Professional Documents
Culture Documents
DOI 10.1007/s10973-014-4215-3
Introduction
The pyrolysis kinetics of lignocellulosic biomass is usually
performed by means of thermogravimetric (TG) analysis
[1]. The pyrolysis of lignocellulosic biomass usually
involves several processes, which would result in some
mass loss steps observed in a TG curve or some peaks
observed in the corresponding differential thermogravimetric (DTG) curve [2]. Unfortunately, a very common
situation is that different processes of decomposition
overlap each other [1]. In general, those overlapped peaks
cannot be separated directly from the DTG curve. To solve
this problem, some researchers proposed the Logistic mixture model [35] and Weibull mixture model [611] to
separate peaks in the DTG curves. Recently, Perejon et al.
[12] found that the FraserSuzuki function can be used to
deconvolute a complex solid-state reaction into its individual processes. Findorakova et al. [13] successfully
deconvoluted the conversion rate curves of the thermal
decomposition of [Zn(2-Clbenz)2(caf)2]. Svoboda et al. [14]
investigated the applicability of the FraserSuzuki function
in the deconvolution of complex crystallization processes
and found that the deconvolution procedure could provide
reasonable, consistent, and accurate results. They also
obtained the correlation between the JohnsonMehlAvrami (JMA) model and parameters of the FraserSuzuki
function. Koga et al. [15] successfully used nine empirical
mathematical functions (including the FraserSuzuki
function, Weibull mixture model, and Logistic mixture
model) for the deconvolution of the DTG curves of the
thermal decomposition of co-precipitated zinc carbonates.
The distributed activation energy model (DAEM) has
proven to describe complex thermally activated solid-state
reactions very well [16], in particular, the pyrolysis of
lignocellulosic biomass and its major components [17, 18].
123
Z. Cheng et al.
Table 1 Ultimate analysis results of lignocellulosic biomass samplesa
Sample
C/%
Cotton stalk
N/%
S/%
O /%
Ash/
%
43.750 6.290
0.536
0.260
47.007
2.157
Oilseed rape
straw
41.390 5.430
0.254
0.429
48.134
4.363
Rice straw
37.895 5.781
0.585
0.408
49.785
5.546
Dry basis
By difference
H/%
Cellulose/%
Hemicellulose/%
Lignin/%
Cotton stalk
34.54
12.30
42.08
47.16
24.31
22.52
Rice straw
42.29
27.43
16.00
Dry basis
123
FraserSuzuki function
The FraserSuzuki function, an exponential function, was
originally developed by Fraser and Suzuki for the resolution of overlapping bands [21, 22]. This function was
usually used for signal processing and data analysis in
chromatography [23, 24]. In this work, the FraserSuzuki
function was used to describe the DAEM processes and
lignocellulosic biomass pyrolysis processes. The Fraser
Suzuki function can be expressed as
(
)
da
ln 2
T p 2
h exp 2 ln 1 2s
;
1
dT
s
w
where a is the conversion degree, T is the temperature, da/
dT is the conversion rate, and h, s, p, and w are the
parameters.
In order to better understand the physical meanings of h,
s, p, and w, a parametric study of the FraserSuzuki
function was carried out. Figure 1 shows the influences of
the parameters on the function curve. From Fig. 1, it can be
obtained that the parameter p is the position parameter, the
parameter w is related to the function curve width, the
parameter h is the height parameter, and the parameter s is
the shape parameter. The FraserSuzuki function curve is
left-skewed if s \ 0 and is right-skewed if s [ 0.
For fitting the data of the DAEM processes and lignocellulosic biomass pyrolysis processes to the FraserSuzuki
function, some estimation of the function parameters is
required. The nonlinear optimization method was
employed to perform this task. The minimum of the
residual sum of squares was sought in order to obtain the
best fit. In this work, DataFit, a professional curve fitting
(nonlinear regression) and data plotting software developed
by Oakdale Engineering [25], was used for the implementation of the nonlinear optimization.
