Assessment of the thermal decomposition kinetics of empty fruit bunch, kernel shell and their blend

Journal Pre-Proof Assessment of the thermal decomposition kinetics of empty fruit bunch, kernel shell and their blend Yesid Javier Rueda-Ordó ñez, Carlos Junior Arias-Hernández, Julián Fernando Manrique-Pinto, Paola Gauthier-Maradei, Waldir Antônio Bizzo PII: DOI: Reference:

S0960-8524(19)31153-8 https://doi.org/10.1016/j.biortech.2019.121923 BITE 121923

To appear in:

Bioresource Technology

Received Date: Revised Date: Accepted Date:

5 June 2019 26 July 2019 27 July 2019

Please cite this article as: Rueda-Ordó ñez, Y.J., Arias-Hernández, C.J., Manrique-Pinto, J.F., Gauthier-Maradei, P., Bizzo, W.A., Assessment of the thermal decomposition kinetics of empty fruit bunch, kernel shell and their blend, Bioresource Technology (2019), doi: https://doi.org/10.1016/j.biortech.2019.121923

This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. Please note that, during the production process, errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

© 2019 Published by Elsevier Ltd.

JOURNAL PRE-PROOF

1

Assessment of the thermal decomposition kinetics of empty fruit bunch, kernel

2

shell and their blend

3 Yesid Javier Rueda-Ordóñeza*; e-mail: [email protected]

5

Carlos Junior Arias-Hernándeza; e-mail: [email protected]

6

Julián Fernando Manrique-Pintoa; e-mail: [email protected]

7

Paola Gauthier-Maradeic; e-mail: [email protected]

8

Waldir Antônio Bizzob; e-mail: [email protected]

PR

O

O

F

4

9 a

GIEMA Research Group, School of Mechanical Engineering, Universidad Industrial de

E-

10

Santander – Carrera 27 Calle 9, Bucaramanga, Santander, Colombia

12

b

PR

11

School of Mechanical Engineering, University of Campinas – Rua Mendeleyev, 200,

13

17 18

Industrial de Santander –Carrera 27 Calle 9, Bucaramanga, Santander, Colombia

*Corresponding author. Tel.: +57 7 6344000 Ext: 2809; e-mail: [email protected]

Abstract

JO

19

AL

16

INTERFASE Research Group, School of Chemical Engineering, Universidad

R N

15

c

U

14

Campinas, SP, 13083-860, Brazil

20

In this work, were studied the thermal and kinetic characteristics of the palm kernel

21

shell (PKS) and empty fruit bunch (EFB) from the African oil palm. Experiments in the

22

inert atmosphere were carried out in a thermogravimetric analyzer. In the kinetic

23

analysis were applied the one-step reaction through iso-conversion methods, mechanism

24

of independent parallel reactions (MIPR), and mechanism of consecutive reactions

1

JOURNAL PRE-PROOF

(MCR). The one-step reaction mechanism overestimated the thermal decomposition of

2

all samples; however, the best was the EFB. The MIPR showed to be representative of

3

the thermal decomposition of all samples, and the proposed correlations between the

4

pre-exponential factor and the heating rate improved the accuracy of the model. The

5

MCR analysis showed that using the same kinetic parameters applied in the MIPR does

6

not affect reliability. Finally, as a conclusion, blending PKS with EFB increase 5%

7

heating value and decrease 50% ash content.

8

Keywords: Biomass; Model-free; Thermogravimetry; Pyrolysis; Combustion.

9

1. Introduction

E-

PR

O

O

F

1

Biomass is currently considered one of the most promising alternatives to fossil fuels,

11

and presents interesting characteristics as a raw material for producing energy via

12

several processes, such as combustion, gasification, liquefaction, and pyrolysis (Rueda-

13

Ordóñez and Tannous, 2015). According to the United States Department of

14

Agriculture, the production of oil palm in Colombia for 2019 is estimated to be 1680 kt,

15

making Columbia the fourth-greatest palm oil producer in the world (IndexMundi,

16

2019). This high level of oil production leads to the production of large amount of

AL

R N

residual biomass, which is mainly composed of empty fruit bunch (EFB), fiber, and kernel shells (PKS). In 2010, the energy potential associated with residual biomass from

JO

18

U

17

PR

10

19

the palm oil industry in Colombia was estimated to be 16000 TJ/year, corresponding to

20

a production of approximately 873 t/year (Escalante-Hernández et al., 2011). Colombia

21

is currently experiencing a post-conflict stage, in which people displaced from rural

22

areas will return to their lands (Reinoso-Pérez et al., 2019; Gutierrez-Sanin and Marín-

23

Jaramillo, 2018). Therefore, there is potential for growth in the farming of crops such as

2

JOURNAL PRE-PROOF

African oil palm. African oil palm crops produce massive quantities of residual

2

biomass. The ratios (t residue)/(t main product) of waste PKS, fibers, and EFB are 0.22,

3

0.63, and 1.06; however, only PKS is commonly used as a solid fuel in boilers.

4

Therefore, investigation of the thermal behavior of these residues is a priority for

5

Colombia, as it could contribute to the development of new technologies for regional

6

agroindustry.

7

Many investigations (Kongkaew et al., 2015; Rueda-Ordoñez & Tannous, 2015; Lee et

8

al., 2017; Da Silva et al., 2018; Mishra and Mohanty, 2018) of the thermal and kinetic

9

characteristics of various types of biomass (rice straw, sugarcane straw, palm kernel

E-

PR

O

O

F

1

shell, empty fruit bunch, palm oil sludge, and waste woods such as eucalyptus and pine)

11

have been conducted in order to understand the phenomena that occur during their

12

thermal degradation, and thus, to lay the foundation for the scale-up of processes

13

utilizing these biomass sources at the industrial level. Investigations of the thermal

14

decomposition of biomass normally include a study of the kinetics of the reactions

15

involved. Thermogravimetry is typically applied for both solid fuels and biomass, as

16

this process is governed by solid-state reactions. The kinetics of the thermal

17

decomposition of biomass are commonly studied by applying three different solid-state

AL

R N

U

reaction mechanisms, namely, mechanisms involving one global reaction, consecutive

JO

18

PR

10

19

reactions, and parallel reactions (Rueda-Ordóñez and Tannous, 2018). Some previous

20

research works (Zakir et al., 2011; Nyakuma et al., 2015; Dewayanato et al., 2016; Lee

21

et al., 2017; Chan et al., 2018) have reported the thermal and kinetic characteristics of

22

residues (EFB and PKS) related to the exploitation of African oil palm. However, no

23

complete and reliable kinetic analysis of EFB, PKS, or their blends was encountered in

24

the literature.

3

JOURNAL PRE-PROOF

As described in Rueda-Ordóñez et al. (2015), the mechanism of independent parallel

2

reactions (MIPR) as applied to the thermal decomposition of biomass, assumes that

3

each biomass component reacts via a single reaction, and that the sum of these reactions

4

determines the overall thermal decomposition behavior. Normally, the biomass

5

components evaluated are hemicellulose, cellulose, and lignin; the activation energies

6

associated with their reactions vary from 90 to 160, from 140 to 220, and from 40 to

7

150 kJ/mol, respectively (Rueda-Ordóñez et al., 2015; Rueda-Ordóñez and Tannous,

8

2017). One previous work (Rueda-Ordóñez and Tannous, 2017) analyzed more than 20

9

investigations using MIPR, and reported that the reaction model most commonly

PR

O

O

F

1

assumed for the reactions in this mechanism was an nth order reaction, with n=1 for

11

hemicellulose and cellulose, and n =1 to 3 for lignin. Yu et al. (2015) implemented the

12

MIPR to investigate the kinetics of the pyrolysis of olive kernel, vine shoots, and

13

bamboo and obtained activation energies of 87.5, 108.5, and 167.3 kJ/mol for

14

hemicellulose, 132.1, 149.6, and 241.8 kJ/mol for cellulose, and 44.1, 38.1, and 74.5

15

kJ/mol for lignin, respectively. Bartocci et al., (2017) investigated the pyrolysis kinetics

16

of pellets composed of 90% fir sawdust and 10% glycerol. The authors assumed four

17

reactions in the MIPR: one for glycerol, and the other three for the lignocellulosic

18

biomass components. The activation energies obtained in their work were 149.7, 230.1,

PR

AL

R N

U

154.3, and 74.5 kJ/mol for hemicellulose, cellulose, lignin, and glycerol, respectively;

JO

19

E-

10

20

these values were within the ranges presented in Rueda-Ordóñez and Tannous (2017).

21

Reports of the application of the mechanism of consecutive reactions (MCR) to the

22

thermal decomposition kinetics of biomass are scarce (Luangkiattikhun et al., 2008;

23

Weerachanchai et al., 2010; Lopes et al., 2016; Rueda-Ordóñez and Tannous, 2018)

24

compared with those in which the MIPR is applied. Normally, for this reaction

4

JOURNAL PRE-PROOF

mechanism, the thermal decomposition kinetics of the biomass are described using two,

2

three, or four consecutive reactions, which are related to low-temperature volatile

3

release (between 150 and 300 °C), high-temperature volatile release (300 to 450 °C),

4

and carbonization (400 to 900 °C). Rueda-Ordóñez and Tannous (2018) observed that

5

the activation energies of these three reactions were on the same order of magnitude as

6

the those related to hemicellulose, cellulose, and lignin using the MIPR.

