Open access peer-reviewed chapter

Analysis of Pyrolysis Kinetic Model for Processing of Thermogravimetric Analysis Data

By Guodong Jiang and Liping Wei

Submitted: March 3rd 2018Reviewed: May 31st 2018Published: November 5th 2018

DOI: 10.5772/intechopen.79226

Downloaded: 323

Abstract

Pyrolysis has profound implications for coal as a raw material to make phase change material (PCM). It is necessary to derive a pyrolysis kinetic model for predicting the yield of volatiles and reaction performance during pyrolysis of coal, which is of significant importance for its thermal processing. The devolatilization of coal is characterized by thermogravimetric analysis (TGA) at different heating rates, and many kinetic models can be achieved by analyzing the TGA data. This work was aimed to find an appropriate model to describe the pyrolysis of coal and took Zhundong coal as an example. Four types of isoconversion kinetic methods, that is, Friedman, Flynn-Wall-Ozawa (FWO), Kissinger-Akahira-Sunose (KAS), Miura-Maki method, and different distributed activation energy models (DAEM) were employed here to fit TGA data for pyrolysis of Zhundong coal. The pre-exponential factors and activation energies obtained from different kinetic models were analyzed. An m-nth-DAEM was developed by considering that m classes of reactions took place with the same pre-exponential factor k0 but different distribution activation energy following logistic distribution or Gaussian distribution. The results showed that the FWO model was better for description of pyrolysis process of Zhundong coal, and the 2-nth-DAEM assuming Gaussian distribution of activation energy gave the best fit for the TGA data of Zhundong coal. The research provides a valuable reference to the development of thermal utilization technology of Zhundong coal.

Keywords

  • phase change material (PCM)
  • Zhundong coal
  • pyrolysis
  • thermogravimetry analysis
  • isoconversional methods
  • kinetics
  • distributed activation energy model

1. Introduction

Energy storage technology can solve the contradiction between energy supply and demand in time and space, so it is an effective way to improve energy utilization [1]. Thermal energy storage is widely used in industrial and civilian applications, so it occupies an extremely important position in the field of energy storage technology. Phase change materials (PCM) will absorb or release a large amount of latent heat for energy storage when the state of matter changes. As a raw material for the preparation of phase change materials, coal has great energy storage density, and the output temperature and energy are relatively stable, so it has good application prospects [2]. Zhundong coalfield is the largest compositive basin in China, which has a large reserve of 390 Gt and is estimated to meet China’s next 100-year coal consumption requirement [3]. In addition, Zhundong coal is more environmentally friendly than other types of coal owing to its extremely low sulfur and ash contents, which has a great significance to coal industry [4, 5]. Therefore, the investigation on Zhundong coal has theoretical and practical significance for the utilization of coal resources of China.

Thermal pyrolysis is the first stage reaction in most of coal thermal conversion processes (combustion, gasification, and liquefaction) and it is also one of the most promising technologies for clean and effective utilization of low-rank coal. During the pyrolysis process, various non-condensible gas and liquid phase matter such as tar are generated, which will lead to the weightlessness of coal samples. The weight losses of the coal range from 10 to 55%, which is affected by pyrolysis temperature, coal size, pyrolysis time and other factors such as pressure, atmosphere, heating rate and mineral content, etc. [6, 7]. The heating rate and pyrolysis temperature have an important influence on the pyrolysis characteristics of coal. The heating rate affects the concentration gradient of the product inside the particle and the ratio of gas phase substance to tar in the volatiles. The pyrolysis temperature determines the pyrolysis of the macromolecule reaction rate in the coal structure and the total release of volatile matters. A thermogravimetric analyzer (TGA) is routinely employed by many researchers to characterize moisture and volatile contents of various coals. It enables the user to monitor the weight change of a certain quality of coal samples as a function of time and temperature, which endue this method to incorporate the advantages of simple operation and good repeatability [8]. From the TG curve, we can obtain the key parameters such as the temperature of initial point of pyrolysis volatilization, the maximum release rate of volatile matter, and its corresponding temperature.

