivity of

11
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Table 5
Combustion performance indices.




Material AL @ 10 C/min AL @ 20 C/min AL @ 50 C/min

Burnout Temperature ‘Tb’ (oC) 698.12 616.66 654.43

Ignition Temperature Ti (oC) 414.39 387.66 444.2
Peak Temperature Tp (oC) 508.9 517.81 532.69
Loss in weight (Rm,mg/min, Maximum rate) 0.5594 1.214 0.9706
Loss in weight (Ra, mg/min, Average rate) 0.078 0.1952 0.1494
CHCI (mg2 min2 OC3) 4.99E-10 3.05E-09 1.38E-09
IG (mg2 min3) 4.19E-05 2.59E-03 1.11E-02
IB (mg2 min4) 2.65E-06 2.65E-04 3.51E-03

Table 6
Derived SLP and MLP NNA's for mass loss (combustion and pyrolysis).

ANN Outputs R2

Models Type Network topology Performance & Transfer Function Input Train Validation Testing

NNA1 SLP 3*1*1 MSE(TRANSIG) 1.Atmoshpere 0.9676 0.9705 0.9683

NNA2 3*1*1 MSEREG(TRANSIG) 2.Heating Rate 0.9688 0.9624 0.9683
NNA3 3*1*1 SSE(TRANSIG) 3.Temperature 0.9680 0.9668 0.9680
NNA4 MLP1 3*8*4*1 MSE(TRANSIG) Output 0.9998 0.9998 0.9998
NNA5 3*8*4*1 MSEREG(TRANSIG) 1.Mass Loss 0.9998 0.9998 0.9998
NNA6 3*8*4*1 SSE(TRANSIG) 0.6613 0.6740 0.6377
NNA7 MLP2 3*8*4*1*1 MSE(TRANSIG) 0.9999 0.9999 0.9999
NNA8 3*8*4*1*1 MSEREG(TRANSIG) 0.9999 0.9999 0.9999
NNA9 3*8*4*1*1 SSE(TRANSIG) 0.4590 0.4483 0.4508
NNA10 MLP2 3*8*4*1*1 MSE(LOGSIG) 0.9550 0.9543 0.9546
NNA11 3*8*4*1*1 MSE(PURELIN) 0.8671 0.8705 0.8673

3.4. Comparative indices of combustion performance heating are more appealing. The varying performance indices were
due to the heterogeneity present in lignite composition.
Factors like CCI, IG, IB, Tb, Ti, Tp, Rm, Ra were taken under
consideration to study the performance of combustion (Table 5). AL 3.5. Validation of experimental results using NNA (neural network-


had the highest weight loss rate @ 20 C/min (1.214 mg/min) and Artificial)
decreased after an increase in heating rate due to improper com-
bustion. The burnout temperature, ignition & peak temperatures, 11 different models were tested. The first nine models included
respectively, all showed sine wave behaviors. Highest values were testing of SLP, MLP models, with each being subjected to MSE,



reported as follows @ 10 C/min Tb (698.12
C), @ 50 C/min Ti MSEREG, SSE transfer functions. Out of all the models, the ones that



(444.2
C), @ 50 C/min Tp (532.69
C). Up to 20 C/min, these yielded the most accurate and precise results were subjected to


temperatures decrease, but as the rate of heating is 50 C/min, it LOGSIG and PURELIN functions. 99.99% (Training), 99.99%(Valida-
again increases. This is due to the complex catalytic activities inside tion), 99.99%(Testing) R2, 0.015 (MSE) proved NNA7 as best model.
the AL subject to high heating rates. The present study suggested The echoing of actual and predicted data (Fig. 9c) cemented the
that the optimum rate of heating with respect to all parameters is NNM application accuracy in predicting percentage loss in mass.
20
C/min, which is preferred industrially also. The highest value of The architecture of NNM and the regression analysis plots were