7

Therefore, in this work, the thermal and kinetic characteristics of palm kernel shell

8

(PKS) and empty fruit bunch (EFB) from the African oil palm were studied. The

9

importance and innovation of this investigation were the implementation of more

10

accurate techniques for the study of the thermal decomposition kinetics of these

11

materials, since the techniques encountered in the literature (Zakir et al., 2011;

12

Nyakuma et al., 2015; Dewayanato et al., 2016; Lee et al., 2017; Chan et al., 2018) for

13

these biomass sources only involve analysis through iso-conversion methods or one-

14

step reaction mechanism. Such iso-conversion methods should only be used for

15

preliminary kinetic analysis, since it tends to overestimate the values of the kinetic

16

parameters, and the error associated with this modeling is very high. This study aimed

17

to investigate the thermal decomposition process of empty fruit bunch (EFB), palm

O

O

PR

E-

PR

AL

R N

U

kernel shell (PKS), and a 50/50 mass% blend of the two under an argon atmosphere

JO

18

F

1

19

using thermogravimetric analysis. This study also included kinetic analysis of these

20

residues using three different reaction mechanisms, namely, one-step reaction, three

21

parallel reactions, and three consecutive reactions, to obtain the representative kinetic

22

parameters for each mechanism.

23

5

JOURNAL PRE-PROOF

2. Materials and method

2

2.1. Physical-chemical characterization of the biomass

3

The agricultural residues analyzed in this work were palm kernel shell (PKS) and empty

4

fruit bunch (EFB) from African oil palm (Elaeis guineensis). First, their physical-

5

chemical characteristics (moisture, proximate analysis, ultimate analysis, and heating

6

value) were determined. These characteristics are key parameters in determining the

7

quality of the biomass as a solid fuel, and thus, this step is of great importance to the

8

selection of the most suitable thermochemical process.

9

The PKS and EFB samples were analyzed as received from a farm located in Puerto

E-

PR

O

O

F

1

Wilches, Colombia, which is in the northeastern region of the country with the

11

coordinates 7° 20′ 54″ N, 73° 53′ 54″ W and an average altitude of 75 meters above sea

12

level. At the farm, the residues were stored in piles open to the air for three weeks,

13

which decreased their natural moisture somewhat. Then, 1 kg of each biomass was

14

separated, ground in a blade mill, and sieved to obtain a mean particle size of 250 µm.

15

For the moisture content determination, the standard ASTM E871-82 was applied using

16

three 10 g samples in a forced convection oven. The blend was then prepared using 50

AL

R N

U

mass% of each biomass on a dry mass basis. The proximate analysis was determined

JO

17

PR

10

18

according to the standards ASTM E1755-01 and ASTM E872-82, while the ultimate

19

analysis was estimated using the mathematical approach of Parikh et al. (2007), in

20

which the percentage of carbon (C), hydrogen (H), and oxygen (O) are determined with

21

an average absolute error of less than 5 mass%. Finally, the higher heating value (HHV)

22

was determined in an oxygen calorimetric bomb (IKA C200) by applying the standard

23

ASTM D240-09. The lower heating value (LHV) in MJ/kg was determined using Eq. 6

JOURNAL PRE-PROOF

1

(1), in which H and M correspond to the percentages of hydrogen and moisture in the

2

biomass, respectively (Monir et al., 2018). 9𝐻 𝑀 + ) 100 100

(1)

F

𝐿𝐻𝑉 = 𝐻𝐻𝑉 − 2.260 (

2.2. Experimental set-up

4

The thermal analysis was carried out using a thermogravimetric analyzer (NETZSCH

5

STA 449F3). The mass loss of the sample was measured using a step of 5 s (12

6

samples/min) from room temperature to 900 °C, with heating rates of 1.25, 2.5, 5, and

7

10 °C/min. In order to obtain reliable and accurate results, the experiments were carried

8

out in triplicate, and prior to each experiment, the samples were dried, and the

9

thermogravimetric analyzer was evacuated with a vacuum pump to exclude trace

PR

E-

PR

O

O

3

oxygen. The initial mass in each experiment was 7.0 ± 0.5 mg. A 50 mL/min (STP)

11

flow of argon provided an inert atmosphere.

12

2.3. Kinetic analysis and methods

13

The thermal decomposition analysis aimed to determine of the stages of dehydration,

14

volatile release, and carbonization, and their corresponding temperature and mass loss

R N

U

ranges. Thus, to allow quantitative comparison of the thermal decomposition curves

JO

15

AL

10

16

obtained at different heating rates, the thermogravimetric data and their derivatives

17

(DTG) were normalized using Eqs. (2) and (3), respectively, in which mi is the initial

18

mass, m is the mass at time t, and w is the normalized mass. In Eq. (3), dm/dt represents

19

the experimental DTG in mg/s, and dw⁄dt is the normalized DTG.

𝒘=

𝒎 𝒎𝒊

(2)

7

JOURNAL PRE-PROOF

𝒅𝒘 𝒅𝒎 𝟏 = 𝒅𝒕 𝒅𝒕 𝒎𝒊

(3)

The determination of the experimental conversion (αe), which represents the conversion

2

of the biomass into the product gases for the selected decomposition range, and the

3

experimental conversion rate (dα/dt)e were carried out using Eqs. (4) and (5). In these

4

equations, wi, w, and wf are the normalized mass at the beginning, at time t, and at the

5

end of the decomposition range.

PR

𝒘𝒊 − 𝒘 𝒘𝒊 − 𝒘𝒇

(4) (5)

PR

𝒅𝛂 𝒅𝒘 𝟏 ( ) =− ( ) 𝒅𝒕 𝒆 𝒅𝒕 𝒘𝒊 − 𝒘𝒇

E-

𝛂𝒆 =

O

O

F

1

The theoretical conversion rate (dα/dt)t presented in Eq. (6) is based on Arrhenius’s law,

7

in which Ea is the activation energy (kJ/mol), A is the pre-exponential factor (s-1), R is

8

the universal gas constant (8.314 kJ/mol K), T is the absolute temperature, and f(α) is

9

the reaction model.

R N

AL

6

𝒅𝛂 −𝑬𝒂 ( ) = 𝑨[𝒇(𝜶)] ⋅ [𝒆𝒙𝒑 ( )] 𝒅𝒕 𝒕 𝑹⋅𝑻

U

2.3.1. Single-step reaction

JO

10

(6)

11

The single-step reaction was evaluated by applying the iso-conversion methods

12

described in Friedman (1964) and Vyazovkin (2001) to obtain the apparent activation

13

energy; master plots were used to obtain the reaction model, and through linearization,

14

the pre-exponential factor. For the analysis, the four heating rates listed above and a

8

JOURNAL PRE-PROOF

range of conversions between 0 and 1 were used, and 121, 137, and 128 levels were

2

defined for EFB, PKS, and their blend, respectively.

3

2.3.1.1. Friedman method

4

The differential method developed in Friedman (1964) was derived by taking the natural

5

logarithm of Eq. (6) to provide the final form of the model presented in Eq. (7).

O

O

F

1

(7)

PR

𝒅𝜶 𝑬𝒂 𝐥𝐧 ( ) = 𝐥𝐧(𝑨) + 𝐥𝐧[𝒇(𝜶)] − ( ) 𝒅𝒕 𝑹⋅𝑻

Thus, in order to calculate the activation energy, for each conversion level, ln(dα/dt)

7

was plotted as a function of 1/T, and the slope of the resulting straight line was

8

multiplied by R to obtain the activation energy.

9

2.3.1.2. Vyazovkin method

PR

E-

6

The nonlinear integral method presented in Vyazovkin (2001) is represented by Eq. (8),

11

in which 𝐼(𝐸𝛼 , 𝑇𝑖 (𝑡𝛼 ))𝛼 = 𝑃(𝑥) ⋅ (𝐸𝛼 /𝑅) and the apparent activation energy 𝐸𝛼 for

12

each conversion level can be determined by minimizing the function ∅.

R N

AL

10

𝒏

𝒏

𝑰(𝑬𝜶 , 𝑻𝒊 (𝒕𝜶 )) ∗ 𝜷𝒋

U

∅ = ∑∑ 𝒋≠𝒊

(8)

𝑰 (𝑬𝜶 , 𝑻𝒋 (𝒕𝜶 )) ∗ 𝜷𝒊

JO

𝒊

13

The function p(x) was solved by applying the eighth-degree rational approximation

14

developed in Pérez-Maqueda and Criado (2000).