8
1=1n
RT A
R1
E
>
>
< 1 0 1 1 n T0 exp
f EdE
dT
b
RT
a
RT A
R1
>
E
>
:
1 0 exp T0 exp
dT f EdE
b
RT
n 6 1
2
n1
8
n=1n
Z T
>
R
A
E
>
> 1 A exp E
exp
f EdE
1
1
n
dT
<
0 b
da
RT
RT
T0 b
Z T
R1 A
dT >
E
A
E
>
>
exp
exp
dT f EdE
:
0 b
RT
RT
T0 b
d /dT/K1
0.012
0.009
0.006
0.010
h = 0.005 K1
h = 0.0075 K1
h = 0.01 K1
0.008
h = 0.0125 K1
h = 0.015 K1
p = 600 K
w = 40 K
s = 0.3
n1
0.006
0.004
h = 0.01 K1
p = 550 K
p = 575 K
p = 600 K
p = 625 K
p = 650 K
w = 40 K
s = 0.3
0.002
0.003
0.000
0.000
450
d /dT/K1
0.015
n 6 1
500
550
600
650
700
450
500
550
T/K
d /dT/K1
0.008
0.006
0.004
0.010
h = 0.01 K1
p = 600 K
w = 20 K
w = 30 K
w = 40 K
w = 50 K
w = 60 K
s = 0.3
0.008
d /dT/K1
0.010
0.006
0.004
0.002
0.002
0.000
0.000
450
500
550
600
650
700
T/K
600
T/K
650
700
450
h = 0.01 K1
p = 600 K
w = 40 K
s = 0.5
s = 0.3
s = 0.1
s = 0.1
s = 0.3
s = 0.5
500
550
600
650
700
T/K
123
Z. Cheng et al.
Table 3 Parameter values of fitted FraserSuzuki functions for simulated DAEM processes together with corresponding correlation coefficients
(Example 1)
DAEM process
h/K-1
13.64
R2
-1
-1
-1
-1
p/K
-2
1.799 9 10
w/K
627.52
-1
-1.853 9 10
51.23
0.9998
1.652 9 10-2
628.26
-1.211 9 10-1
56.04
0.9997
1.522 9 10-2
628.68
-5.285 9 10-2
60.83
0.9994
1.409 9 10-2
628.89
1.379 9 10-2
65.44
0.9990
1.311 9 10-2
628.98
7.691 9 10-2
69.83
0.9984
1.226 9 10-2
628.99
1.360 9 10-1
73.99
0.9970
Table 4 Parameter values of fitted FraserSuzuki functions for simulated DAEM processes together with corresponding correlation coefficients
(Example 2)
DAEM process
R2
p/K
w/K
1.064 9 10-2
770.02
-4.911 9 10-2
86.93
0.9991
1.209 9 10-2
716.55
-5.099 9 10-2
76.49
0.9991
1.362 9 10-2
669.83
-5.220 9 10-2
67.94
0.9992
1.522 9 10-2
628.68
-5.283 9 10-2
60.83
0.9993
1.689 9 10-2
592.17
-5.302 9 10-2
54.86
0.9993
1.861 9 10-2
559.58
-5.282 9 10-2
49.81
0.9994
Table 5 Parameter values of fitted FraserSuzuki functions for simulated DAEM processes together with corresponding correlation coefficients
(Example 3)
DAEM process
h/K-1
13.64
-1
R2
-1
-1
p/K
-2
1.586 9 10
w/K
595.57
-2
-4.957 9 10
58.44
0.9994
1.522 9 10-2
628.68
-5.285 9 10-2
60.83
0.9993
1.464 9 10-2
661.74
-5.568 9 10-2
63.23
0.9992
1.409 9 10-2
694.75
-5.851 9 10-2
65.66
0.9992
123
h/K-1
13.64 -1
-1
-1
-1
p/K
-2
1.677 9 10
1.522 9 10-2
A = 10
R2
w/K
629.43
-2
-8.393 9 10
54.86
0.9986
628.68
-5.283 9 10-2
60.83
0.9993
1.333 9 10-2
627.79
-2.341 9 10-2
69.90
0.9997
1.156 9 10-2
627.02
-3.452 9 10-3
80.86
0.9999
0.020
Simulated data
F-S predicted curve
d /dT/K1
d /dT/K1
n = 1.0
0.015
0.010
0.010
0.005
0.005
0.000
Residuals/K1
Residuals/K1
0.015 n = 1.25
Simulated data
F-S predicted curve
0.0005
0.0000
0.0005
500
550
600
650
700
0.000
0.0005
0.0000
0.0005
500
550
600
T/K
n = 1.5
Simulated data
F-S predicted curve
0.010
750
n = 1.75
Simulated data
F-S predicted curve
0.010
0.005
0.000
0.0005
0.0000
0.0005
500
550
600
650
700
750
800
Residuals/K1
Residuals/K1
0.005
0.000
0.0005
0.0000
0.0005
500
550
600
650
T/K
700
750
800
T/K
0.015
0.0125
n = 2.0
Simulated data
F-S predicted curve
0.010
0.005
n = 2.25
Simulated data
F-S predicted curve
0.0100
d /dT/K1
d /dT/K1
700
0.015
d /dT/K1
d /dT/K1
0.015
650
T/K
0.0075
0.0050
0.000
0.0005
0.0000
0.0005
500
550
600
650
700
750
800
T/K
Residuals/K1
Residuals/K1
0.0025
0.0000
0.0005
0.0000
0.0005
500
550
600
650
700
750
800
850
T/K
(solid lines). The residuals between the simulated data and the curve
predicted by the FraserSuzuki function are plotted underneath the
figures. (Example 1)