CHCI (1.38E-09 mg min2 OC3) was obtained at 50 C/min. Both IG shown in Fig. 9a and b. NNM authenticated the data that was ob-
and IB showed an increase with the incre ase in heating rates. This tained using TGA. Overfitting of the networks was avoided by N-
was supported by enhanced autocatalytic reactions in AL. The ratio fold cross-validation and a high correlation coefficient was due to
of fixed carbon to volatile matter ratio is the Fuel ratio. Combustion collinearity present in the data. It confirmed the authenticity of the
efficiency and fuel ratio (FR) are inversely related. FR of 1.47 for AL experiments carried out. It projected the difference between the
made it evident that it is a stable combustion fuel. actual versus predicted system under consideration. Supplemen-
FR also allows inferring stability of combustion flame intensity. tary figures G1-G4 represent different performance and transfer
As FR variability is not large, the combustion system has got good functions used in validating NNA models. The data set used in NNA
stability of flame. Char formation rate of material is dependent on model testing and validation are separately provided in the sup-
its FC content. AL had a reasonable value of FC for a valid amount of plementary data MS-EXCEL file.
char to be formed, but the same char is useful in catalyst prepa-
ration. The comprehensive parameter of combustion is CHCI,
3.6. Characterization of Australian lignite coal
reflecting the ignition ease along with the velocity of burning and
temperature of burnout. More excellent the CSI value, the greater
The characterization of Australian lignite coal was carried out
the pace of burning of sample and the faster char burning. A
using highly pre
cised analytical instruments like X-Ray diffraction,
comparative literature survey also showed the same trends (Lin,
Fourier transform infrared spectroscopy, scanning electron micro-
2015; Wang, 2018; Parvez, 2017). Therefore, the study of indices
scope, particle size analysis. It involved phase study, identification
revealed that for effective combustion of lignite coal, lower rates of
of functional groups, heavy metals, the composition of elements. It

12
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 9. Architecture of NNA (a), Regression plots (b), MSE value (c) Actual Vs Predicted Values (d).

also involved morphological study and determination of the size of was responsible for the formation of aromatic structures in the coal.
particles for the process under consideration. These properties will In the spectra 2000-1200 cm1 wave-number limit, C]C alkene
be beneficial for Australian lignite coal systems as a dominant fuel bonding was noted and at 1643 cm1 sharp peak was evident. At
in their combustion or pyrolytic processes. Fig. 9 represents the 1464 cm1 eCH2 group aliphatic in its nature was reported with a
graphs and images for characterization techniques used, sharp peak. 1200-400 cm1 wavenumber limit indicated SieO and
respectively. CeO bonds peak, thereby confirming the presence of silicate min-
erals. 900-750 cm1 wave-number limit indicated small peaks
confirming the presence of organosilicon compounds in the coal
3.6.1. XRD, FTIR, SEM, PSA analysis sample. For the ash of coal, the characteristic peaks of different
The XRD analysis (Fig. 10a) indicated rich minerals in the groups were observed in the zone of 1200e400 cm1 wave-
Australian lignite coal sample. The principal components were number. The peaks mainly reported the presence of siloxanes,
aluminosilicates, dimorphs pyrite, and marcasite, followed by oxysilicon, and disilicoxanes groups in the coal sample. The sharp
gypsum, barite, hydrated sulfates, calcite, and siderite [61]. The peaks were reported at wavenumber of 1056 cm1, 796 cm1,
phases evident before combustion were anatase, rutile, mullite, 432 cm1 respectively [11,14,62,63].
magnetite, quartz, hematite. After combustion, principally, quartz The SEM (Fig. 10c) and PSA (Fig. 10d) were used for morpho-
phase formation took place. The intensity of peaks was in the range logical and size of particles study, respectively. The coal sample
of 0e13200 counts and between 20 and 70
two-theta angle [50]. (SEM images) showed that Australian lignite coal had a large
The presence of various functional groups in the long-range amount of minerals present (catalytic, non-catalytic), and coal was
wavenumber zones was evident in the FTIR spectrogram microporous. ICP-OES, XRD, and FTIR confirmed the minerals pre-
(Fig. 10b) of Australian lignite coal. eOH and eCH2 aliphatic groups sent. The distribution range of 0e600 and 0e120 mm for the particle
were reported in the 3800-2700 cm1 wavenumber limit. At 3696 size (coal, ash) was observed. The bulk of particles for the coal were
and 3620 cm1, peaks of eOH groups (sharp and minor) were in the size range of 150e350 mm. Post-combustion, the bulk of ash
spotted, indicating minerals of clay being present. At 3360 cm1, particles was in the range of 10e50 mm. Therefore, the study sug-
peaks of phenols, alcohols, and carboxylic acids were evident. This gested that the post-combustion product was beneficial for other
was because of the fact of hydrogen bonding with eOH groups. At process industries [16,64e66].
2948 cm1 peak of the eCH2 group was noted. This aliphatic group