It is necessary to obtain an accurate knowledge of the pyrolysis process, since all the models of coal thermal conversion processes involve pyrolysis, and the kinetic behavior of these products in pyrolysis is fundamental to the optimization of their use. However, the actual phenomena of pyrolysis are complex and the kinetic parameters of which are difficult to obtain. To date, a large number of kinetic models have been proposed in the literature [5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15]. The models for processing TGA data and explaining the kinetic process of coal pyrolysis include first-order single reaction model (SRM), isoconversional or model-free methods, distributed activation energy model (DAEM). The SRM is the simplest, and it is applicable only over a narrow range of temperatures, thus limiting its applicability. Isoconversional or model-free methods are also used to process TGA data in which the kinetics parameters can be obtained without an explicit kinetics model [9]. Friedman method [10], Kissinger method [11], Flynn-Wall-Ozawa [12], and Miura-Maki method [13] are some of the outstanding methods in this classification. Several TGA curves at different heating rates for the same value of conversion are required for these models to determine activation energy for each conversion point. In recent years, nonlinear regressions (least square optimization) are increasingly used to process the pyrolysis process of coals. More often than not, the distributed activation energy model (DAEM), originally developed by Pitt [14], has become the distinguished representative of them. It assumes that the coal pyrolysis should include a large number of independent, parallel, and irreversible n-order reactions with different activation energies. This distribution of activation energy corresponding to these infinite reactions is often represented by a distribution function of activation energy fE, which is usually used as logistic distribution or Gaussian distribution. Based on the DAEM, a two-Gaussian DAEM (2-DAEM) [15] has been developed by dividing the pyrolysis process into two steps. According to these indexes, the distributed activation energy model will be better improved and applied.

In the present paper, TGA experiments with pyrolysis of Zhundong coal at different heating rate have been carried out. The data was first processed using various isoconversional methods and then processed using the distributed activation energy model (DAEM). An m-nth-DAEM was developed to predict the weight loss of Zhundong coal, considering that mclasses of reactions take place having the same pre-exponential factor k0and different distribution of activation energy.

2. Experiment

The experimental tests were with a TGA (Netzsch Model STA449F3). Before the pyrolysis experiments, coals were grounded and the finer particle fractions in a size range 0.075–0.15 (mm) were selected. Table 1 shows the results of the proximate and ultimate analyses of the coal samples. Small samples were used to ensure uniform heating and to avoid problems of transport phenomena through the sample bed in the crucible. The instrument was first purged for 30 min with nitrogen to remove any remains of air, then further heated from room temperature to 1000°C at a heating rate of 10 Kmin1, 20 Kmin1, 30 Kmin1and 50 Kmin1, respectively. The tests were conducted in an N2inert atmosphere, and the samples were maintained at the maximum temperature for 10 min until a constant mass was obtained at this temperature. All experiments at given heating rate were conducted two times and averaged to eliminate measure error. The experiments showed reproducibility within the range of ±3.7%.

Proximate analysis (%)Ultimate analysis (%)
MadAadVadFCadCdHdOdNdSt,d
9.9615.4622.4952.0970.553.497.640.620.53

Table 1.

Proximate and ultimate analyses of raw Zhundong coal samples.

Typical decomposition data of Zhundong coal obtained from the TGA experiments [9] are shown in Figure 1. The key characteristic parameters related to the main pyrolysis process are summarized in Table 2. The initial decomposition temperature is denoted as T0, the corresponding temperature at the maximum point of the weight loss rate is denoted as Tp, the maximum weight loss rate at Tpis denoted as R, and Vfis the weight loss corresponding to the final temperature. As can be seen from Figure 1, the mass fraction of releasing volatiles decreases with the increase of heating rate at the same temperature. The reason is that the increasing heating rate results in a thermal hysteresis, which makes the temperature inside the sample lower than the appearance, thereby leading to the value of 1αshifts to lager temperature. As can be seen from Figure 1, the TGA curve first decreases slightly due to small molecular gases being removed as the weak bonds are decomposed. Then tar-based volatile components and the formation of char result in a significant decrease of the TGA curve. At the end of the curve, it is close to horizontal, where a cross-linking reaction occurs. The TGA curve contains the main chemical reactions in the coal pyrolysis process, so it is the basis for the kinetic model investigation on coal pyrolysis.

Figure 1.

TGA curves of Zhundong coal samples during pyrolysis at different heating rates taken from literature [9].

