1-s2.0-s0016236113008387-main (1)
TRANSCRIPT
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 1/12
Computational fluid dynamics modeling of biomass fast pyrolysis
in a fluidized bed reactor, using a comprehensive chemistry scheme
Pelle Mellin ⇑, Efthymios Kantarelis, Weihong Yang
KTH Royal Institute of Technology, Division of Energy and Furnace Technology, Brinellvägen 23, SE 100 44 Stockholm, Sweden
h i g h l i g h t s
We have used an advanced kinetic model for pyrolysis coupled with CFD in 3D. The yields are compared to experimental results and shows a not too far-off prediction.
The simulations are very time consuming but makes it possible to explore secondary reactions.
For example, a number of thermal cracking reactions are applied to the tar components to see the effect.
a r t i c l e i n f o
Article history:
Received 18 June 2013
Received in revised form 6 August 2013
Accepted 5 September 2013
Available online 17 September 2013
Keywords:
CFD
Fluidized bedFast pyrolysis
Pyrolysis oil
Pyrolysis
a b s t r a c t
The CFD modeling for fast pyrolysis has previously focused on the major pyrolysis products; liquid, char
and gas. This paper introduces a new approach to biomass pyrolysis; integrating a complex scheme of
reactions including formation of such components as levoglucosan. The 3-D simulation takes into account
the complex breakdown of each biomass subcomponent, the fluid dynamics of the process as well as the
heat and momentum transfer of three Eulerian phases.
The pyrolysis products include reference species that reflects the composition of the bio oil, gas fraction
and char fraction. A number of reactions are in addition applied to account for the thermal cracking of tar
compounds and the final compositions are compared to experimental yields. The results show that the
predicted pyrolysis products reflect the experimental yields satisfactorily, apart from the water content
which is under predicted. Most importantly though, the approach is computationally feasible and it
should be useful for future work.
2013 Elsevier Ltd. All rights reserved.
1. Introduction
Pyrolysis of biomass is considered as a potential source of fuel
for various applications such as transportation, combined heat
and power production as well as reduction agents. From a practical
standpoint, fast pyrolysis means processing with optimized
conditions for liquid production which implies fast heating rate
and a temperature not exceeding 500 C.Modeling such processes is challenging in many ways and even
more so in fluidized bed reactors. Those challenges are linked to
the complex thermo-physical environment of fluidized bed
reactors and the complex structure of biomass and the decomposi-
tion with an immense number species as products.
Much of the earlier CFD simulation work assumed three
superficial components as representatives for the liquid, gaseous
and solid components. Some studies elaborated on the non-unifor-
mity of a reacting biomass particle and some studies focused more
on the kinetics and the composition of the samples related to
pyrolysis products [1]. No one has yet been able to device a
complete model which takes into account all phenomena of fast
pyrolysis in fluidized beds but both approaches improved the
outcome and lead to increased understanding.
From the CFD direction, Papadikis [2] argues that: ‘‘The most
crucial of the assumptions is that the particle is assumed to
maintain uniform temperature along its radius’’. In contrast,Dupont et al. [1] claims that ‘‘the accurate knowledge of the
reaction kinetics appears to be a crucial parameter for a reliable
modeling of the pyrolysis process’’. This work can be seen as a
unifying effort, where the in-depth chemistry is included in a
comprehensive CFD model which accounts for heat and mass
transfer as well as entrainment of particles.
Previously, a CFD model was developed (described in Mellin
et al. [3]) with a two phase framework for time-effectively
studying the behavior of the gas. Now this model has been
extended in order to report the specific components of pyrolysis
and thus obtain a composition of the liquid, gas and char. Therefore
a new kinetic model has been implemented as well as a third
0016-2361/$ - see front matter 2013 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.fuel.2013.09.009
⇑ Corresponding author. Tel.: +46 87909022; fax: +46 8207681.
E-mail address: [email protected] (P. Mellin).
Fuel 117 (2014) 704–715
Contents lists available at ScienceDirect
Fuel
j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / f u e l
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 2/12
Eulerian phase which in greater detail renders the biomass flow.
Experiments are in addition made alongside the simulation to
compare with the results. More information on the basis of the
numerical model can be found in the previous publication [3], as
well as more details on tuning of the drag law. The far-reaching
purpose of the work is up-scaling of the technology.