13
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Fig. 10. XRD graph (a), FTIR graph (b), SEM AL (c1,c2) AL ash (c3,c4), Particle size distribution (d).

4. Conclusion (Pyrolysis). The average activation energy (Ea) was 200 kJ/mol
(combustion) and 450 kJ/mol(pyrolysis). Iso-conversional simula-
The characterization results confirmed that the AL was highly tion methods (Vyzovkin and OFW) predicted the same values with
composed of functional groups (alcohols, carboxylic acids, phenols- more accuracy and precision (135e165 kJ/mol combustion,
OH, eCH2, -Si-O, eCH3) along with minerals like aluminosilicates, 210e240 kJ/mol pyrolysis). The master plot method validated three
dimorphs pyrite, marcasite, gypsum, barite, hydrated sulfates, reaction mechanisms for combustion, namely contracting volume
calcite, and siderite. Phases before combustion were anatase, rutile, (R3), two-dimensional diffusional (D2), and power-law P2/3,
mullite, magnetite, quartz, hematite. After combustion, quartz whereas pyrolysis followed contracting volume (R3). The reaction
phase formation took place due to loss of the anatase phase. Ash mechanism shifted from (R3, D2, P2/3) to (R3) as the process shifted
generated had mineral components with a strong affinity towards from combustion to pyrolysis. It confirmed combustion as a multi-
cement formation and favored co-processing. NNA corroborated step reaction process and pyrolysis as a single-step process. TGA
the authenticity of the TGA data with MLP networks NNA 7,8 as the analysis made it evident that combustion and pyrolysis processes
best fit models along with performance function(TRANSIG), trans- prefer lower heating rates. The research study is highly beneficial in
fer functions (MSE, MSEREG) R2 (0.9999). The combustion indices the crucial design, construction, maintenance, and operation of all
indicated that higher rates of heating would be a disaster for the systems employing lignite as the primary fuel.
combustion and proposed 20
C/min as the optimum rate of
heating with the CHCI value of 3.05E-9. Sinusoidal behavior of
ignition, burnout, and peak temperatures (Ti, Tb, Tp) proved highly Author contribution
complex autocatalytic reactions in AL. The pre-exponential factors
calculated by KAS, FOW, Freidman, Doyle models were in the range Sathiya Prabhakaran: Conceptualization; Data curation; Formal
of 1010 to 1014 min1 (combustion) and 1011 to 1051 min1 analysis; Funding acquisition; Investigation; Methodology; Project
administration; Resources; Software; Validation; Visualization;
14
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