123
Z. Cheng et al.
0.0125
0.0125
A = 1010.64 s 1
0.0100
0.0075
d /dT/K1
Simulated data
F-S predicted curve
0.0050
0.0050
0.0025
0.0000
0.0000
0.0005
0.0000
0.0005
600
Simulated data
F-S predicted curve
0.0075
0.0025
Residuals/K1
Residuals/K1
d /dT/K1
0.0100
A = 1011.64 s 1
0.0005
0.0000
0.0005
650
700
750
800
850
900
600
650
700
T/K
A = 1013.64 s 1
A = 1012.64 s 1
0.0150
Simulated data
F-S predicted curve
d /dT/K1
0.0125
0.0100
0.0075
0.0075
0.0050
0.0050
0.0025
0.0025
0.0000
0.0000
Residuals/K1
d /dT/K1
Residuals/K1
Simulated data
F-S predicted curve
0.0100
0.0005
0.0005
0.0000
0.0000
0.0005
0.0005
550
600
650
700
750
500
550
600
T/K
0.0200
A = 1014.64 s 1
700
750
0.0100
0.0075
0.0050
Simulated data
F-S predicted curve
0.0150
d /dT/K1
Simulated data
F-S predicted curve
0.0125
A = 1015.64 s 1
0.0175
0.0150
0.0125
0.0100
0.0075
0.0050
0.0025
0.0025
Residuals/K1
0.0000
0.0005
0.0000
0.0005
500
650
T/K
0.0175
d /dT/K1
800
0.0175
0.0150
0.0125
Residuals/K1
750
T/K
0.0000
0.0005
0.0000
0.0005
525
550
575
600
625
650
675
700
T/K
475
500
525
550
575
600
625
650
T/K
123
0.0125
0.0100
d /dT/K1
E0 = 175.4 kJ mol1
Simulated data
F-S predicted curve
0.0075
0.0125
0.0100
0.0050
0.0025
0.0025
0.0000
0.0000
0.0005
0.0000
0.0005
500
550
600
650
700
Simulated data
F-S predicted curve
0.0075
0.0050
Residuals/K1
Residuals/K1
d /dT/K1
0.0150
0.0005
0.0000
0.0005
500
550
600
T/K
E0 = 195.4 kJ mol1
d /dT/K1
0.0125
Simulated data
F-S predicted curve
0.0100
0.0075
0.0100
750
0.0050
0.0025
0.0025
0.0000
0.0000
0.0005
0.0000
600
650
700
750
800
T/K
Simulated data
F-S predicted curve
0.0075
0.0050
Residuals/K1
d /dT/K1
0.0125
Residuals/K1
700
1
0.0150 E0 = 205.4 kJ mol
0.0150
0.0005
550
650
T/K
0.0005
0.0000
0.0005
600
650
700
750
800
T/K
Fig. 4 Simulated
kinetic
curves
(dots,
A = 1013.64 s-1,
r = 3.5 kJ mol-1, b = 20 K min-1, n = 1.5, and E0 = 175.4,
185.4, 195.4, and 205.4 kJ mol-1) fitted with the FraserSuzuki
hi exp 2 ln 1 2si
; 5
dT
wi
si
i1
123
Z. Cheng et al.
0.0175
0.0150
0.0125