9

JOURNAL PRE-PROOF

2.3.2. Determination of the reaction model for single-step reaction

2

The kinetic reaction model was determined by comparing the experimental and

3

theoretical reaction model data through master plots, as presented in Eq. (9). The

4

mathematical forms of the reaction models (first and nth order, F1, Fn, one-dimensional

5

diffusion, D1, two-dimensional diffusion D2, three-dimensional diffusion D3, and

6

Ginstling-Brounshtein three-dimensional diffusion, D4) analyzed in this work have

7

been presented in several investigations on solid-state kinetics, and are commonly

8

known as the reaction model table (Vyazovkin et al., 2011; Rueda-Ordóñez and

9

Tannous, 2015).

(9)

PR

𝐴 ⋅ 𝐸𝑎 𝑝(𝑥) 𝑔(𝛼) 𝑝(𝑥) 𝛽⋅𝑅 = = 𝑔(0.5) 𝐴 ⋅ 𝐸𝑎 𝑝(𝑥 ) 𝑝(𝑥0.5 ) 0.5 𝛽⋅𝑅

E-

PR

O

O

F

1

2.3.3. Mechanism of independent parallel reactions

11

The parallel reaction scheme is represented by Eq. (10), where 𝐹i represents the volatile

12

mass fraction related to each pseudo-component, and j is the number of reactions. In

13

this work, three parallel reactions (Table 1) corresponding to the reactions of the three

R N

U

main pseudo-components, i.e., hemicellulose, cellulose, and lignin, were assumed, as

JO

14

AL

10

15

has been reported in previous works (Santos et al., 2011; Rueda-Ordóñez et al., 2015). 𝒋

𝒋

𝒅𝜶 𝒅𝜶𝒊 𝑬𝒊 = ∑ 𝑭𝒊 = ∑ 𝑭𝒊 𝑨𝒊 𝐞𝐱𝐩 (− ) 𝒇(𝜶𝒊 ) 𝒅𝒕 𝒅𝒕 𝑹𝑻 𝒊

𝒊

10

(10)

JOURNAL PRE-PROOF

2.3.4. Mechanism of consecutive reactions

2

The consecutive reaction scheme presented in Eq. (11) was applied assuming three

3

consecutive reactions (Table 1) and following the procedure recommended in Rueda-

4

Ordóñez and Tannous (2018). The kinetic constants were calculated according to the

5

Arrhenius equation.

O

F

1

PR

O

𝒅[𝑨] −𝑬𝟏 = −(𝑨𝟏 ) [𝐞𝐱𝐩 ( )] [𝑨] 𝒅𝒕 𝑹𝑻 𝒅[𝑩] −𝑬𝟏 −𝑬𝟐 = (𝑨𝟏 ) [𝐞𝐱𝐩 ( )] [𝑨] − (𝑨𝟐 ) [𝐞𝐱𝐩 ( )] [𝑩] 𝒅𝒕 𝑹𝑻 𝑹𝑻

(11)

PR

𝒅[𝑫] −𝑬𝟑 = (𝑨𝟑 ) [𝐞𝐱𝐩 ( )] [𝑪] 𝒅𝒕 𝑹𝑻

E-

𝒅[𝑪] −𝑬𝟐 −𝑬𝟑 = (𝑨𝟐 ) [𝐞𝐱𝐩 ( )] [𝑩] − (𝑨𝟑 ) [𝐞𝐱𝐩 ( )] [𝑪] 𝒅𝒕 𝑹𝑻 𝑹𝑻

According to Rueda-Ordóñez and Tannous (2018), in this reaction scheme, thermal

7

decomposition is divided into three stages: 1) soft pyrolysis or torrefaction followed by

8

2) pyrolysis and 3) final carbonization. In the first reaction, component A (dry biomass)

9

reacts to produce B (torrefacted biomass) with a pre-exponential factor A1 and activation

R N

precursor of the liquid fraction in pyrolysis, is released along with other gases to

JO

11

energy E1. In the second reaction, it is assumed that the volatile mass, which is a

U

10

AL

6

12

produce a carbonaceous solid (C); the associated kinetic parameters were A2 and E2. The

13

final reaction is related to the release of gas and vapors by the carbonaceous solid (C) to

14

produce the final residue in the thermal decomposition process.

15

Finally, the theoretical conversion of the different reaction mechanisms was calculated

16

by applying the fourth-order Runge-Kutta method. Determination of the kinetic

11

JOURNAL PRE-PROOF

parameters was carried out by comparing the experimental and theoretical conversion

2

rate data using the quality of fit, which was evaluated through the average deviation

3

(AD) presented in Eq. (12), which is based on the least squares method. The reliability

4

criterion was AD < 3%. In Eq. (12), (y)experimental and (y)theoretical are the experimental and

5

theoretical conversion rate data, respectively. N represents the number of experimental

6

points considered, with N = 1043, 1021, and 1149 for EBF, PKS, and their blend at 1.25

7

°C/min, respectively. The application of this methodology requires initial values for the

8

kinetic parameters; therefore, the parameters obtained from the iso-conversion analysis

9

were used, as recommended in the literature (Vyazovkin et al. 2011; Anca-Couce et al.,

O

O

PR

2014; Rueda-Ordóñez and Tannous, 2018).

E-

10

F

1

𝟐

PR

𝑵 √∑𝒊=𝟎[(𝒚)𝒆𝒙𝒑𝒆𝒓𝒊𝒎𝒆𝒏𝒕𝒂𝒍 − (𝒚)𝒕𝒉𝒆𝒐𝒓𝒆𝒕𝒊𝒄𝒂𝒍 ] 𝑵 𝑨𝑫(%) = 𝟏𝟎𝟎 (𝒚)𝒆𝒙𝒑𝒆𝒓𝒊𝒎𝒆𝒏𝒕𝒂𝒍,𝒎𝒂𝒙𝒊𝒎𝒖𝒎

]

AL

[

(12)

In order to obtain reliable kinetic parameters, the activation energy, reaction order, and

12

fraction (consecutive and parallel reactions) for each step of the reaction mechanisms

13

analyzed were considered independent of the heating rate. It is necessary to clarify this,

U

since in this kind of analysis, average kinetic parameters followed by a standard

JO

14

R N

11

15

deviation are normally presented in the literature (Santos et al., 2011; Rueda-Ordóñez et

16

al., 2015; Xavier et al., 2016), which could cause confusion to the readers. Then, to

17

obtain accurate values according to Eq. (12), only the pre-exponential factor was

18

considered to be a function of the heating rate, and was assumed to follow a first or

19

second order polynomial. This assumption was based on numerous works (Brown et al.,

20

1980; White et al., 2011; Vyazovkin et al., 2011) that have reported that the pre-

12

JOURNAL PRE-PROOF

exponential factor is only a fitting parameter in solid state kinetics, and does not have a

2

valid physical meaning.

3

3. Results and discussion

4

3.1. Biomass characteristics

5

Table 2 presents the physical-chemical and thermal characteristics of the empty fruit

6

bunch (EFB), palm kernel shell (PKS), and the 50/50 mass% blend of the two. The

7

proximate analyses (moisture (M), volatile material (VM), fixed carbon (FC), and ash)

8

showed that the EFB contained less VM than PKS, and therefore, the blend presented

9

an intermediate VM content. The VM content directly influences the thermal

E-

PR

O

O

F

1

decomposition process, with lower VMs enhancing the combustion and gasification

11

processes (Kuo et al., 2014; Ku et al., 2016; Trop et al., 2016), and higher values

12

enhancing pyrolysis (Vecino-Mantilla et al., 2014). Thus, based on their VM contents,

13

EFB could be more suitable for combustion or gasification, while the blend and PKS

14

could be more suitable for pyrolysis. The VM content of the EFB in this work was

15

lower than in previous reports in the literature (Escalante-Hernández et al., 2011;

16

Asadullah et al., 2013), but that of the PKS and blend were similar to those previously

AL

R N

U reported.

JO

17

PR

10

18

The ash content of the EFB was 5 mass% higher than that of PKS, which is typical for

19

these residues according to Vassilev et al. (2017) and Vecino-Mantilla et al. (2014);

20

however, the ash content of PKS was slightly lower than those commonly reported in

21

the literature (typically 4-10 mass%) (Escalante-Hernández et al., 2011; Vassilev et al.,

22

2017; Asadullah et al., 2013). As expected, the blend presented an intermediate ash

23

content, which was expected improve its thermal decomposition compared to EFB, 13

JOURNAL PRE-PROOF

since high ash contents in biomass can result in slagging, fouling, and corrosion, as

2

reported in Vassilev et al. (2013). However, according to Dupont et al (2011) and

3

Abdullah and Gerhauser (2008), the potassium in ash improves gasification

4

performance, since this element acts as a catalyst, thus increasing the gas production.

5

The ultimate analysis of EFB presented in Table 2 was compared with the analysis

6

presented in Abdullah and Gerhauser (2008) for the same biomass source. The carbon

7

content in this work was 5% lower, the hydrogen 1% lower, and the oxygen 3% higher.

8

For PKS, the carbon content was 5% higher, the hydrogen 1% higher, and the oxygen

9

3% lower than the results presented for the same biomass in Escalante et al. (2011).

E-

PR

O

O

F

1

These differences were attributed to the fact that in this work, these values were

11

determined by applying the correlations from Parikh et al. (2007).