Roles/Writing - original draft; Writing - review & editing. G. Swa- [17] Sp SP, S G, Joshi VV. Thermogravimetric analysis of hazardous waste: pet-
coke, by kinetic models and Artificial neural network modeling. Fuel 2020:
minathan: Mentoring; Investigation; Writing - review & editing;
119470.
Supervision; Formal analysis. Viraj V Joshi: Conceptualization; Data [18] Sp SP, Swaminathan G, Joshi VV. Energy conservation e a novel approach of
curation; Formal analysis; Funding acquisition; Investigation; co-combustion of paint sludge and Australian lignite by principal component
Methodology; Project administration; Resources; Software; Vali- analysis, response surface methodology and artificial neural network
modeling. Environmental Technology & Innovation; 2020, 101061.
dation; Visualization; Roles/Writing - original draft; Writing - re- [19] Vyazovkin S, Burnham AK, Criado JM, Pe
rez-Maqueda LA, Popescu C,
view & editing. Sbirrazzuoli N. ICTAC Kinetics Committee recommendations for performing
kinetic computations on thermal analysis data. Thermochim Acta
2011;520(1):1e19.
Funding [20] Vyazovkin S, Chrissafis K, Di Lorenzo ML, Koga N, Pijolat M, Roduit B, et al.
ICTAC Kinetics Committee recommendations for collecting experimental
This research did not receive any specific grant from funding thermal analysis data for kinetic computations. Thermochim Acta 2014;590:
1e23.
agencies in the public, commercial, or not-for-profit sectors. [21] Sriram A, Swaminathan G. Pyrolysis of Musa balbisiana flower petal using
thermogravimetric studies. Bioresour Technol 2018;265:236e46.
Declaration of competing interest [22] Lim ACR, Chin BLF, Jawad ZA, Hii KL. Kinetic analysis of rice husk pyrolysis
using kissinger-akahira-sunose (KAS) method. Procedia Eng 2016;148:
1247e51.
The authors declare that they have no known competing [23] Shukla SK, Srivastava D, Srivastava K. Synthesis, spectral and thermal degra-
financial interests or personal relationships that could have dation kinetics of the epoxidized resole resin derived from cardanol. Adv
Polym Technol 2015;34(1).
appeared to influence the work reported in this paper. [24] Venkatesh M, Ravi P, Tewari SP. Isoconversional kinetic analysis of decom-
position of nitroimidazoles: friedman method vs flynnewalleozawa method.
Acknowledgements J Phys Chem 2013;117(40):10162e9.
[25] Li B, Liu G, Gao W, Cong H-Y, Bi M-S, Ma L, et al. Study of combustion
behaviour and kinetics modelling of Chinese Gongwusu coal gangue: model-
The authors gratefully acknowledge the National Institute of fitting and model-free approaches. Fuel 2020;268:117284.
Technology, Tiruchirappalli, Tamil Nadu, India, for providing facil- [26] Bhange DS, Pandya NA, Jha RK, Ramaswamy V. Non-isothermal kinetic studies
ities for this research. Also, the Ministry of Human Resource and of the template decomposition from silicalite-1 framework-high temperature
X-ray diffraction and thermogravimetric analysis. Microporous Mesoporous
Development, Government of India, is acknowledged for the pro- Mater 2008;113(1):64e71.
vision of institute fellowship. The authors would like to acknowl- [27] Liu NA, Fan WC. Critical consideration on the Freeman and Carroll method for
edge South Ural State University, Russia, for lending simulated evaluating global mass loss kinetics of polymer thermal degradation. Ther-
mochim Acta 1999;338(1):85e94.
kinetic software. [28] Clark SM. A kinetic study of the I to II phase transition in ammonium chloride,
ammonium bromide and ammonium iodide. J Non-Cryst Solids 2016;451:
Appendix A. Supplementary data 161e7.
[29] Gupta MC, Viswanath SG. Kinetic compensation effect in the thermal degra-
dation of polymers. J Therm Anal 1996;47(4):1081e91.
Supplementary data to this article can be found online at [30] Gurnule WB, Das N. Kinetic study of non-isothermal decomposition of
https://doi.org/10.1016/j.energy.2021.122949. copolymer resin derived from 2, 4-dihydroxypropiophenone, 1, 5-
diaminonaphthalene and formaldehyde. Mater Today Proc 2019;15:611e9.