In this paper, the pilot reactor is firstly described with analysis
results of the pyrolysis products. Then the numerical model is
described with focus on the additions to the previous work; finally
the results, comparison with experiment, discussion and conclu-
sion is presented. Throughout the work we made extensive use
of User Defined Functions (UDF) and solution methods included
in the commercial software package ANSYS Fluent 14.5.
2. Pilot reactor
A pilot fast pyrolysis setup with has been assembled at KTH,Sweden. The setup includes a preheater for the fluidizing-gas, the
fluidized bed reactor, a cyclone and a scrubber. A schematic
overview of the plant is shown in Fig. 1, with the domain of the
CFD model indicated.
The biomass is introduced by a screw feeder directly into the
fluidized bed at height of 5.75 cm above the distributor plate.
The biomass is fed at rate of 2 kg/h, which is the designed capacity
of the rig. The biomass is a mix of pine and spruce, see proximate
and ultimate analysis in Table 1.
The gas species and the main groups of bio oil components are
given in Table 2 which is reported in Kantarelis et al. [4], therein
referred to as case S/B: 0. All specific components of the bio oil
can be found in the same paper as well as the measurements
methods and more detailed descriptions of the plant.
Nomenclature
Alphabetic letters A pre-exponential factor (s1)a interfacial area concentration (m2 m3)C heat capacity (J kg-1 K-1)c coefficient (–)d diameter (m)E activation energy (J mol1)e coefficient of restitution (–) g gravitational acceleration (m s2)h heat transfer coefficient (W m2 K1)K momentum exchange (kg m3 s1)k heat conductivity (W m1 K1)m mass (kg)Q heat transfer (W)q heat transfer per surface area (W m2)R universal gas constant (J mol1 K1)S source term, due to e.g. reactions ([kg, J. . .]m3 s1)T temperature (K)t time (s)u intrinsic velocity (m s1)v velocity (m s1)Y mass fraction (–)
Greek lettersa volume fraction (–)q density (kg m3)l viscosity (Pa s)p pi (–)s stress (Pa)u sphericity (–)
Dimensionless numbersRe Reynolds number (-)Pr Prandtl number (-)Nu Nusselt number (-)
Common subscripts g gass sandb biomassi any specie
mf minimum fluidizationD drag
Biomass
Windbox
Bubblingfluidized
bedreactor
Char
Fluidizing gaspreheater
Circulation
with cooling
Cyclone
Liquid
Scrubber
Gas
CFDmodel
Fig. 1. Cross section of the bubbling fluidized bed reactor system with boundariesindicated for the CFD model.
Table 1
Biomass composition [4].
Analysis Parameter
Moisture
(wt%)
Volatile
matter
(wt%)db
Ash
(wt%)dbFixed
carbon
(wt%)db
HHV
(MJ/kg)
Proximate
analysis
9.8 83 0.31 16.6 20.46
C, wt%db H, wt%db Oa,
wt%db
N, wt%db S, ppmdb
Ultimate
analysis
50.7 6.1 42.71 0.18 <120
a
By difference.db Dry basis.
P. Mellin et al. / Fuel 117 (2014) 704–715 705
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 3/12
3. Modeling approach
The Euler–Euler multiphase framework with three phases is ap-
plied in the computation; one gas phase and two granular phases:
sand and biomass. Several CFD studies treated the biomass parti-
cles as dilute in the sand and vapor phase, which is a reasonable
assumption. The biomass particles are then tracked in a lagrangian
manner and treated as an additional layer to the two-phase
framework.
When such an arrangement is made the phase coupling be-
tween the two Eulerian phases and the lagrangian one are often
simplified. Usually the drag of the two Eulerian phases correctly
influences the momentum of the lagrangian phase; while not the
other way around. Often the conservation equations for the Euleri-an phases do not contain a term for drag caused by the lagrangian
phase.
In this case where the amount of biomass particles can have
non-negligible effect on the other phases in the system, the
three-Eulerian system was chosen. In addition the Eulerian defini-
tion of a phase in Fluent, allow the use of a stiff chemistry solver
and chain reactions for solid species.
The model with phases and interactions are shown in Fig. 2. The
mass exchange includes drying and reactions. The defined pyroly-
sis reactions correspond to the scheme developed by Ranzi et al. [5]
with cracking reactions proposed by Blondeau and Jeanmart [6].
The mass exchange is two-ways since the cracking of some com-
pounds produce solid char.