12

The higher heating values determined for the residues were on the same order of

13

magnitude as those in Abdullah and Gerhauser (2008) and Vecino-Mantilla et al.

14

(2014), which were around 20 MJ/kg. According to Jenkins et al. (1998), the ash

15

content has a significant influence on the heating value, with a 1% increase in ash

16

content reducing the heating value by 0.2 MJ/kg. The heating value of Abdullah and

17

Gerhauser (2008) should have been slightly greater, since the biomass presented a lower

AL

R N

U

ash and a greater VM value, and thus did not follow the expected behavior. The HHV in

JO

18

PR

10

19

the work of Escalante-Hernández et al. (2011) could be estimated from their reported

20

LHV and moisture values using Eq. (1), resulting in an HHV of 19.98 MJ/kg. This

21

value was expected, since the sample presented a higher VM.

14

JOURNAL PRE-PROOF

3.2. Thermal decomposition

2

Fig. 1 present the TG and DTG experiments of the thermal decomposition of EFB,

3

blend, and PKS in an inert argon atmosphere at the studied heating rates. The thermal

4

decomposition process can be divided into three stages: moisture evaporation, volatile

5

release, and carbonization. For all the residues, the first stage or moisture evaporation

6

took place from room temperature to 150 °C. Between 150 °C and 200 °C, the mass

7

variation was less than 3% and was associated with the evaporation of some extractives.

8

In this work, the effect and reactions associated with the aforementioned process were

9

considered negligible. Volatile release occurred in the temperature range between 200

PR

O

O

F

1

°C and 400 °C, and was related to the main mass loss in the thermal decomposition

11

process.

12

Fig. 1(a) presents the curves of the thermal decomposition of EFB in an argon

13

atmosphere, in which average volatilized mass was 57.81 ± 2.53% with a maximum of

14

60.69% and a minimum of 55.31% at heating rates of 10 and 2.5 °C/min, respectively,

15

and corresponding to a 4.0% deviation from the average. Fig. 1(b) presents the DTG

16

curves for EFB, which were characterized by the presence of a single peak that was

17

attributed to the thermal decomposition of hemicellulose and cellulose at 277.7 °C,

PR

AL

R N

U

288.8 °C, 295.7 °C, and 306.8 °C for the different heating rates.

JO

18

E-

10

19

As shown in Fig. 1(c), the average volatilized mass for PKS was 60.05 ± 0.6 %, which

20

was associated with the two major peaks in the DTG curve in Fig. 1(d). The first peak

21

occurred at 252, 260, 270 and 276 °C, depending on the heating rate, and was related to

22

hemicellulose decomposition, while the second peak was located at 314, 325, 333, and

23

345 °C and was related to cellulose decomposition.

15

JOURNAL PRE-PROOF

Finally, the thermal decomposition behavior of the blend is presented in Fig. 1(e). The

2

average volatilized mass was 59.43 ± 1.32 %, which was intermediate between that of

3

the EFB and PKS, with 60% volatilized mass as a general trend. Fig. 1(f) presents the

4

DTG of the blend, in which a shoulder followed by a pronounced peak was observed.

5

This behavior is typical of a process with overlapped hemicellulose and cellulose

6

reactions. The major peak was related to the thermal decomposition of cellulose, and

7

occurred at temperatures of 286.3 °C, 297.8 °C, 308.2 °C, and 312.5 °C for the different

8

heating rates. The shoulders related to the thermal decomposition of hemicellulose were

9

located at 238.5 °C, 248 °C, 260.8 °C, and 265.3 °C.

E-

PR

O

O

F

1

The carbonization stage corresponds to the thermal decomposition process, which

11

occurred in the temperature range 400 °C–700 °C, and was only related to the thermal

12

decomposition of the remaining biomass (lignin-based structure). EFB presented the

13

lowest mass loss, 6.88%, at 1.25 °C/min and the blend presented the highest mass loss,

14

12.92%, at 5 °C/min. The residue remaining at 700 °C is mainly composed of bio-char

15

and ash. Excluding the ash, EFB presented 21.96%, 22.21%, 20.7%, and 20.49%

16

biochar at the heating rates evaluated. PKS presented 25.36%, 24.60%, 25.75%, and

17

26.34 % biochar, and the blend presented 23.19%, 21.2%, 20.13%, and 23.08% biochar.

AL

R N

U

Comparing these results of the FC determined through proximate analysis (Table 1)

JO

18

PR

10

19

with the previously reported values in Table 3, a perceptible difference of about 10

20

mass% dry mass was observed. This difference could be explained simply by the

21

different techniques and standards used in the proximate analysis, which was carried out

22

in a muffle oven using very different experimental conditions than those of the

23

previously reported data.

16

JOURNAL PRE-PROOF

3.3. Results of the kinetic analysis

2

3.3.1. Single-step reaction through iso-conversion methods

3

The thermal decomposition of biomass consists of several processes that can occur

4

simultaneously or separately depending on the experimental conditions; therefore, it is

5

governed by several complex reactions. However, for some applications, e.g., reactor

6

modeling and design, the use of a single-step reaction to describe the overall process is

7

common, although this represents a rough approximation. In this work the iso-

8

conversion methods described in Friedman (1964) and Vyazovkin (2001) were used to

9

obtain the kinetic parameters to describe the process using a single-step reaction for all

O

O

PR

E-

the samples analyzed.

PR

10

F

1

11

Fig. 2(a) presents the apparent activation energies as a function of the conversion

13

determined using the iso-conversion methods. The activation energy of empty fruit

14

bunch (EFB) varied with the conversion. Using the method of Friedman (1964), the Ea

15

varied from 130.32 to 272.38 kJ/mol, with an average of 180.41±33.80 kJ/mol, while

16

using that of Vyazovkin (2001), a range of 103.41 to 228.23 kJ/mol and an average of

R N

164.77 ± 32.55 kJ/mol was calculated. In Lee et al. (2017), the authors analyzed the pyrolysis kinetics of EFB using the Ozawa method and calculated an average activation

JO

18

U

17

AL

12

19

energy of 169.76 kJ/mol, which was similar to that obtained in this work using the

20

method of Vyazovkin (2001). However, the Friedman (1964) method gave an Ea 20

21

kJ/mol higher than that given in Lee et al. (2017). This difference has been observed in

22

other works which used both the integral and differential methodologies. Thus, it was

17

JOURNAL PRE-PROOF

assumed that the results determined using the method of Vyazovkin (2001) were more

2

reliable and accurate.

3

The percentage of variation of the data with respect to the standard deviation was found

4

to be 18.73% and 19.75%; these values were considered to be too high according to

5

recommendations of the ICTAC (Vyazovkin et al., 2011). Thus, the apparent activation

6

energy determined for the EFB thermal process assuming a single-step reaction

7

mechanism could be used for modeling purposes; however, caution should be exercised

8

when using these results, since their high variation provides no guarantee of accurate or

9

reliable modeled data.

E-

PR

O

O

F

1

For palm kernel shell (PKS), the apparent activation energy determined using the

11

Friedman method was 185.18±4.50 kJ/mol, corresponding to a variation of 2.43%, and

12

that found with the Vyazovkin method was 183.60±7.04 kJ/mol and had 3.43%

13

variation. In this case, the activation energy could be considered stable, and the average

14

values determined through both iso-conversion methods were acceptable for kinetic

15

analysis. The differences between the apparent activation energies determined using the

16

methods were related to the basis of each method. The method of Friedman (1964) is

17

differential, while that of Vyazovkin (2001) is integral; this difference caused the

AL

R N

U

different results. Lee et al. (2017) also analyzed the pyrolysis kinetics of PKS through

JO

18

PR

10

19

the Ozawa method and reported an apparent activation energy of 205.34 kJ/mol. This

20

value was slightly higher than that obtained in this investigation (about 15 kJ/mol);

21

however, this could be due to the nature and geographical origin of the palm species,

22

which affect the thermal behavior of the resulting PKS samples.

18

JOURNAL PRE-PROOF

The iso-conversion analysis of the blend gave results similar to those of EFB. The

2

apparent activation energies determined using the Friedman and Vyazovkin methods

3

were 178.53±26.36 kJ/mol and 164.90±30.00 kJ/mol, with variation percentages of

4

14.77% and 18.18%, respectively. From an analysis of the overall behavior in Fig. 2(a)

5

and the TG/DTG curves presented in Fig. 1(e) and (f), the influence of the EFB was

6

found to be greater than that of PKS in the thermal decomposition process of the blend.

7

This phenomenon was concluded to be due to experimental error during the

8

manipulation of the biomass blend, since three repetitions of this TG analysis were

9

performed, and all gave similar results.

E-

PR

O

O

F

1

Few studies of biomass blends have been published in the literature (Mallick et al.,

11

2018; Salema et al., 2019), and such investigations are normally focused on blends of

12

biomass with coal. Thus, this investigation contributes to the analysis of this common

13

blend. Compared with the results in Lee et al. (2017) and Salema et al. (2019), which

14

reported activation energies of 169.76 kJ/mol and 205.34 kJ/mol for EFB and PKS, the

15

blend in this work presented intermediate behavior.