β(K∙min−1)T0(K)Tp(K)R(%∙min−1)Vf(%)
10425.39718.950.9035.8
20458.89731.251.7536.4
30482.89735.252.8538.0
50489.86752.755.3437.9

Table 2.

Pyrolysis characteristic values of coal samples.

3. Theoretical basis for thermal kinetics

For non-isothermal and heterogeneous systems, the thermal analysis kinetics is usually evolved from the theory of isothermal and homogeneous systems. The kinetics of reactions in isothermal and homogeneous systems for solids are generally described by Eq. (1)

dV/Vdt=kfVVE1

where V is total mass of volatiles released at time ‘t’,Vis the total mass of volatiles originally available for the reaction, fVVis the reaction model mechanism function, and kis the kinetic rate constant. The reaction rate constant varies with temperature and is assumed to follow the Arrhenius law [16].

k=k0eERTE2

In the above formula, k0refers to pre-exponential factor, Eis the activation energy,Ris universal gas constant, and Tis the thermodynamic temperature. This equation is basically applicable to most elementary reactions and complex reactions, where k0and Eare two important parameters that need to be combined with experimental and theoretical models. If the TGA experiment is conducted at a linear heating rate of β(Kmin1), the temperature and time coordinates will be related to each other as:

T=T0+βtE3

Thus,β=dT/dt. Substituting Eq. (2) and (3) into Eq. (1) gives,

βdV/dVdT=k0eERTfVVE4

Assuming α=VV, then Eq. (4) becomes,

βdT=k0eERTfαE5

For different reaction models, there are differences in the expression of fα. One of the most commonly used reaction models is the reaction order model, which can be expressed as,

fα=1αnE6

The core issue of the general reaction kinetics of coal pyrolysis is determining the three factors k0, fα, and E[17], which is used as a key indicator to describe the process of devolatilization. In general, the integral form of the dynamical mechanism function fαis,

gα=0αfαE7

The above is the basic knowledge of thermal analysis dynamics. Next, two methods, that is, the isoconversional model and the distributed activation energy model (DAEM), will be used to align the TGA data of Zhundong coal pyrolysis.

4. Isoconversional methods

4.1. Isoconversional theories

The basic assumption of the isoconversional method is: under different conditions of temperature rise, the activation energy corresponding to the same conversion rate remains the same [18], and the pyrolysis is specified to be a first-order reaction, that is, n = 1. Thus, gαcan be written as,

gα=ln1αE8

The activation energy and pre-exponential factor for each conversion point can be obtained by analyzing several TGA curves at different heating rates for the same value of conversion α. Next, we will focus on the typical isoconversional method for discussion, including a temperature differential method and three temperature integration methods.

4.1.1. Friedman method

Friedman model is a typical representative of the temperature differential method. Taking the logarithm of the Eq. (5) on both sides of the equal sign to obtain,

lnβdT=lnk0fαERTE9

The above formula is the general formula of the Friedman method. Thus, the plot of lnβdTversus 1/Tat the same conversion levels should be a straight line, the slope and intercept of which can be employed to specify the values of Eand k0, respectively.

4.1.2. Flynn-Wall-Ozawa method

Combining Eq. (5), Eq. (8) can be rewritten as

gα=0αfα=T0Tk0βeERTdT0Tk0βeERTdT=k0EβRpuE10

where u=E/RTand pucan be simplified by using integral approximation,

pu=ueuu2du=euu2uueudu2=euu2u2u3deu
=euu22euu3+u6u4deu
=euu22euu3+6euu4uueud6u4=
=euu212!u+3!u24!u3+E11

Taking the first two items in parentheses on the right side of Eq. (11), we can get the approximate expression of pu,

pu=euu2u3E12

Performing logarithmic operation on both sides of the Eq. (12),

lnpu=u+lnu23lnuE13

For a given solid fuel, in the temperature range of the thermal reaction stage, the value of uis generally ranging from 20 to 60 [19], which is, 20 ≤ u ≤ 60, then −1 ≤ (u − 40)/20 ≤ 1, take s=u40/20, thus it gives,

u=20s+40E14

Substituting Eq. (14) into Eq. (13), it gives,

lnpu=u3ln40+ln38+ln1+10s193ln1+s2E15

Using the Taylor series expansion for the two logarithmic expressions to the right of Eq.(15) and taking a first-order approximate expression of each logarithmic, the following expression can be obtained:

lnpu5.33081.0516uE16
pu=0.0048e1.0516uE17

Substituting Eq. (17) into Eq. (10) yields,

gαk0EβR0.0048e1.0516ERTE18

Eq. (18) can be rewritten as,

lnβ=lnk0ERgα5.3311.052ERTE19

Eq. (19) is the integral form of the FWO model. The plot of lnβversus 1/Tat the same αvalue for different heating rates gives the Eand k0value corresponding to the selected α.