[31] Cai H, Liu J, Kuo J, Buyukada M, Evrendilek F. Thermal characteristics, kinetics,
References gas emissions and thermodynamic simulations of (co-)combustions of textile
dyeing sludge and waste tea. J Clean Prod 2019;239:118113.
[1] Bejarano PA, Levendis YA. Single-coal-particle combustion in O2/N2 and O2/ [32] Duan Y, Li J, Yang X, Hu L, Wang Z, Liu Y, et al. Kinetic analysis on the non-
CO2 environments. Combust Flame 2008;153(1):270e87. isothermal dehydration by integral master-plots method and TGeFTIR study
[2] Sisodia GS, Singh P. The status of renewable energy research on India. Energy of zinc acetate dihydrate. J Anal Appl Pyrol 2008;83(1):1e6.
Proc 2016;95:416e23. [33] da Silva JCG, de Albuquerque JG, Galdino WVdA, de Sena RF, Andersen SLF.
[3] Oertel C, Matschullat J, Zurba K, Zimmermann F, Erasmi S. Greenhouse gas Single-step and multi-step thermokinetic study e deconvolution method as a
emissions from soilsda review. Geochemistry 2016;76(3):327e52. simple pathway for describe properly the biomass pyrolysis for energy con-
[4] Zhou Z, Hu X, You Z, Wang Z, Zhou J, Cen K. Oxy-fuel combustion charac- version. Energy Convers Manag 2020;209:112653.
teristics and kinetic parameters of lignite coal from thermo-gravimetric data. [34] Xiao R, Yang W, Cong X, Dong K, Xu J, Wang D, et al. Thermogravimetric
Thermochim Acta 2013;553:54e9. analysis and reaction kinetics of lignocellulosic biomass pyrolysis. Energy
[5] Yang ZR, Huang JJ, Song SS, Wang ZQ, Fang YT. Insight into the effects of 2020;201:117537.
additive water on caking and coking behaviors of coal blends with low-rank [35] Doyle CD. Kinetic analysis of thermogravimetric data. J Appl Polym Sci
coal. Fuel 2019;238:10e7. 1961;5(15):285e92.
[6] Shi Z. Effect of hydrothermal treatment on structure and liquefaction behavior [36] Hu Q, Tang Z, Yao D, Yang H, Shao J, Chen H. Thermal behavior, kinetics and
of Baiyinhua coal. Fuel Process Technol 2017;167. 7-654-2017 v.167. gas evolution characteristics for the co-pyrolysis of real-world plastic and tyre
[7] Gai Z, Zhang R, Bi J. Characteristics of soot from rapid pyrolysis of coal and wastes. J Clean Prod 2020;260:121102.
petroleum coke. Energy Fuels 2017;31(4):3759e67. [37] Drozin D, Sozykin S, Ivanova N, Olenchikova T, Krupnova T, Krupina N, et al.
[8] Liu Z, Guo X, Shi L, He W, Wu J, Liu Q, et al. Reaction of volatiles e a crucial step Kinetic calculation: software tool for determining the kinetic parameters of
in pyrolysis of coals. Fuel 2015;154:361e9. the thermal decomposition process using the Vyazovkin Method. Software
[9] Chen T-Y, Yeh T-L, Lee Y-T. Comparison of power plants efficiency among 73 2020;11:100359.
countries. J Energy 2013;2013:916413. [38] Corte
s AM, Bridgwater AV. Kinetic study of the pyrolysis of miscanthus and its
[10] Ashman PJ, Mullinger PJ. Research issues in combustion and gasification of acid hydrolysis residue by thermogravimetric analysis. Fuel Process Technol
lignite. Fuel 2005;84(10):1195e205. 2015;138:184e93.
[11] UN environmental programme report on gas emissions. 2019. [39] Damartzis T, Vamvuka D, Sfakiotakis S, Zabaniotou A. Thermal degradation
[12] Abdallah L, El-Shennawy T. Reducing carbon dioxide emissions from elec- studies and kinetic modeling of cardoon (Cynara cardunculus) pyrolysis using
tricity sector using smart electric grid applications. J Eng 2013;2013:845051. thermogravimetric analysis (TGA). Bioresour Technol 2011;102(10):6230e8.
[13] Mahlia TMI, Lim JY, Aditya L, Riayatsyah TMI, Pg Abas AE, Nasruddin. Meth- [40] Huang J, Liu J, Kuo J, Xie W, Zhang X, Chang K, et al. Kinetics, thermodynamics,
odology for implementing power plant efficiency standards for power gen- gas evolution and empirical optimization of (co-)combustion performances of
eration: potential emission reduction. Clean Technol Environ Policy spent mushroom substrate and textile dyeing sludge. Bioresour Technol
2018;20(2):309e27. 2019;280:313e24.
[14] Jankovi
c B, Mani
c N, Radovi