16

Comparing the overall iso-conversion kinetic behavior of EFB, PKS, and the blend with

17

that of other agricultural residues, their activation energies were found to be on the same

AL

R N

U

order of magnitude, e.g., Mishra and Mohanty (2018) analyzed areca nut husk, pine

JO

18

PR

10

19

sawdust, and sal sawdust, and obtained Ea values of 184.61, 168.58, and 181.53 kJ/mol,

20

respectively. Da Silva et al. (2018) investigated the kinetic and thermal behavior of

21

eucalyptus Benthamii, eucalyptus Dunii, and slash pine, and reported activation

22

energies of 142.98, 147.71, and 155.46 kJ/mol, respectively.

19

JOURNAL PRE-PROOF

3.4. Determination of the reaction model and pre-exponential factor

2

In order to perform a complete kinetic analysis, the kinetic triplet must be completed.

3

Therefore, we determined which reaction model best described the thermal behavior of

4

each biomass and finally, the pre-exponential factor related to the reaction.

5

Table 3 presents the coefficients of determination obtained from the comparison of the

6

experimental and modeled data for the different reaction models using master plots. For

7

all the residues evaluated in this work, the tridimensional diffusion (D3) reaction model

8

best described the experimental behavior. The pre-exponential factor and activation

9

energy were both determined through the linearization of the conversion rate equation

E-

PR

O

O

F

1

using the D3 reaction model. This reaction model is only a mathematical representation

11

of the physical-chemical behavior of the thermal decomposition process. Nonetheless, it

12

was considered to be a rough approximation of the real behavior, since the biomass is

13

composed of multiple components that are related to several complex reactions in the

14

process analyzed.

15

Fig. 2(b), (c), and (d) present the experimental and modeled conversion of EFB, PKS,

16

and the blend as a function of temperature, respectively. In Fig. 2(b), the EFB

AL

R N

conversion was modeled using the kinetic parameters presented in Table 4 and compared with the experimental data obtained at heating rates of 1.25, 2.5, 5, and 10

JO

18

U

17

PR

10

19

°C/min. The modeled curves overestimated the thermal behavior, assuming faster

20

reaction than observed in the experimental data for all the heating rates. However, at 5

21

and 10 °C/min, this effect was only detected at conversions higher than 0.85 and lower

22

than 0.10.

20

JOURNAL PRE-PROOF

Fig. 2(c) compares the experimental and modeled conversion of PKS, for which the

2

kinetic modeling was the worst, with notable differences between the curves. In this

3

case, the model overestimated the data for all the heating rates analyzed, and thus is not

4

recommended for any kind of analysis. This phenomenon occurred due to the complex

5

behavior of PKS. As shown in Fig. 1(d), the DTG exhibited two peaks, which indicated

6

that at least two separate reactions governed the overall thermal decomposition process.

7

Thus, the single-step reaction mechanism is not recommended for the description of the

8

thermal decomposition of PKS.

9

Fig. 2(d) compares the theoretical and modeled conversions as a function of the

10

temperature for the thermal decomposition of the blend. As expected, the blend

11

presented intermediate behavior with respect to that of EFB and PKS, and the modeled

12

curves corresponded to the experimental ones with slight differences, mainly at

13

conversions higher than 0.7.

14

Finally, from the kinetic representations of the thermal decomposition of the tested

15

biomasses by a single-step reaction, it was concluded that this reaction mechanism is

16

not suitable for representing their thermal behavior in analyses with a high accuracy

17

level. The empty fruit bunch (EFB) and the blend could be described using this model

O

O

PR

E-

PR

AL

R N

U

in conditions in which high accuracy representation is not a priority. However, this

JO

18

F

1

19

kinetic analysis is not recommended in any case for the thermal analysis of palm kernel

20

shell (PKS).

21

3.4.1. Mechanism of independent parallel reactions

22

As previously discussed, the single-step reaction did not provide high accuracy

23

modeling of the thermal behavior of the residues analyzed, and therefore, the 21

JOURNAL PRE-PROOF

mechanism of independent parallel reactions (MIPR) was considered. The application

2

of this reaction mechanism to the residues analyzed in this work was not encountered in

3

the literature; that is, information on the application of the MIPR to EFB and PKS is

4

scarce, since the iso-conversion methodology above is typically applied (Zakir et al.,

5

2011; Nyakuma et al., 2015; Dewayanato et al., 2016; Lee et al., 2017; Chan et al.,

6

2018).

7

Thus, Fig. 3 presents the experimental and modeled conversion rates and normalized

8

masses of the different samples (EFB, PKS, and the blend) evaluated at 10 °C/min. In

9

the analysis of this reaction mechanism, three parallel reactions, R1, R2, and R3, were

10

assumed, each of which was associated with the thermal decomposition of a different

11

pseudo-component: R1 with hemicellulose, R2 with cellulose, and R3 with lignin.

12

According to the modeling data, the thermal decomposition of these pseudo-

13

components occurred in the temperature ranges 150-300 °C, 215-372 °C and 100-900

14

°C, respectively. This behavior was expected, since these ranges were similar to those

15

obtained by other researchers (Yang et al., 2007; Cai et al., 2013).

16

As shown in Fig. 3(a), 3(c), and 3(e), the reaction R1 occurred in the same temperature

17

range for all three of the analyzed samples; however, the highest conversion rate

O O

PR

E-

PR

AL

R N

U

amplitude was registered for PKS (1 s-1). This indicated that the hemicellulose of PKS

JO

18

F

1

19

reacted faster than that in EFB or the blend. This effect was related to the nature of each

20

biomass; however, in the case of the blend, the influence of EFB was higher than that of

21

PKS.

22

The cellulose reaction (R2), however, presented different behavior for each biomass.

23

The decomposition temperature range was higher for PKS than for EFB and the blend.

22

JOURNAL PRE-PROOF

The decomposition temperature ranges related to R2 were 225-325 °C (EFB), 250-375

2

°C (PKS), and 225-350 °C (blend). The differences among the residues was attributed to

3

the nature of the cellulose in each biomass. The cellulose reaction of the blend presented

4

intermediate behavior, which was expected, but the shape of its conversion rate curve

5

more closely resembled that of the EFB curve, indicating the strong influence of this

6

residue in the blend.

7

Finally, the third reaction (R3), which was related to the thermal decomposition of

8

lignin, occurred over a broad temperature range from 150 to 700 °C, as is commonly

9

observed in MIPR investigations (Sfakiotakis and Vamvuka, 2015; Yu et al., 2015;

10

Xavier et al., 2016). This wide reaction range is considered to be normal, since this

11

biomass component is known to have an amorphous structure in the solid state, and thus

12

releases several volatile components as the temperature is increased and is the precursor

13

of the final carbonaceous solid in pyrolysis (Hatakeyama, and Hatakeyama, 2009).

14

Table 4 summarizes the kinetic parameters obtained using the three-parallel reaction

15

mechanism for the studied samples. The kinetic parameters for this reaction mechanism

16

were obtained using fitting, as described previously in the materials and methods

17

section.

O O

PR

E-

PR

AL

R N

U

In investigations of the thermal decomposition kinetics of biomass using the MIPR

JO

18

F

1

19

(Santos et al., 2011; Rueda-Ordóñez et al., 2015; Xavier et al., 2016), the kinetic

20

parameters for each reaction are typically presented as average values with the

21

corresponding standard deviations obtained from the analyses at several heating rates.

22

Therefore, this averaged kinetic parameter, e.g. the activation energy, may provide good

23

results for one heating rate, but not others. In this work, in order to obtain reliable

23

JOURNAL PRE-PROOF

kinetic parameters, the average value methodology was avoided. Instead, only one

2

activation energy was selected for each reaction and fraction, and was used for all the

3

heating rates. The pre-exponential factor was used as a fitting parameter. This process

4

was not carried out randomly, but instead was conducted by applying a polynomial

5

function to describe its variation for the heating rates analyzed.

6

Therefore, in Table 4 the activation energy related to the thermal decomposition

7

reaction of hemicellulose was 85 kJ/mol for EFB, 107 kJ/mol for PKS, and 80 kJ/mol

8

for the blend. The activation energy of cellulose was 194 kJ/mol (EFB), 200 kJ/mol

9

(PKS), and 190 kJ/mol (blend), and that of the lignin reaction was 49 kJ/mol for PKS

10

and EFB, and 49.23 kJ/mol for the blend. According to Rueda-Ordóñez et al. (2015),

11

who collected kinetic parameters from more than 20 works, the pseudo-component

12

activation energy of hemicellulose in the MIPR varies between 75 and 150 kJ/mol, that

13

of cellulose ranges from 125 to 225 kJ/mol, and that of lignin is 25 and 100 kJ/mol.

14

Thus, the values determined in the present study were within the ranges presented in

15

this previous work.

16

The pre-exponential factor was used as a fitting parameter, and its variation with the

17

heating rate (β) was correlated using a first or second order polynomial as appropriate.