4.1.3. Kissinger-Akahira-Sunose method

Taking the first item in parentheses on the right side of Eq. (11), the approximate expression of pucan be rewritten as,

pu=eu/u2E20

Substituting Eq. (20) into Eq. (10) yields,

gα=k0Eeu/βRu2E21

Performing Logarithmic operations on the two sides of Eq. (21) and then substituting u=E/RTinto the formula,

lnβT2=lnk0REgαERTE22

Eq. (22) is the holonomic form of the KAS model. Using Eq. (22), we can estimate both Eand k0from the Arrhenius plot of lnβT2versus 1/T at the selected αvalue and βvalues.

4.1.4. Miura-Maki method

In the Miura-Maki integral method, the approximation of the temperature integral is the same as the KAS method. Rearranging Eq. (22) and inserting Eq. (8) gives,

1αexpk0RT2βEexpERT=EsTE23

To derive the Miura-Maki integral method, the assumption of distributable activation energy model is used to approximate the right side of Eq. (23) as a step function at E=Esfor a selected temperature. Furthermore, the activation energy Escan be chosen to meet EsT=0.58[12]. Substituting Eq. (8) into Eq. (22) yields

lnβT2=lnk0RElnln1αERTE24

Inserting the value of EsTinto Eq. (24) gives

lnβT2=lnk0RE+0.6075ERTE25

Eq. (25) is the final expression of Miura-Maki method. The linear expression used to calculate the activation energy is similar to the KAS method, but there are some differences in the calculation results of the pre-exponential factor.

Generally speaking, the core methods of the above four models for calculating activation energy and pre-exponential factor are the same. The procedure to estimate fEand k0using the four methods is as follows:

  1. Measure αversus Trelationships at least three different heating rates;

  2. Calculate the values of lnβdT, for Friedman model, lnβ, for FWO model, and lnβT2, for KAS/Miura-Maki model, at selected αvalues from the αversus Trelationships obtained for different heating rates, respectively.

  3. Plot lnβdTversus 1/T, for Friedman model,lnβversus 1/T, for FWO model,and lnβT2versus 1/T, for KAS/Miura-Maki model, at the selected αand then determine the activation energies Efrom the slopes and k0from the intercept, respectively.

The Friedman model is a representative of the differential method in the isoconversional method. The FWO model, KAS model, and Miura-Maki method are three typical integration methods. The integral equations all involve a general temperature integral, 0Tk0βeERTdT, which cannot be solved exactly, and different approximation methods have been proposed for this integral [13]. The KAS method and the Miura-Maki method express the above integral expression by a step-by-step integration formula, as shown in Eq. (11), and adopt the first item of the integration formula, which produces a lower accuracy. The FWO model performs two terms of the step-integration method, which improves the prediction accuracy. Nevertheless, Taylor approximation for the logarithm gives some errors. Two approximations may cause the FWO method to have some deviation from the experimental value. The Friedman method avoids the problem of integral approximation, but the numerical differential calculation will amplify the experimental noise. Unlike the three methods, the Miura-Maki method uses not only the temperature integral approximation, the same as KAS method, but also the assumption in the DAEM, as discussed below. It is obvious that all the isoconversional methods will cause some inevitable errors, so it is necessary to use a method with a good linear correlation to describe the TGA data.