A transient second order formulation is used with a fixed timestep. QUICK is used for the volume fraction coupling, which corre-
sponds to third order accuracy. No turbulence is accounted for
since the flow is assumed laminar. The residuals are allowed to de-
cline below 1 103 before progressing to the next time step.
Interested readers can find further information on the computa-
tional solution procedure in the ANSYS Fluent theory guide [7];
see section on The Pressure-Based Segregated Algorithm. In the
same guide the full set of conservation equations and correlations
used in this publication can be found.
3.1. Species conservation and reactions
The multi-fluid CFD calculation is based on continuity equations
that are solved for all iterations in each time step, in each cell and
phase. Continuity equations account for mass, momentum, energy
and species; for the fluid phase the following momentum Eq. (1) is
solved, where a denotes volume fraction, q the density, v the veloc-
ity, _m g the mass exchange between phases and S the source term
defined by the user.
@
@ t ða g q g Þ þ r ða g q g ~v g Þ ¼ _m g þ S g ð1Þ
For all the species in a phase, such as the compounds in the
fluid, the following analogous Eq. (2) is solved, where Y denotesthe mass fraction of any specie i.
@
@ t ða g q g Y iÞ þ r ða g q g Y i~v g Þ ¼ _m g þ S g ð2Þ
Heterogeneous reactions between phases is included in the
sources term S and for this simulation the primary pyrolysis reac-
tions (1–15, in Fig. 3) corresponding to the scheme developed by
Ranzi et al. [5] is implemented. All reactions are considered first or-
der kinetically controlled, however the heat supply rate to the par-
ticle and the concept of competing reactions makes the picture
more complex. Heat of the reactions is collected from Calonaci
et al. [8]. The primary pyrolysis reactions are in detail described, to-
gether with kinetics and reaction heat, in Appendix A (see Table A).
All the tar species are assumed rapidly evaporating from thebiomass particle, primarily since the fluidization provides swift
flushing of the pyrolysis vapors but also because the general
knowledge is lacking on this particular step (see Dofour et al. [9]
for more discussion on the evaporative step). Hence the gaseous
species trapped in the metaplast, as described in Ranzi et al. [5]
are also directly released into the gas phase.
The thermal cracking of tar is usually effective above 500 C
according to Fagbemi et al. [10] and is commonly known to be fa-
vored by contact with char particles and long residence times.
Hoekstra et al. [11] on the other hand performed an experimental
study indicating that low-mineral char actually is not catalytically
active in a fluidized bed, while in the same study observing that tar
cracking can occur above 400 C. In any case a need to model the
thermal cracking is obvious as the reactor temperature is almost500 C and the gas residence time quite long, around 2 s. In the
Table 2
Pyrolysis product composition, with the rest being mostly water [4].
Product Composition
Gas Specie H2 CH4 CO CO2 C2 C3 C4 C6 C2–C3 ole fins Tota l
wt%wb 0.088 1.581 9.249 8.620 0.461 0.272 0.039 0.005 0.356 20.672
Oil Group Acids Ketones Aldehydes Furans Sugars Phenols Methoxy-phenols Catechols Benzenes Total
wt%wb 9.218 14.044 0.406 4.600 1.741 1.560 6.305 2.039 0.271 34.574
Char Element C H O N Ash Total
wt%wb 14.432 0.541 2.598 0.162 0.280 18.022
wb Wet basis.
H e a t
M a s s
M o m
e n t u m
Heat
Momentum
H
e a t M o m e n t u m
FeedstockSolid 1
(biomass, b)
Solid 2
(sand, s)
Fluid
(gas, g)
E n t r a i n e d s o l i d
Gas and vaporsGas (fluidizing)
Fig. 2. The model in principal with phases and exchange of heat, mass and
momentum.
706 P. Mellin et al./ Fuel 117 (2014) 704–715
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 4/12
model the cracking is considered as extra-particle reactions ap-
plied to the macromolecules with products according to the in-
tra-particle reactions from Blondeau and Jeanmart [6] with
kinetics from Park et al. [12]. These are hypothetical reactions that
have not been proven experimentally but are still helpful to ex-
plore this aspect; note that some solid Char forms due to reaction
3, 4 and 9. The thermal cracking reactions shown in Fig. 3 and are
in detail described in Appendix A, see Table B.