O

O

PR

E-

PR

AL

R N

U

The variation of the pre-exponential factor in the PKS and EFB showed a linear trend,

JO

18

F

1

19

and therefore, was represented using first-order polynomial. However, in the case of the

20

blend, a linear correlation could not be achieved for reactions R2 and R3, and therefore

21

a second-order polynomial was proposed for each reaction.

22

In summary, the application of the mechanism of independent parallel reactions (MIPR)

23

resulted in a very good representation of the experimental data. It was also concluded

24

JOURNAL PRE-PROOF

that reliable and accurate results can be achieved by controlling the variation of the pre-

2

exponential factor while holding the other kinetic parameters constant. Thus, it is

3

recommended that future works correlate the pre-exponential factor with the heating

4

rate in order to avoid errors in the results, which should add scientific rigor to the

5

proposed models.

6

3.4.2. Mechanism of consecutive reactions

7

To obtain a complete kinetic analysis that contributes to the development of the thermal

8

analysis of biomass, the implementation of a mechanism involving three consecutive

9

reactions was also analyzed. This approach is commonly used in solid-state reactions;

E-

PR

O

O

F

1

however, for the analysis of the thermal decomposition of biomass, the more typical and

11

accepted method is the MIPR. Nonetheless, several works in the literature

12

(Luangkiattikhun et al., 2008; Weerachanchai et al., 2010; Lopes et al., 2016; Rueda-

13

Ordóñez and Tannous, 2018) have applied this reaction mechanism to represent the

14

thermal decomposition behavior of biomass.

15

Fig. 4 presents the normalized mass as a function of the temperature obtained by

16

applying the mechanism of consecutive reactions (MCR). The results of the analysis of

AL

R N

EFB using the MCR are presented in Fig. 4 (a) and (b); the model represented the normalized mass experimental data very well. However, in the normalized DTG, the

JO

18

U

17

PR

10

19

model underestimated the experimental behavior between 275 and 325 °C. This was

20

attributed to the change in the inclination of the curve, which was related to the

21

interactions among the components in the reaction. Fig. 4(b) presents the concentration

22

as a function of temperature. The initial component, A, decomposed at temperatures up

23

to 330 °C to form B, which reached a maximum of 0.74 at 300 °C, and then rapidly

25

JOURNAL PRE-PROOF

decomposed. Component C was rapidly formed and decomposed slowly. At 900 °C, its

2

concentration was 0.5, which indicated that this component was a precursor of the final

3

carbonaceous solid.

4

Fig. 4 (c) and (d) show the normalized mass and DTG of PKS, and the concentrations of

5

the components as a function of temperature, respectively. The model overestimated the

6

data related to the first peak in the DTG and underestimated that of the second peak;

7

however, the model fitted the experimental data with an average deviation of less than

8

4%. The blend data presented in Fig. 4 (e) and (f) presented the best fitting of the

9

experimental data. The concentrations of PKS and the blend showed similar trends to

E-

PR

O

O

F

1

those of the EFB.

11

As observed in a previous work (Rueda-Ordóñez and Tannous, 2018), the kinetic

12

parameters such as the activation energy, pre-exponential factor, and reaction model

13

obtained using the mechanism of parallel reactions could be used in the consecutive

14

reactions mechanism with reliable results. Thus, in this work the same kinetic

15

parameters presented previously in Table 3 were used for the modeling of the MCR, and

16

the results were satisfactory, with an average deviation of less than 4% for all the

17

analyses.

AL

R N

U

4. Conclusions

JO

18

PR

10

19

The characterization showed that blending EFB with PKS improve EFB thermal

20

characteristics. It is recommended using the EFB for combustion or gasification, and the

21

blend and PKS for pyrolysis based on their volatile content. The single-step reaction

22

serves to describe the EFB and blend thermal behavior when high accuracy is not a

23

priority. The MIPR and MCR represented very well the thermal decomposition of the

26

JOURNAL PRE-PROOF

biomasses, and the kinetic parameters were the same. Finally, an aspect to highlight is

2

the reliability obtained correlating the pre-exponential factor with heating rate, then, is

3

highly recommended for future works in the field.

4

Acknowledgements

5

This research did not receive any specific grant from funding agencies in the public,

6

commercial, or not-for-profit sectors.

7

References

8

[1]. Abdullah, N., Gerhauser, H., 2008. Bio-oil derived from empty fruit bunches.

O PR

E-

10

Fuel 87, 2606–2613. https://doi.org/10.1016/j.fuel.2008.02.011 [2]. Anca-Couce, A., Berger, A., Zobel, N., 2014. How to determine consistent

PR

9

O

F

1

biomass pyrolysis kinetics in a parallel reaction scheme. Fuel 123, 230–240.

12

https://doi.org/10.1016/j.fuel.2014.01.014

AL

11

[3]. Asadullah, M., Rasid, N., Kadir, S., Azdarpour, A., 2013. Production and detailed

14

characterization of bio-oil from fast pyrolysis of palm kernel shell. Biomass

16

[4]. Bartocci, P., Anca-Couce, A., Slopiecka, K., Nefkens, S., Evic, N., Retschitzegger, S., Barbanera, M., Buratti, C., Cotana, F., Bidini, G., Fantozzi, F.,

JO

17

Bioenergy 59, 316-324. https://doi.org/10.1016/j.biombioe.2013.08.037

U

15

R N

13

18

2017. Pyrolysis of pellets made with biomass and glycerol: Kinetic analysis and

19

evolved gas analysis. Biomass Bioenergy 97, 11-19.

20

[5]. Brown, W.E., Dollimore, D., Galwey, A.K., 1980. Theory of solid-state kinetics.

21

In: Bamford, C.H., Tipper, C.F.H. (Ed.) Chemical kinetics volume 22: reactions in

22

the solid state. Elsevier scientific publishing company, Amsterdam, pp. 41-113.

27

JOURNAL PRE-PROOF

1

[6]. Cai, J., Wu, W., Liu, R., Huber, G., 2013. A distributed activation energy model

2

for the pyrolysis of lignocellulosic biomass. Green Chem. 15, 1331-1340.

3

https://doi.org/10.1039/C3GC36958G [7]. Chan, Y., Quitain, A., Yusup, S., Uemura, Y., Sasaki, M., Kida, T., 2018.

F

4

Liquefaction of palm kernel shell to bio-oil using sub- and supercritical water: An

6

overall kinetic study. J. Energ. Inst. 92, 535-541.

7

https://doi.org/10.1016/j.joei.2018.04.005

11

O

PR

10

kinetic evaluation by single-step for waste wood from reforestation. Waste Manag. 72, 265-273. https://doi.org/10.1016/j.wasman.2017.11.034

E-

9

[8]. Da Silva, J., Alves, J., Galdino, W., Andersen, S., Sena, R., 2018. Pyrolysis

[9]. Dewayanato, N., Azman, A., Ahmad, N., Mohd, M., 2016. Study of thermal

PR

8

O

5

12

degradation of biomass wastes generated from palm oil milling plant. CHEMICA

13

Jurnal Teknik Kimia 3, 31-37. doi: 10.26555/chemica.v3i2.5860 [10]. Dupont, C., Nocquet, T., Da Costa Jr., J.A., Verne-Tournon, C., 2011. Kinetic

15

modelling of steam gasification of various woody biomass chars: Influence of

16

inorganic elements. Bioresour. Technol. 102, 9743-9748.

[11]. Escalante-Hernández, H., Orduz-Prada, J., Zapata-Lesmes, H.J., Cardona-Ruiz, M., Duarte-Ortega, M., 2011. Atlas del Potencial Energético de la Biomasa

JO

19

R N

18

https://doi.org/10.1016/j.biortech.2011.07.016

U

17

AL

14

20 21

Residual en Colombia, first ed., Ediciones UIS, Bucaramanga.

[12]. Fedepalma, 2018. Con récord en producción de aceite de palma, sector palmero

22

colombiano cierra 2017 con balance positivo. http://web.fedepalma.org/con-

23

record-en-produccion-de-aceite-de-palma-sector-palmero-colombiano-cierra-

24

2017-con-balance-positivo (accessed 17 August 2018).

28

JOURNAL PRE-PROOF

1

[13]. Friedman, H.L., 1964. Kinetics of thermal degradation of char-forming plastics

2

from thermogravimetry. Application to a phenolic plastic. J. Polym. Sci.: Polym.

3

Symp. 6, 183–195. https://doi.org/10.1002/polc.5070060121 [14]. Gutierrez-Sanin, F., Marín-Jaramillo, M., 2018. Tierras en el posconflicto: ¿en el

F

4

fondo cual es el problema?, Análisis Político. Enero-Abril, 18-38.