4.2. Processing thermogravimetric analysis data by isoconversional methods

In order to obtain the key parameters of the Friedman, FWO, KAS, and Miura-Maki models, the value of αis chosen to ranges from 0.05 to 0.95 with an interval of 0.05 in this paper [10]. The activation energy and pre-exponential factor at selected αvalues can be obtained by the slope and intercept of the linear fit, respectively, as shown in Figures 46 in literature [9]. The activation energy and pre-exponential factor obtained with the isoconversional and Miura-Maki methods with respect to the conversion extent αhave been shown in Figure 2. It can be noted that activation energies obtained from the four methods were almost the same, and they all increase continuously with increased conversion rates. This is because the refractory molecules in coal must require more energy to be released in the form of volatiles. Note also that as the reaction proceeds, the logarithmic value of the pre-exponential factor first increases and then levels off. This indicates that as the temperature rises, the frequency of molecular collisions in the coal rises and the reaction tends to be strenuous; when the reaction proceeds to a certain degree, that is, E=170kJmol1, most of the molecules are in an activated state, and the reaction rate tends to be level, that is, pre-exponential factors fluctuate within an order of magnitude. The relationship between lnk0and Efor Zhundong coal are also shown in Figure 2. It shows that before the activation energy was 170 kJmol1, lnk0and Eshowed a clear linear relationship, the so-called “kinetic compensation effect” [20], which would bring about some unavoidable errors to the solution of the activation energy and the pre-exponential factor of the reaction. When the activation energy is greater than 170 kJmol1, the value of the pre-exponential factor fluctuates above and below lnk0=23. The fixed value of k0obtained here is to be used as a constant value of the k0in the distributed activation energy model. The range of kinetic parameters obtained for the four models is shown in Table 3. Since the three integration methods use certain approximation methods in the derivation process, and the differential method will amplify the experimental noise, a method with better linear correlation is used here to discuss the distribution of activation energy in the Zhundong coal. The linear correlation coefficients of the four models are listed in Table 4. It can be seen that when the value of αreaches 0.75 or more, the linear correlation coefficient square R2of the four methods shows a downward trend. For the FWO method, the linear correlation coefficient is kept above 0.92 when αis between 0.15 and 0.95, and the value of R2is higher than the other three methods. It shows that the FWO method describes the experimental data more accurately, and the obtained kinetic parameters are more reliable over a wide range of temperatures. Figure 3 was obtained by differentiating the αversus E (calculated by FWO method) relationship. It can be seen that the fEshows a unimodal distribution, with the peak at 173 kJmol1and the activation energy mainly in the range of 160–180 kJmol1. The calculated activation energy distribution can be well fitted using the Gaussian function, where the correlation coefficient is 0.96. Comparison between the experimental and the simulated data by the FWO method (the best fit among the three methods) is shown in Figure 4. The error between the simulated values and the experimental results of αis less than 12.3%, so FWO model achieves a good agreement with the experimental data over the whole range of conversion. But it needs more than three TGA curves to obtain the distributions of kinetic parameters, which will produce some inconvenience. Jain et al. [23] used Indian coal as an experimental sample and found that the results obtained by using the isoconversional methods at higher temperature had poor linear correlation. Therefore, the activation energy calculated by the isoconversional methods can only reflect the condition within a certain temperature range. However, the kinetic parameters obtained from isoconversional methods can provide inputs for selecting appropriate distributions for the DAEM.

Figure 2.

Comparison of activation energy and pre-exponential factor obtained with four different methods for Zhundong coal samples.

MethodEmin(kJ·mol1)Emax(kJ·mol1)kmin(s1)kmax(s1)
FWO55237120.161.58E + 10
KAS48.5231.69.461E + 10
Miura-Maki48.5231.6100.483.82 E + 9
Friedman69.7247528.755.07E + 9

Table 3.

Summary of the kinetic parameters obtained from different methods.

R2αmethods0.150.250.350.450.550.650.750.850.95
FWO0.9990.9980.9980.9990.9990.9980.9900.9730.927
KAS/Miura-Maki0.9980.9960.9970.9990.9980.9980.9880.9690.916
Friedman0.9990.9960.9970.9980.9970.9930.9730.9740.881

Table 4.

Linearly dependent coefficients for different methods at various αvalues.

Figure 3.

Pyrolysis activation energy distributions f E based on FWO model.

Figure 4.

Comparison of calculation results from FWO method ( α cal ) and experimental data α exp of conversion.