c I, Jankovi

c M, Rajaci

c M. Model-free and model- [41] Parisi GI, Kemker R, Part JL, Kanan C, Wermter S. Continual lifelong learning
based kinetics of the combustion process of low rank coals with high ash with neural networks: a review. Neural Network 2019;113:54e71.
contents using TGA-DTG-DTA-MS and FTIR techniques. Thermochim Acta [42] Schmidhuber J. Deep learning in neural networks: an overview. Neural
2019;679:178337. Network 2015;61:85e117.
[15] Heydari M, Rahman M, Gupta R. Kinetic study and thermal decomposition [43] Abiodun OI, Jantan A, Omolara AE, Dada KV, Mohamed NA, Arshad H. State-of-
behavior of lignite coal. Int J Chem Eng 2015;2015:481739. the-art in artificial neural network applications: a survey. Heliyon 2018;4(11):
[16] Alonso MJG, Alvarez D, Borrego AG, Mene
ndez R, Marb
an G. Systematic effects e00938.
of coal rank and type on the kinetics of coal pyrolysis. Energy Fuels [44] Beigi M, Torki-Harchegani M, Tohidi M. Experimental and ANN modeling
2001;15(2):413e28. investigations of energy traits for rough rice drying. Energy 2017;141:

15
S.S. Prabhakaran, G. Swaminathan and V.V. Joshi Energy 242 (2022) 122949

2196e205. by X-ray diffraction, Raman and FTIR spectroscopy. Energy 2010;35(12):