3.2. Energy conservation and phase exchange
The conservation equation for energy in the gas phase is givenby (3), where h g is the internal heat transfer coefficient, h gs is the
heat transfer coefficient between gas-sand and h gb is the heat
transfer coefficient between gas–biomass. As the conservation
equations are based volumetric flow variables, the heat transfer
coefficient is multiplied by the interfacial area concentration a gs
and a gb respectively.
@
@ t ða g q g h g Þ þ r ða g q g ~v g h g Þ ¼ a g
@ p
@ t þ s g : r~v g
r~q g þ S þ h gba gbðT g T bÞ þ h gsa gsðT g T sÞ ð3Þ
h gs is estimated based on the Nusselt number correlation (4) from
Gunn [13] which is applied for the heat transfer between the sandand the gas phase (5).
Biomass C6H8.46O3.9
tw%80.7tw%16.1tw%01.81tw%39.03tw%72.24
lom%87.2lom%56.0lom%99.11lom%20.04lom%75.44
2 1 9 11 12
15 H2O 1 CellA C2H4 1 Acetone 11 CO2
16 Char 2 7 6 H2O 1 LignOH 1
3 CH4
CO2 CO2
4 1 Xylan CH4 H2 H2O
1 LVG 8 C2H4 Char 2 Methanol
CO2 10.1 p-Coumaryl2 CH4
H2 10.08 Phenol2 C2H4
CH2O 0.35 LignCC CO2
CO2 Methanol2 H2
CO2 H2 Ethanol2 Char
H2 OCO 2 H2O H2O 1 Lign
CH4 CO2 Char 2 CH4 14
Acetaldehyde2 Formaldehyde2 C2H4
Char 2 Methanol2 H2 11 Lumped-phenol2 11 H2O
HAA2 Ethanol2 CO2 12 CO2
Glyoxal2 H2 r ahCO 2 Formaldehyde2
HMFU2 CH4 10.3 p-Coumaryl2 Methanol2
Acetone2 C2H4 10.2 Phenol2 Acetaldehyde2
Char 2 10.35 Acrylic-acid2 CH4
C2H4
H2
End product Char 2
Solid intermediete product Acetone2
Tar subject to thermal cracking
Thermal cracking reactions (1-10)
CO2 CO2 CO2 CO2 CO2 12.5 CO2 CO CO2 12 CO2 CO2
C2H4 H2 C2H4 C2H4 H2 11.5 H2 H2 H2 C2H4 C2H4
C2H4 Char 2 Char C2H4 11.75 C2H4 Char 2
81.05%
Cell
C6H10O5
LignOH
18.95%
100%
11.14%
88.86%
60.21%39.79%
41.68%
HCell1 HCell2
100%
2.08.0
0.41
1
0.495
1.32
15.0
1
58.32%
0.7
7.052.061.0
Primary pyrolysis reactions (1-15)
OngiLHngiLCngiLlleCH
0.25 5.735
54.08.0
40% 60%99.97% 100% 100%
C5H8O4 C15H14O4 C16H10O6(OCH3)4 C17H13O4(OCH3)5
5
0.03%
13
51.4521.057.08.0
2
0.2
52.016.0
521.059.0
521.052.0
521.08.09.0
4.11.0
5.02.0
1
0.7
0.65
0.6
1
1.8
6.4
15
Acrylic-acid
3 0.5
HMFU Acetone p-Coumaryl Phenol Xylan LVG
1
1 2 3 4 5 6 7 8 9 10
1.5 0.5
1.25
1
1
2
1
2
2
HAA Glyoxal Lumped-phenol
0.675 0.6
0.65
0.5
5.5
0.2
0.2
526.052.0 0.4
52.02.0
1
3
3
3
2.5
1 0.5
1.5
2.5 1.5
1
2
10 100%
Fig. 3. Shows the reactions pathways implemented in the model, with greyed-out percentages of the predicted results from the model, at 10.75 s.