0.0125
d /dT/K1
d /dT/K1
0.0150
0.0100
0.0075
Simulated data
F-S predicted curve
0.0000
0.0000
Residuals/K1
Residuals/K1
= 3.5 kJ mol1
0.0025
Simulated data
F-S predicted curve
0.0025
0.0075
0.0050
= 1.5 kJ mol1
0.0050
0.0100
0.0005
0.0000
0.0005
0.0005
0.0000
0.0005
550
600
700
650
550
600
650
T/K
700
T/K
0.0150
0.0100
0.0075
= 5.5 kJ mol1
0.0050
0.0025
0.0050
= 7.5 kJ mol1
Simulated data
F-S predicted curve
0.0000
0.0005
0.0000
0.0005
500
0.0075
0.0025
Simulated data
F-S predicted curve
0.0000
Residuals/K1
d /dT/K1
0.0100
Residuals/K1
d /dT/K1
0.0125
0.0005
0.0000
0.0005
550
600
650
700
750
500
550
600
T/K
650
700
750
T/K
Table 7 Parameter values of fitted mixed FraserSuzuki functions for cotton stalk, oilseed rape straw, and rice straw
FS function
Parameter value
Cotton stalk
Rice straw
h/K-1
2.888 9 10-3
2.478 9 10-3
2.903 9 10-3
p/K
592.79
561.17
581.65
-6.089 9 10-2
3.320 9 10-1
-4.649 9 10-3
w/K
2
-1
83.00
4.266 9 10
3.774 9 10-3
p/K
635.99
608.60
595.73
-6.552 9 10-2
-5.120 9 10-2
-5.592 9 10-2
h/K-1
-3
34.55
-3
43.02
41.75
9.698 9 10-4
5.242 9 10-4
6.888 9 10-4
p/K
639.75
677.13
638.05
1.982 9 10-1
-6.394 9 10-2
-6.784 9 10-3
w/K
243.65
264.01
291.28
123
90.70
3.635 9 10
h/K
w/K
3
82.62
0.006
Experimental data
F-S predicted curve
F-S 1
F-S 2
F-S 3
0.004
0.003
Experimental data
F-S predicted curve
F-S 1
F-S 2
F-S 3
0.006
dw/dT/K1
dw/dT/K1
0.005
R 2 = 0.9989
0.002
0.005
0.004
0.003
R 2 = 0.9995
0.002
0.001
Residual/K1
Residual/K1
0.001
0.000
0.0002
0.0000
0.0002
450
0.000
0.0002
0.0000
0.0002
400
500
550
600
650
700
750
800
850
600
700
800
900
1000
T/K
900
T/K
500
Conclusions
0.006
Experimental data
F-S predicted curve
F-S 1
F-S 2
F-S 3
dw/dT/K1
0.005
0.004
0.003
R 2 = 0.9991
0.002
Residual/K1
0.001
0.000
0.0002
0.0000
0.0002
450
500
550
600
650
700
750
800
850
900
T/K
References
1. White JE, Catallo WJ, Legendre BL. Biomass pyrolysis kinetics:
a comparative critical review with relevant agricultural residue
case studies. J Anal Appl Pyrolsis. 2011;91(1):133.