6

[15]. Hatakeyama H., Hatakeyama T., 2009. Lignin Structure, Properties, and

O

5

Applications. In: Abe A., Dusek K., Kobayashi S. (eds) Biopolymers. Advances

8

in Polymer Science 232. Springer, Berlin, Heidelberg

9

https://doi.org/10.1007/12_2009_12

12

PR

E-

11

[16]. IndexMundi, 2019. Palm oil production by country in 1000 Mt. Retrieved from https://www.indexmundi.com/agriculture/?commodity=palm-oil

PR

10

O

7

[17]. Jenkins, B., Baxter, L., Miles, T., Miles, T., 1998. Combustion properties of biomass. Fuel Process. Technol. 54, 17-46. https://doi.org/10.1016/s0378-

14

3820(97)00059-3

17 18

https://doi.org/10.1016/j.egypro.2015.11.552

[19]. Ku, X., Lin, J., Yuan, F., 2016. Influence of torrefaction on biomass gasification performance in a high-temperature entrained-flow reactor. Energy Fuels 305,

JO

19

analysis of the pyrolysis of rice straw. Energy Procedia 79, 663-670.

R N

16

[18]. Kongkaew, N., Pruksakit, W., Patumsawad, S., 2015. Thermogravimetric kinetic

U

15

AL

13

20 21

4053-4064. https://doi.org/10.1021/acs.energyfuels.6b00163

[20]. Kuo, P., Wu, W., Chen, W., 2014. Gasification performances of raw and torrefied

22

biomass in a downdraft fixed bed gasifier using thermodynamic analysis. Fuel

23

117, 1231-1241. https://doi.org/10.1016/j.fuel.2013.07.125

29

JOURNAL PRE-PROOF

1

[21]. Lee, X., Lee, L., Gan, S., Thangalazhy-Gopakumar, S., Ng, H., 2017. Biochar potential evaluation of palm oil wastes through slow pyrolysis: Thermochemical

3

characterization and pyrolytic kinetic studies. Bioresour. Technol. 236, 155-163.

4

https://doi.org/10.1016/j.biortech.2017.03.105

F

2

[22]. Lopes, F.C.R., Tannous, K., Rueda-Ordóñez, Y.J., 2016. Combustion reaction

6

kinetics of guarana seed residue applying isoconversional methods and

7

consecutive reaction scheme. Bioresour. Technol. 219, 392-402.

O

O

5

[23]. Luangkiattikhun, P., Tangsathitkulchai, P., Tangsathitkulchai, M., 2008. Non-

9

isothermal thermogravimetric analysis of oil-palm solid wastes. Bioresour. Technol. 99, 986-997.

E-

10

PR

8

[24]. Mallick, D., Poddar, M. K., Mahanta, P., Moholkar, V. S., 2018. Discernment of

12

synergism in pyrolysis of biomass blends using thermogravimetric analysis.

13

Bioresour. Technol. 261, 294-305.

PR

11

[25]. Mishra, R., Mohanty, K., 2018. Pyrolysis kinetics and thermal behavior of waste

15

sawdust biomass using thermogravimetric analysis. Bioresour. Technol. 251, 63-

16

74. https://doi.org/10.1016/j.biortech.2017.12.029

empty fruit bunch in a downdraft reactor: A pilot scale approach. Bioresour. Technol. Reports 1, 39-49. https://doi.org/10.1016/j.biteb.2018.02.001

JO

19

R N

18

[26]. Monir, M.U., Aziz, A.A., Kristanti, R.A., Yousuf, A., 2018. Co-gasification of

U

17

AL

14

20

[27]. Nyakuma, B., Ahmad, A., Abdullah, T., Oladokun, O., Aminu, D., 2015. Non-

21

isothermal kinetic analysis of oil palm empty fruit bunch pellets by

22

thermogravimetric analysis. Chem. Eng. Trans. 45, 1327-1332.

23

https://doi.org/10.3303/CET1545222

30

JOURNAL PRE-PROOF

1

[28]. Parikh, J., Channiwala, S., Ghosal, G., 2007. A correlation for calculating

2

elemental composition from proximate analysis of biomass materials. Fuel 86,

3

1710-1719. https://doi.org/10.1016/j.fuel.2006.12.029 [29]. Pérez-Maqueda, L., Criado, J., 2000. The accuracy of Senum and Yang's

F

4

approximations to the arrhenius integral. J. Therm. Anal. Calorim. 60, 909-915.

6

https://doi.org/10.1023/a:1010115926340

[30]. Reinoso-Pérez, Y., Estrella-Martínez, D., Alturo-Mendigaña, S., Gámez-Móvil,

O

7

O

5

E., 2019. El postconflicto como proceso de restablecimiento de derechos a las

9

víctimas del desplazamiento forzado, Saber, ciencia y libertad, enero-junio, 35-47.

E-

11

https://doi.org/10.18041/2382-3240/saber.2019v14n1.5202

[31]. Rueda-Ordóñez, Y.J., Tannous, K., 2015. Isoconversional kinetic study of the

PR

10

PR

8

12

thermal decomposition of sugarcane straw for thermal conversion processes.

13

Bioresour. Technol. 196, 136-144. https://doi.org/10.1016/j.biortech.2015.07.062 [32]. Rueda-Ordóñez, Y.J., Tannous, K., Olivares-Gómez, E., 2015. An empirical

AL

14

model to obtain the kinetic parameters of lignocellulosic biomass pyrolysis in an

16

independent parallel reactions scheme. Fuel Process. Technol. 140, 222-230.

18

[33]. Rueda-Ordóñez, Y.J., Tannous, K., 2017. Análisis cinético de la descomposición térmica de biomasas aplicando un esquema de reacciones paralelas

JO

19

https://doi.org/10.1016/j.fuproc.2015.09.001

U

17

R N

15

20 21

independientes. UIS Ingenierías 16, 119-128.

[34]. Rueda-Ordóñez, Y.J., Tannous, K., 2018. Drying and thermal decomposition

22

kinetics of sugarcane straw by nonisothermal thermogravimetric analysis.

23

Bioresour. Technol. 264, 131-139. https://doi.org/10.1016/j.biortech.2018.04.064

31

JOURNAL PRE-PROOF

1

[35]. Salema, A. A., Ting, R. M. W., Shang, Y. K., 2019. Pyrolysis of blend (oil palm

2

biomass and sawdust) biomass using TG-MS. Bioresour. Technol. 274, 439-446.

3

[36]. Santos, K., Lira, T., Gianesella, M., Lobato, F., Murata, V., Barrozo, M., 2011. Bagasse pyrolysis: a comparative study of kinetic models. Chem. Eng. Commun.

5

199, 109-121. https://doi.org/10.1080/00986445.2011.575906

O

6

F

4

[37]. Sfakiotakis, S., Vamvuka, D., 2015. Development of a modified independent parallel reactions kinetic model and comparison with the distributed activation

8

energy model for the pyrolysis of a wide variety of biomass fuels. Bioresour.

9

Technol. 197, 434-442. https://doi.org/10.1016/j.biortech.2015.08.130

PR

[38]. Trop, P., Agrez, M., Urbancl, D., Goricanec, D., 2016. Co-gasification of torrefied

E-

10

O

7

wood biomass and sewage sludge. Comput. Aided Chem. Eng. 38, 2229-2234.

12

https://doi.org/10.1016/B978-0-444-63428-3.50376-3

16 17 18

[40]. Vassilev, S.V., Baxter, D., Andersen, L.K., Vassileva, C.G., 2013. An overview of the composition and application of biomass ash. Part 2. Potential utilisation, technological and ecological advantages and challenges. Fuel 105, 19-39. https://doi.org/10.1016/j.fuel.2012.10.001

[41]. Vassilev, S.V., Vassileva, C.G., Song, Y.C., Li, W.Y., Feng, J., 2017. Ash

JO

19

AL

15

biomasa. Palmas 37 (Especial Tomo II) 149-156.

R N

14

[39]. Van Dam, J., 2016. Subproductos de la palma de aceite como materias primas de

U

13

PR

11

20

contents and ash-forming elements of biomass and their significance for solid

21

biofuel combustion. Fuel 208, 377-409.https://doi.org/10.1016/j.fuel.2017.07.036

22

[42]. Vecino-Mantilla, J.S., Gauthier-Maradei, M.P., Álvarez-Gil, P.J., Tarazona-

23

Cardenas, S.Y., 2014. Comparative study of bio-oil production from sugarcane

24

bagasse and palm empty fruit bunch: Yield optimization and bio-oil

32

JOURNAL PRE-PROOF

1

characterization. J. Anal. Appl. Pyrolysis. 108, 284-294.

2

https://doi.org/10.1016/j.jaap.2014.04.003

3

[43]. Vyazovkin, S., 2001. Modification of the integral isoconversional method to account for variation in the activation energy. J. Comput. Chemi. 22, 178-183.

5

https://doi.org/10.1002/1096-987x(20010130)22:2<178::aid-jcc5>3.0.co;2-#

O

F

4

[44]. Vyazovkin, S., Burnham, A., Criado, J., Pérez-Maqueda, L., Popescu, C.,

7

Sbirrazzuoli, N., 2011. ICTAC Kinetics Committee recommendations for

8

performing kinetic computations on thermal analysis data. Thermochim. Acta.

9

520, 1-19. https://doi.org/10.1016/j.tca.2011.03.034

PR

[45]. Weerachanchai,

P., Tangsathitkulchai, C., Tangsathitkulchai, M., 2010.