P. Mellin et al. / Fuel 117 (2014) 704–715 707
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 5/12
h gs ¼ 6k g a g asNus
d2
s
ð4Þ
Nus ¼ ð7 10a g þ 5a2 g Þð1 þ 0:7Re0:2s Pr
1=3Þ
þ ð1:33 2:4a g þ 1:2a2 g ÞRe0:7s Pr
1=3 ð5Þ
For the heat transfer to a particle immersed in a fluidized bed,
several expressions exist and a commonly used one (6) is describedby Mickley and Fairbanks [14] (with adaption for CFD by Papadikis
et al. [15]).
h ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffikmðqC ÞmS
r ð6Þ
For the most important heat transfer to the biomass phase; the
Eulerian framework requires a separate contribution from each
phase (sand and gas). As result the expression (6) has been modi-
fied to account only for the heat conduction from the solid phase,
see (7) with km now is e sks and (qC p)m being esqs(C p)s.
hsb ¼
ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffia2s ksqsC p;spt
r ð7Þ
The exchange coefficient h gb (8) between the gas and the bio-mass is instead calculated according to the well-known Ranz–Mar-
shall equation [16,17], with correction for the local gas volume
fraction. An implicit assumption is then made; that heat transfer
from gas inside the bed occurs through convection and not by con-
duction as implied in the theory by Mickley and Fairbanks [14]. By
implementing the heat transfer in parallel no discontinuity at an
interface of dense region is encountered.
h gb ¼ k g Nubdb
ð8Þ
Nub ¼ 2 þ 0:6Re1=2b Pr1=3 g : ð9Þ
3.3. Momentum conservation and phase exchange
The momentum equation for the gas is expressed in (10), where
p is pressure, s g equals the stress tensor, g is the gravitational force,
K gs is the phase momentum exchange coefficient for gas-sand, K gb
the exchange coefficient for gas-biomass, _m gb~v g is the momentum
transfer associated to mass transfer between gas and biomass.
Equivalent equations are solved for the two granular phase.
@
@ t ða g q g ~v g Þ þ r ða g q g Y g ~v g ~v g Þ ¼ a g rp þ rs g
þ a g q g ~ g þ K gsð~v g ~v sÞ þ K gbð~v g ~v bÞ þ _m gb~v g ð10Þ
K gs is estimated by (10), according to the work by Syamlal and
O’Brien [18]. We have in addition tuned the drag law according pre-
dict the right minimum fluidization velocity umf (c 1 and c 2 in (11)are approximated to 9.19 and 0.28). See Ref. [3] for how for the
gas-sand drag law tuning was carried out. The drag coefficient c Dis given by (12) and the terminal velocity v r ,s for the sand particles
are defined in (13) with coefficients in (14).
K gs ¼3asalq f 4v
2r ;sdsc DResv r ;s
j~v s ~v g j ð11Þ
c D ¼ 0:63 þ 4:8 ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiRes=v r ;s
p !2
ð12Þ
v r ;s ¼ 0:5 A0:06Res þ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ð0:06ResÞ2
þ0:12Resð2B AÞ þ A2q
ð13Þ
A ¼ a4:14 g B ¼
ac 1 g ;a g > 0:85
c 2a1:28 g ;a g 6 0:85
(
K gb is described by (15) which is the drag law from Morsi and
Alexander [19] with u added as an additional input to the drag coef-
ficient (u is set to 0.6). c D is the drag coefficient given by (16) and c 1,
c 2, c 3, c 4 are coefficients defined in (17)–(20) respectively.
K gb ¼18l g
qbd2
b
c DReb24
j~v b ~v g j ð15Þ
c D ¼ 24
Rebð1 þ c 1Re
b2b
Þ þ c 3Rebc 4 þ Reb
ð16Þ
c 1 ¼ expð2:3288 6:4581u þ 2:4486u2Þ ð17Þ
c 2 ¼ 0:0964 þ 0:5565u ð18Þ
c 3 ¼ expð4:905 13:8944uþ 18:4222u2 10:2599u3Þ ð19Þ
c 4 ¼ expð1:4681 þ 12:2584u 20:7322u2 þ 15:8855u3Þ ð20Þ
K sb is the exchange coefficient for sand and biomass, the parti-
cle–particle drag term developed by Syamlal [20] given by (21) is
used (with esb is set to 0.9). g 0,sb is the radial distribution function,
calculated according to Lun et al. [21].
K sb ¼ 3ð1 þ esbÞðp=2 þ c fr ;sbp
2=8Þasqsabqbðds þ dbÞ2 g 0;sb
2p qsd3s þ qbd
3b
j~v s ~v bj
ð21Þ
Bulk viscosity for the sand phase is calculated according to Lun
et al. [21]. The solid shear viscosity is generally assumed to be a
sum of three components, the collisional viscosity calculated
according to Ref. [22], the kinetic viscosity according to Ref. [18]
and the frictional viscosity (assumed negligible due solids volumefraction being far from the maximum packing limit). The granular
temperature is a component of both the collisional viscosity and
the kinetic viscosity and is estimated based on kinetic theory, from
Ref. [18].