5. Distributed activation energy model

5.1. Distributed activation energy model theory

The DAEM assumes that the original materials of coal contain plenty of constituents, which are numbered i = 1, 2…j. These components have different reaction activation energies [21]. Only when the reaction temperature reaches a certain value will they be released in the form of a volatile fraction. These reactions for diverse components are expressed as n-order reactions and are all irreversible and independent of the reactions in other components. The instantaneous volatilized mass fraction for the ithconstituent is denoted by Vit,and the total released mass fraction for the ithconstituent is Vit. According to Eq. (4) and Eq. (6), the reaction rate of the mass fraction of releasing volatiles can be expressed as,

βdVi/dVidT=k0eEiRT1ViVinE26

To some solid state reactions, the pre-exponential factor is connected with the temperature through the following relationship,

k0=k0,TγE27

Combining Eq. (27), the integrating solution to Eq. (26) may be rewritten as,

ViVi=111nk0,β0TTxeEi/RTdT1/1n,n1ViVi=1expk0,β0TTxeEi/RTdT,n=1E28

Considering the continuous reactants have different activation energies, the DAEM devotes to describe the energy distribution of the infinite complicated set of reactions, that is j, as a function fE. Thus, the amount of volatiles loss that has an activation energy between Eand E+dEcan be expressed as,

dV=VfEdEE29

The integration of Eq. (29) brings about,

0fEdE=1E30

The value of dVis actually the same concept as Vi, which accounts for the total released mass fraction for the ithconstituent which has an activation energy between Eand E+dE. Let replace Viand Viwith dVand dV,respectively. Introducing Eq. (29) to Eq. (28), Eq. (28) has following expression.

dV=V111nk0,β0TTxeE/RTdT1/1nfEdE,n1dV=V1expk0,β0TTxeE/RTdTfEdE,n=1E31

By integrating Eq. (31), the following expression can be obtained,

10VdVV=10111nk0,β0TTxeE/RTdT1/1nfEdE,n110VdVV=101expk0,β0TTxeE/RTdTfEdE,n=1E32

Substituting Eq. (30) into Eq. (32), the equation yields,

1α=011nk0,β0TTxeE/RTdT1/1nfEdE,n11α=0expk0,β0TTxeE/RTdTfEdE,n=1E33

A double model [15] was developed assuming that the pyrolysis process occurs in two steps with different kinetic behaviors. The pyrolysis process is divided into two steps: the tar and light hydrocarbon gas formation during the primary pyrolysis and the char condensation, cross-linking reactions, and a further gas production during the secondary pyrolysis. Two sets of parallel reactions occur, sharing the same pre-exponential factor but not the same distributed activation energy. The distribution function of activation of energy fEin 2-DAEM equation can be written as,

fE=ωf1E+1ωf2EE34

where ωis a parameter that weighs the different reaction classes, varying from 0 to 1. This parameter describes how many volatiles are released during the various reactions.fεE, where εdenotes the stage number, is a Gaussian or logistic function of the form with a mean activation energy E0and a standard deviation σ,

fE=1σ2πeEE0/2σ2fE=π3σeπEE03σ1+eπEE03σ2E35

In the 1-DAEM case, there are four parameters to be estimated: the mean activation energy E0, the standard deviation σ, the temperature exponent x, and the chemical reaction order n. In the 2-DAEM case, there are seven parameters to be estimated: two mean activation energies E01and E02, two standard deviations σ1and σ2, the temperature exponent γ, the chemical reaction order, and ω.

In this work, the model-fitting exercise was carried out using MATLAB TM (R2014b). It is essential to use an appropriate method to compute the inner dTintegral of Eq. (33). An approximation proposed by Cai and Liu [21] was validated by literature [24], and was applied to estimate the general temperature integral,

0TTγeERTdT=RTγ+2EeERT0.99954E+0.580580.044967γRTE+2.54+0.94057γRTE36

To deal with the outer dEintegral of Eq. (33), Simpson’s 1/3 rule [22] was employed. In the numerical integration of Eq. (33), E0+3σvalue has been used for the upper limit of the outer dEintegral. Therefore, for given values of the parameters, the DAEM equation can be numerically solved.

The kinetic parameters of the DAEM are determined using an objective function. An effective criterion for achieving the objective function is to minimize the differences of squares. The functional form of the objective function used in this paper is,

OF=xlαx,calαx,expi2E37

where αx,caland αx,expare the calculated and experimental values of the residual mass fraction, lis data number, and the sub-index xmakes reference to data point x. In this paper, the objective function (OF) has been optimized using the pattern search method [23], which is a class of direct search methods. The Gaussian and logistic DAEMs show an interrelationship between activation energies and the pre-exponential factor. Thus, E0obtained for DAEM differed significantly over values of k0, and it was difficult to figure out their relationship. To remove this ambiguity, the k0was set to be a fixed value according to the results from isoconversional methods. The initial value of the mean activation energy E0and standard deviation σsettled in pattern search method used the results from isoconversional methods.