[45] Xie C, Liu J, Zhang X, Xie W, Sun J, Chang K, et al. Co-combustion thermal 5347e53.
conversion characteristics of textile dyeing sludge and pomelo peel using TGA [63] ~ oz E, Moliner R. FTIR study of the evolution of coal structure
Ibarra J, Mun
and artificial neural networks. Appl Energy 2018;212:786e95. during the coalification process. Org Geochem 1996;24(6):725e35.
[46] Bryant GW, Browning GJ, Emanuel H, Gupta SK, Gupta RP, Lucas JA, et al. The [64] Vuthaluru HB, Zhang D-k. Effect of coal blending on particle agglomeration
fusibility of blended coal ash. Energy Fuels 2000;14(2):316e25. and defluidisation during spouted-bed combustion of low-rank coals. Fuel
[47] Engin B, Atakül H. Air and oxy-fuel combustion kinetics of low rank lignites. Process Technol 2001;70(1):41e51.
J Energy Inst 2018;91(2):311e22. [65] Ramesh R, Somasundaran P. SEM/EDX studies on association of mineral
[48] Santhosh Raaj S, Arumugam S, Muthukrishnan M, Krishnamoorthy S, matter in coal. Coal Prep 1990;8(1e2):93e9.
Anantharaman N. Characterization of coal blends for effective utilization in [66] Sung D-J, Kang S-H. Measurement of coal particle size and shape with
thermal power plants. Appl Therm Eng 2016;102:9e16. concentrated coal-water mixtures. Kor J Chem Eng 1997;14(1):1e7.
[49] Ye DP, Agnew JB, Zhang DK. Gasification of a South Australian low-rank coal
with carbon dioxide and steam: kinetics and reactivity studies. Fuel
1998;77(11):1209e19.
NOMENCLATURE
[50] Qiu T, Yang J-G, Bai X-J, Wang Y-L. The preparation of synthetic graphite
materials with hierarchical pores from lignite by one-step impregnation and TGA: Thermogravimetric analysis
their characterization as dye absorbents. RSC Adv 2019;9(22):12737e46. ICP-OES: Inductively Coupled Plasma Optical Emission Spectroscopy
[51] Yuzbasi NS, Selçuk N. Air and oxy-fuel combustion characteristics of biomass/ FTIR: Fourier Transform Infrared Spectroscopy
lignite blends in TGA-FTIR. Fuel Process Technol 2011;92(5):1101e8. XRD: X-Ray Diffraction
[52] Khatami R, Levendis YA. An overview of coal rank influence on ignition and SEM: Scanning Electron Microscope
combustion phenomena at the particle level. Combust Flame 2016;164: A: Pre exponential factor (min1)
22e34. B, C, D: Constant
[53] Kang T-J, Jang N-H, Kim H-T. The study on characteristics upgraded low rank a: Conversion
coal (Lignite-IBC) by changed temperature and particle size In: Yao T editor. Ea: Activation Energy (kJ/mol)
Conference the study on characteristics upgraded low rank coal (Lignite-IBC) K: Reaction Rate Constant
by changed temperature and particle size, Tokyo. Springer Japan, p. 222-228. R: Universal Gas Constant (kJ/mol*K)
[54] Quast KB, Readett DJ. Minerals and inorganic components associated with T: Absolute Temperature (K)
South Australian lignites. In: Coal science II: American chemical society; 1991. t: Time (min)
p. 20e30. w: Weight of sample at time t (mg)
[55] Zhang Z, Zhu M, Li J, Zhang K, Xu G, Zhang D. Experimental study of ignition wf: Weight of residue at time the end of reaction (mg)
and combustion characteristics of single particles of zhundong lignite. Energy w0: Initial weight of sample (mg)
Fuels 2018;32(4):4221e6. X: Weight of sample undergoing reaction (mg)
[56] PawlakeKruczek H, Czerep M, Niedzwiecki L, Karampinis E, Violidakis I, Vs: Versus
Avagianos I, et al. Drying of lignite of various origins in a pilot scale toroidal M: Moisture content (%)
fluidized bed dryer using low quality heat. Energies 2019;12(7). VM: Volatile Matter (%)
[57] Zakrzewski M, Sciazko A, Komatsu Y, Akiyama T, Hashimoto A, Kaneko S, et al. F: Fixed Carbon (%)
Numerical analysis of single and multiple particles of Belchatow lignite dried GCV: Gross Calorific Value (kJ/kg)
in superheated steam. Heat Mass Tran 2018;54(8):2215e30. Al: Australian Lignite
[58] Zhang H, Dou B, Zhang H, Li J, Ruan C, Wu C. Study on non-isothermal kinetics FC, SW, CR, KC: Freeman-Carroll, Sharp-Wentworth, Coats-Redfern, Kennedy-Clark
and the influence of calcium oxide on hydrogen production during bitumi- CHCI, IG, IB, FR, Co, CV: Characteristic combustibility Index, Ignition Index, Burnout
nous coal pyrolysis. J Anal Appl Pyrol 2020;150:104888. Index, Fuel Ratio, Conversion, Calorific Value
[59] Chapter 3 Kinetic models for solid state reactions. In: Galwey AK, Brown ME, Tb, Ti, Tp, Rm, Ra: Burnout temperature, Ignition temperature, Peak temperature,
editors. Studies in physical and theoretical chemistry. Elsevier; 1999. Maximum weight loss, Average weight loss
p. 75e115. C, H, N,S: Carbon, Hydrogen, Nitrogen, Sulphur
[60] Khawam A, Flanagan DR. Solid-state kinetic models:basics and mathematical NNA: Artificial Neural Networks
fundamentals. J Phys Chem B 2006;110(35):17315e28. ASTM, USEPA, UATR, APHA: American Society for Testing & Materials, U.S. Environ-
[61] Takagi H, Maruyama K, Yoshizawa N, Yamada Y, Sato Y. XRD analysis of mental Protection Agency, Universal Attenuated Total Reflectance, American
carbon stacking structure in coal during heat treatment. Fuel 2004;83(17): Public Health Association
2427e33.
[62] Sonibare OO, Haeger T, Foley SF. Structural characterization of Nigerian coals

16