3.4. Solution strategy and time step size
The rate of various reactions sometimes differs by several or-
ders of magnitude. As the products of some reactions are reactants
in others, the large difference in rate can cause stability problems.
The system can in that sense be called stiff and for this work we
have used a Stiff Chemistry Solver; which is applied to each
time-step and first solves the flow field with all reaction rates set
to zero. This gives an initial solution which aids in convergenceof the next step which is computing the complete reacting flow
field. The used Stiff Chemistry Solver is integrated in ANSYS Fluent
and is based on the Double precision Variable-coefficient Ordinary
Differential Equation solver (DVODE); see for example Brown et al.
[23]. This solver introduces an error and as a result the Stiff Chem-
istry Solver was only used until the calculation stabilized.
When performing the calculation, a first stage without any bio-
mass feeding was used to obtain a continuously converging simu-
lation with only fluidizing gas and sand. After 1.4 s, the biomass
feeding was switched on and a short time step of 1 105 s was
used. After a short elapse of time, a step size of 1 104 s was
used. Much later in the calculation, the step size could be increased
to a final value of 5 104 s with an average of 25 iterations per
time step. The physical time computed in total is 10.75 s whichcorresponds to over two months of nonstop computation.
708 P. Mellin et al./ Fuel 117 (2014) 704–715
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 6/12
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 7/12
most of the water formation occurred due to secondary tar reac-
tions (STR) and only a small amount during primary biomass
decomposition.
A considerably less amount of work has been done on the
homogenous reactions that do not (or only slightly) alter theamount of liquid. In the context of this model, those would
be the reactions that follow the formation of e.g. Levoglucosan
and contribute to the vast variety of molecules found in pyro-
lysis oil. These are likely the bulk source of water; as seen in
the work by Zhang et al. [27], most of the typical pathways
of continued levoglucosan decomposition involve sheddingwater.
(a) (b)
(c) (d) 0 . 9
7 9
0 . 0
0 0
0 . 0
0 0
0 . 0
2 1
0 . 9
7 9
0 . 0
0 0
0 . 0
0 0
0 . 0
2 1
0 . 8
7 7
0 . 0
1 3 0 . 0 9 3
0 . 0
1 7
0 . 8
0 8
0 . 0 3 0
0 . 1
4 6
0 . 0
1 6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
C H O Ash
E l e m
e n t a l w t f r a c t i o n o f c h a r
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
Gas Liquid Char
W
t f r a c t i o n o f p y r o l y s i s
p r o d u c t s
Without tar cracking With tar cracking With tar cracking incl. unreacted biomass Experiment
0 . 4
6 3
0 . 0
6 7
0 . 4
7 0
0 . 4
5 8
0 . 0
6 5
0 . 4
7 7 0
. 5 7 6
0 . 0
6 7
0 . 3
5 7
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1.0
C H O
E l e m e n t a l w t f r a c t i o n o f
o r g a n i c l i q u i d
0 . 1 1 7
0 . 1
9 4
0 . 1
3 8
0 . 2
4 9
0 . 0 6 3
0 . 1
3 4
0 . 2
4 6
0 . 1
9 8
0 . 3
0 3
0 . 1
1 9
0 . 1
4 3
0 . 0 6 4
0 . 2
8 3
0 . 4
7 7
0 . 0
4 6
0.0
0.1
0.2
0.3
0.4
0.5
0.6
CH4 H2 CO2 CO C2H4
M o l f r a
c t i o n o f p e r m e n e n t
g a s e s
Without tar cracking With tar cracking Experiment
H2O
fraction
0 . 1 9
1
0 . 5
9 3
0 . 1
0 1 0 . 2
5 9
0 . 4 3 8
0 . 1
4 5 0 . 2
4 8 0
. 4 2 1
0 . 1 7 8
0 . 2 0
7 0 . 3
4 6
0 . 1 8
0
0 . 1 5 9
0 . 1
1 5
0 . 2
5 4
0 . 1
5 2
fraction
Organic
Fig. 4. The pyrolysis products, at 10.75 s. Where (a) is the elemental composition of the biomass, (b) the volume fraction of gas species (c) the elemental composition incl. ash
of the char fraction, (d) the overall pyrolysis products summed up on wet basis.