5.2. Processing thermogravimetric analysis data by distributed activation energy models

In the distributed activation energy models (DAEMS) adopted here, the pre-exponential factor k0is shared by all the reactions. From the literature [20], it is known that there exists an interrelation between the pre-exponential factor k0and the mean activation energy E0, which means that several pairs of k0and E0values can fit equally well for a given set of experimental data. Thereby, the value of k0has to be fixed to remove the mutual influence between k0and E0. Nevertheless, when the value of k0is selected within two orders of magnitude, the observed correlation between k0and activation energy does not adversely influence the utility of kinetic models [24]. In this paper, the value of k0is set to be 1.0E+10S1, which is based on the results from the isoconversional methods discussed above. The initial guess value of E0S1used in the pattern search method of DAEMs is also based on the results from isoconversional methods of FWO model, as shown in Figure 3, which is corresponding to the peak activation energy and can be set to be 173kJmol1. The expected initial value of σwas set to be 20kJmol1, which is in accordance with the peak width of the fEobtained from isoconversional methods. The value of the objective equation (OF) can be used as a benchmark for evaluating the prediction results obtained by different DAEMs.

The DAEMs were then applied to fit with the experimental data of β=20Kmin1. Figure 5 shows the comparison of the various 1-DAE models. The kinetic parameters of these models and the parameters achieved with these models are listed in Table 5. It can be noticed from Figure 5 that 1-DAE models with Gaussian/logistic distribution do not show good match with experimental values, which suggests that it is not appropriate to describe the pyrolysis process based on the single activation energy distribution for Zhundong coal. 2-DAEMs consider that pyrolysis occurs in two steps assuming that the first and the second stage share the same activation energy. The comparison of the various 2-DAE models with experimental results for Zhundong coal at a heating rate of β=20Kmin1are shown in Figure 6, and the kinetic parameters extracted from these models are listed in Table 6. Looking at the value of OFlisted in Tables 5 and 6, it can be concluded that 2-nth-DAEM assuming the Gaussian distribution gives best fit with the experimental data. From the values extracted with the 2-DAE models assuming the Gaussian distribution, as shown in Table 6, it can be seen that the mean activation energy for the primary pyrolysis (where breakage of weaker bonds takes place) is lower than the mean activation energy for the secondary pyrolysis (where breakage of tougher bonds takes place). However, the difference in activation energy between the two stages is not very large. This behavior is more visible with a view to the plots showed in Figure 7 where the fEfunction versus activation energy for 2-nth-DAEM with Gaussian distribution is exhibited. It can be concluded that the amount of volatile production attributed to primary pyrolysis stage is a little more than the secondary stage for the Zhundong coal. The original 2-DAEM put forward by Caprariis et al. [15] considers coal pyrolysis as a first order reaction, that is, n = 1, and the pre-exponential factor is a constant. This work takes all these factors into account and will be more conducive to better predictions.

Figure 5.

Comparison of experimental data and simulated data with the 1-DAEM methods: (a) Gaussian distribution; (b) logistic distribution.

E0kJmol1σkJmol1γnOF
n=1Gaussian169.4224.45−0.44410.0493
Logistic175.1179.84−0.40010.2148
n1Gaussian172.9618.730.1275.7660.0058
Logistic172.9619.350.1325.7860.0060

Table 5.

Estimated values of kinetic parameters [1-DAEMs with Gaussian/logistic distribution functions].

Figure 6.

Comparison of experimental data and simulated data with the 2-DAEM methods: (a) Gaussian distribution; (b) logistic distribution.

ModelE01kJmol1σ1kJmol1E02kJmol1σ2kJmol1γnωOF
n=1Gaussian150.0020.00173.0036.85−0.50010.4000.0492
Logistic151.0241.89171.0825.06−0.51410.3040.0160
n1Gaussian150.0028.22173.0026.85−0.5022.8720.6460.0001
Logistic150.0029.57181.8334.50−0.5132.8260.7650.0005

Table 6.