E-

10

O

6

Comparison of pyrolysis kinetic models for thermogravimetric analysis of

12

biomass. Suranaree J. Sci. Technol. 17, 387-400.

13

PR

11

[46]. White, J.E., Catallo, W.J., Legendre, B.L., 2011. Biomass pyrolysis kinetics: a comparative critical review with relevant agricultural residue case studies. J. Anal.

15

Appl. Pyrolysis. 91, 1–33. https://doi.org/10.1016/j.jaap.2011.01.004

AL

14

[47]. Xavier, T., Lira, T., Schettino Jr., M., Barrozo, M., 2016. A study of pyrolysis of

17

macadamia nut shell: parametric sensitivity analysis of the ipr model. Brazilian J.

[48]. Yang, H., Yan, R., Chen, H., Lee, D., Zheng, C., 2007. Characteristics of

JO

19

Chem. Eng. 33, 115-122. https://doi.org/10.1590/0104-6632.20160331s00003629

U

18

R N

16

20

hemicellulose, cellulose and lignin pyrolysis. Fuel 86, 1781-1788.

21

https://doi.org/10.1016/j.fuel.2006.12.013

22

[49]. Yu, H., Liu, F., Ke, M., Zhang, X., 2015. Thermogravimetric analysis and kinetic

23

study of bamboo waste treated by Echinodontium taxodii using a modified three-

33

JOURNAL PRE-PROOF

1

parallel-reactions model. Bioresour. Technol. 185, 324-330.

2

https://doi.org/10.1016/j.biortech.2015.03.005

3

[50]. Zakir, K., Yusup, S., Murni, M., Uemura, Y., Vuoi, S., Rashid, U., Inayat, A., 2011. Kinetic study on palm oil waste decomposition, in M.A. Bernardes (Ed.),

5

Biofuel’s Engineering Process Technology, InTechOpen, Rijeka, pp. 523-536.

6

https://doi.org/10.5772/16458

O

F

4

JO

U

R N

AL

PR

E-

PR

O

7

34

JOURNAL PRE-PROOF

Figure 1. TG and DTG curves of thermal decomposition of (a) and (b) EFB, (c) and (d)

2

PKS, and (e) and (f) blend, in an atmosphere of argon at the studied heating rates.

3

Figure 2. (a) Activation energy as a function of conversion; (b) EFB conversion (c)

4

PKS conversion and (d) Blend conversion as a function of temperature.

5

Figure 3. Conversion rate and normalized mass as a function of temperature at

6

10°C/min of (a) and (b) EFB, (c) and (d) PKS, and (e) and (f) blend.

7

Figure 4. Normalized mass and DTG, and concentration as a function of the

8

temperature of (a) and (b) EFB, (c) and (d) PKS, and (e) and (f) blend obtained with

9

MCR.

PR

O

O

F

1

E-

10

PR

11 12

Table 1. Mechanism of parallel and consecutive reactions tested in this study.

14

Table 2. Physical-chemical properties of the EFB, PKS and their blend.

15

Table 3. Comparison of reaction models and their related kinetic parameters.

16

Table 4. MIPR Kinetic parameters for each of the studied samples.

JO

19

R N

18

U

17

AL

13

20 21 22 23 24

35

JOURNAL PRE-PROOF

JO

2

U

R N

AL

PR

E-

PR

O

O

F

1

3

Figure 1. TG and DTG curves of thermal decomposition of (a) and (b) EFB, (c) and (d)

4

PKS, and (e) and (f) blend, in an atmosphere of argon at the studied heating rates.

5 6 7

36

JOURNAL PRE-PROOF

PR

E-

PR

O

O

F

1

AL

2

Figure 2. (a) Activation energy as a function of conversion; (b) EFB conversion (c)

4

PKS conversion and (d) Blend conversion as a function of temperature.

6

JO

7

U

5

R N

3

8 9

10 11 12 13

37

JOURNAL PRE-PROOF

JO

2

U

R N

AL

PR

E-

PR

O

O

F

1

3

Figure 3. Conversion rate and normalized mass as a function of temperature at

4

10°C/min of (a) and (b) EFB, (c) and (d) PKS, and (e) and (f) blend.

5 6 7

38

JOURNAL PRE-PROOF

JO

2

U

R N

AL

PR

E-

PR

O

O

F

1

3

Figure 4. Normalized mass and DTG, and concentration as a function of the

4

temperature of (a) and (b) EFB, (c) and (d) PKS, and (e) and (f) blend obtained with

5

MCR.

6 7

39

JOURNAL PRE-PROOF

1 2 3

Table 1. Mechanism of parallel and consecutive reactions tested in this study. Parallel reactions  r1=k1[C]

1) Dry biomass (A)Torrefied r1=k1[A] biomass (B)

F

(C)

O

1) Cellulose products

Consecutive reactions

2) Torrefied biomass r2=k2[B] (B)carbonaceous solid (C)

3) Lignine (L)  products

3) Carbonaceous (C)Final residue (D)

O

2) Hemicellulose (HC)  r2=k2[HC] products

PR

r3=k3[L]

4

E-

5

PR

6 7 8

AL

9

12 13

JO

14

U

11

R N

10

15 16 17 18 19

40

solid r3=k3[C]

JOURNAL PRE-PROOF

1 2 Table 2. Physical-chemical properties of the EFB, PKS, and their blend.

8.19 ± 0.036 3.74 ± 0.06

5.43 ± 0.05

VM

78.92 ± 0.30 85.38 ± 1.57

83.68 ± 0.56

19.71 ± 0.20 20.22 ± 0.21

19.97 ± 0.21

LHV

18.43

18.63

AL U

JO

9

18.89

R N

6

8

PR

HHV 4

7

E-

12.88 10.88 10.89 Ultimate analysis (mass%) Dry basis 44.12 45.78 45.01 5.56 5.86 5.75 41.48 43.95 43.14 Heating Value (MJ/Kg) Dry basis

PR

C H O

O

Ash

FC

5

O

(Biomass) EFB PKS Blend Proximate analysis (mass%) Dry basis M 6.43 ± 0.23 6.28 ± 0.16 7.17 ± 0.24

10 11 12 13 14 15

41

F

3

JOURNAL PRE-PROOF

1 2 Table 3. Comparison of reaction models and their related kinetic parameters 𝑹𝟐 D1

D2

D3

PKS

0.7551 0.7449 0.8861 0.9397 0.9209

158.24

24.57

BLEND

0.8592 0.923

170.07

28.55

0.9616 0.9738 0.9698

EPR

7 8

AL

9

U

JO

14

R N

10

15 16 17 18 19

42

30.11

O

175.77

6

13

LnA (ln s −1 )

0.8617 0.9197 0.9632 0.9778 0.9729

5

12

E (kJ/mol)

EFB

4

11

D4

F

F1

Kinetic Parameters

O

Biomass

PR

3

JOURNAL PRE-PROOF

JO

U

R N

AL

PR

E-

PR

O

O

F

1

43

JOURNAL PRE-PROOF

Table 4. MIPR Kinetic parameters for each of the studied samples. EFB

β (°C/min) AD

1.25 5.319

Ea (kJ/mol) A equation F

R1 (HC) 107.00 𝑙𝑛𝐴 = 0,0813𝛽 + 17,641 0.346

β (°C/min) AD

1.25 3.399

Ea (kJ/mol) A equation F

R1 (HC) 80.00 𝑙𝑛𝐴 = 0,0863𝛽 + 11.533 0.3417

Β (°C/min) AD

1.25 3.136

R2 (C) 194.00 𝑙𝑛𝐴 = 0,0074𝛽 + 35,272 0.3631 Fitting Error 5.00 4.818 PKS R2 (C) 200.00 𝑙𝑛𝐴 = 34,762 0.37 Fitting Error 5.00 3.156 BLEND R2 (C) 190.00

PR

O

O

2.50 4.487

PR

E-

2.50 3.117

𝑙𝑛𝐴 = 0,136𝛽2 + 0.1466𝛽 + 34.246

AL

R N U

2.50 2.767

R3 (L) 49.00 𝑙𝑛𝐴 = 0,2716𝛽 + 1.1125 0.3112

F

Ea (kJ/mol) A equation F

R1 (HC) 85.00 𝑙𝑛𝐴 = 0,0607𝛽 + 12,998 0.3257

0.344 Fitting Error 5.00 4.926

44

10.00 3.211

R3 (L) 49.00 𝑙𝑛𝐴 = 0,2009𝛽 + 1.1526 0.284 10.00 3.479 R3 (L) 49.20

𝑙𝑛𝐴 = 0,05𝛽2 + 0.4296𝛽 + 2.0551

0.3143 10.00 1.890

U

R N

AL

PR

E-

PR

O

O

F

JOURNAL PRE-PROOF

45

JOURNAL PRE-PROOF

Highlights

U

R N

AL

PR

E-

PR

O

O

F

 Blending PKS with EFB increase 5% heating value and decrease 50% ash content.  Correlating the pre-exponential factor with heating rate increase accuracy.  Kinetic parameters are equal for consecutive or parallel reactions.

46