2. Van de Velden M, Baeyens J, Brems A, Janssens B, Dewil R.
Fundamentals, kinetics and endothermicity of the biomass pyrolysis reaction. Renew Energy. 2010;35(1):23242.
123
Z. Cheng et al.
3. Barbadillo F, Fuentes A, Naya S, Cao R, Mier JL, Artiaga R.
Evaluating the logistic mixture model on real and simulated TG
curves. J Therm Anal Calorim. 2007;87(1):2237.
4. Cao R, Naya S, Artiaga R, Garca A, Varela A. Logistic approach
to polymer degradation in dynamic TGA. Polym Degrad Stab.
2004;85(1):66774.
5. Naya S, Cao R, de Ullibarri IL, Artiaga R, Barbadillo F, Garca
A. Logistic mixture model versus Arrhenius for kinetic study of
material degradation by dynamic thermogravimetric analysis.
J Chemom. 2006;20(34):15863.
6. Cai J, Alimujiang S. Kinetic analysis of wheat straw oxidative
pyrolysis using thermogravimetric analysis: statistical description
and isoconversional kinetic analysis. Ind Eng Chem Res.
2009;48(2):61924.
7. Cai J, Chen SY, Liu RH. Weibull mixture model for isoconversional kinetic analysis of biomass oxidative pyrolysis. J Energy
Inst. 2009;82(4):23841.
8. Cai J, Liu R. Weibull mixture model for modeling nonisothermal
kinetics of thermally stimulated solid-state reactions: application
to simulated and real kinetic conversion data. J Phys Chem B.
2007;111(36):106816.
9. Cai J, Liu R. Application of Weibull 2-mixture model to describe
biomass pyrolysis kinetics. Energy Fuels. 2008;22(1):6758.
10. Kuo-Chao L, Keng-Tung W, Chien-Song C, Wei-The T. A new
study on combustion behavior of pine sawdust characterized by
the Weibull distribution. Chin J Chem Eng. 2009;17(5):8608.
11. Yoshikawa M, Yamada S, Koga N. Phenomenological interpretation of the multistep thermal decomposition of silver carbonate
to form silver metal. J Phys Chem C. 2014;118(15):805970.
12. Perejon A, Sanchez-Jimenez PE, Criado JM, Perez-Maqueda LA.
Kinetic analysis of complex solid-state reactions. A new deconvolution procedure. J Phys Chem B. 2011;115(8):178091.
13. Findorakova L, Svoboda R. Kinetic analysis of the thermal
decomposition of Zn (II) 2-chlorobenzoate complex with caffeine. Thermochim Acta. 2012;543:1137.
14. Svoboda R, Malek J. Applicability of Fraser-Suzuki function in
kinetic analysis of complex crystallization processes. J Therm
Anal Calorim. 2013;111(2):104556.
15. Koga N, Goshi Y, Yamada S, Perez-Maqueda L. Kinetic
approach to partially overlapped thermal decomposition processes. J Therm Anal Calorim. 2013;111(2):146374.
16. Koga N, Goshi Y, Yamada S, Perez-Maqueda LA. Kinetic
approach to partially overlapped thermal decomposition processes. J Therm Anal Calorim. 2013;111(2):146374.
17. Wu W, Cai J, Liu R. Isoconversional kinetic analysis of distributed activation energy model processes for pyrolysis of solid
fuels. Ind Eng Chem Res. 2013;52(40):1437683.
18. Gb Varhegyi, Bz Bobaly. Jakab E, Chen H. Thermogravimetric
study of biomass pyrolysis kinetics. A distributed activation
energy model with prediction tests. Energy Fuels. 2010;25(1):
2432.
123