Estimated values of kinetic parameters [2-DAEMs with Gaussian/logistic distribution functions].

Figure 7.

Distribution activation energy curves as a function of activation energy for 2-nth-DAEM with Gaussian distribution functions.

The kinetic parameters extracted with 2-nth-DAEM with Gaussian distribution functions (best models to describe data at 20 Kmin1for Zhundong) were used to predict the experimental data obtained at different heating rates of 10Kmin1, 30Kmin1,and 50Kmin1, respectively. The results obtained with 2-nth-DAEM are shown in Figure 8. It can be seen that the 2-nth-DAEM with Gaussian distribution shows good match with other three heating rate curves.

Figure 8.

Comparison of predicted results with the experimental data at other heating rates with 2-nth-DAEM with Gaussian distribution functions: (a) β = 10 K ∙ min − 1 ; (b) β = 30 K ∙ min − 1 ; (c) β = 50 K ∙ min − 1 .

6. Conclusions

As a raw material for the preparation of phase change materials, coal is of great significance to energy utilization. In this work, Zhundong coal was taken as an example to investigate the pyrolysis process of it by means of thermogravimetric tests. In order to evolve appropriate kinetic parameters to characterize the pyrolysis process of coal, various methods and potential assumptions for processing TGA data were rigorously analyzed. The four isoconversional methods, that is, Friedman, FWO, KAS, and Miura-Maki method were used to calculate the activation energy and pre-exponential factor of Zhundong coal samples, respectively. It was found that the FWO model gave the best fit for the experimental data. During the pyrolysis process, the distribution of the activation energy of Zhundong coal, fE, shows a unimodal distribution, with a peak at the activation energy of 173 kJmol1.In addition, Several DAEMs have been theoretically derived, considering the reaction order, the dependence of the pre-exponential factor on temperature, the steps mechanism for the pyrolysis, and the form of activation distribution, wherein the 2-nth-DAEM assuming Gaussians distribution of activation energy gave the best fit for the experimental data of Zhundong coal.

Acknowledgments

This work is currently supported by the National Natural Science Foundation of China through contract No.51606153 and China Postdoctoral Science Foundation (2017M3200).

Eactivation energy (kJ·mol−1)
E0mean activation energy (kJ·mol−1)
fEdistribution curve of the activation energy E
ithe number of constituents in original materials
krate coefficient s−1
k0pre-exponential factor s−1
nThe reaction order
Runiversal gas constant (kJ·mol−1·K−1)
R∞maximum weight loss rate (%·min−1)
Ttemperature (K)
T0initial decomposition temperature (°C)
Tppeak temperature (°C)
Vtotal mass of volatiles released at time ‘t’ (%)
Vfthe corresponding weight loss percentage when the pyrolysis reaches the final temperature (%)
V∗total amount of volatile yields (%)
αConversion
βheating rate (K∙min−1)
γConstant
ωweighing factor
σstandard deviation (kJ·mol−1)
Subscript
calcalculated value
expexperimental value

© 2018 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution 3.0 License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

How to cite and reference

Link to this chapter Copy to clipboard

Cite this chapter Copy to clipboard

Guodong Jiang and Liping Wei (November 5th 2018). Analysis of Pyrolysis Kinetic Model for Processing of Thermogravimetric Analysis Data, Phase Change Materials and Their Applications, Mohsen Mhadhbi, IntechOpen, DOI: 10.5772/intechopen.79226. Available from:

chapter statistics

323total chapter downloads

More statistics for editors and authors

Login to your personal dashboard for more detailed statistics on your publications.

Access personal reporting

Related Content

This Book

Next chapter

Introductory Chapter: Phase Change Material

By Mohsen Mhadhbi

Related Book

First chapter

The Effects of Novel Additives Used in PVA/Starch Biohybrid Films

By Eyyup Karaogul, Ertugrul Altuntas, Tufan Salan and Mehmet Hakki Alma

We are IntechOpen, the world's leading publisher of Open Access books. Built by scientists, for scientists. Our readership spans scientists, professors, researchers, librarians, and students, as well as business professionals. We share our knowledge and peer-reveiwed research papers with libraries, scientific and engineering societies, and also work with corporate R&D departments and government entities.

More About Us