Fig. 5. Shows the velocity profile of the different phases along the volume fraction, at 10.75 s. Note that volume fraction of biomass is displayed with a logarithmic scale.
710 P. Mellin et al./ Fuel 117 (2014) 704–715
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 8/12
In fact several groups of primary pyrolysis products are known
to produce water when they react (dehydrate) homogenously in
the vapor phase [11]. A study by Faix et al. [28] concerning lignin
clearly shows water formation by TGMS, at temperatures usually
above primary biomass decomposition. A common product of lig-
nin is methoxyphenols, such as isoeugenol, and forms by dehydra-
tion according to [29]. The formation mechanism of sugars(monosaccharaides) is dehydration according to [30] and a com-
mon product is furans which will also form by dehydration [31].
In the model all furans are represented by hydroxymethylfurfural
(HMFU) with a yield of about 1% for ‘‘Without tar cracking’’ and
0.66% ‘‘With tar cracking’’. This is too low as Table 2 states about
4.6% furans on wet basis (5.1% on dry basis) in the experiment.
As a result of further dehydration of organic tar, the oxygen con-
tent of the liquid will decrease as well. As of now it is too high as
visible in Fig. 4(a).
4.2. Instantaneous results: Volume fraction and velocity
Fig. 5 shows the velocity magnitude and volume fraction of gasbiomass and sand along a surface in the reactor. The biomass is
quite well distributed in the fluidized bed and as visible some
floats on top of the bed, in the splash zone.
The volume fraction along the centerline of the reactor is shown
in Fig. 6. The velocity in y-direction for the three phases is shown in
Fig. 7; the acceleration at the outlet is very rapid (for the gas phase
velocity is at most 21.82 m/s). The biomass phase velocity strongly
correlates with the sand phase in the dense region and correlates
with the gas phase in the less dense region. This means that themomentum is easily transferred to the biomass particles and more
easily so by the sand phase. The pressure in the reactor is also
shown in Fig. 7. A decline in pressure clearly correlates with an
acceleration of the gases. The pressure drop over the bed in the
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
0 0.3 0.6 0.9
V o l u m e f r a c t i o n
V
u
m
a
o
Reactor height, meactor height m
Gas
Sand
Biomass
Fig. 6. Shows the volume fraction of gas, sand and biomass along the centerline, at
10.75 s.
Atmosphericpressure
101
103
105
107
109
-1
-0.5
0
0.5
1
1.5
2
2.5
3
0 0.3 0.6 0.9
P r e s s u r e ,
k P a
V e l o c i t y , m / s
Reactor height, m
Gas y-velocity
Sand y-velocity
Biomass y-velocity
Predicted pressure
Fig. 7. The velocity for gas, sand and biomass as well as gas pressure, as function of
reactor height, at 10.75 s.
Fig. 8. Shows the temperature of all phases, at 10.75 s. A surface along the Z = 0
plane shows a color map in degrees centigrade.
0
50
100
150
200
250
300
350
400
450
500
0 0.3 0.6 0.9
H e a t t r a n s f e r c o e f f i c i e n
t , W / m 2 / K
Reactor height, m
Gas-biomass
Sand-biomass
Total
Fig. 9. The heat transfer coefficient at the interface of gas-biomass and sand-biomass, at 10.75 s.
P. Mellin et al. / Fuel 117 (2014) 704–715 711
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 9/12
experiment is 1212.1 ± 128.9 Pa while the predicted value is
1246 Pa.
4.3. Instantaneous results: temperature and heat transfer
The temperature of the phases is shown in Fig. 8. The tempera-
ture is quite uniform in the lower part of the reactor and the gas
and sand temperature is close. The low temperature zone caused
by biomass feeding is small but clearly visible in the biomass
phase. As the volume fraction of biomass is low, no significant
amount of heat is transferred from the gas and sand which meansthe temperature does not drop noticeably.
The heat transfer coefficient is presented in Fig. 9, as the fluctu-
ations indicate the local volume fraction of gas and sand is factored
in. In the lower part (dense region) of the bed the heat transfer
from the sand dominates while higher up, the heat transfer from
gas obviously dominates. At the outlet of the reactor, the heat
transfer coefficient for the gas phase increases as the Reynolds
number rises at higher velocities.
Since the temperature of the sand and gas is very close, the heattransfer coefficients of each may be added together which is shown
as a black line in Fig. 9. The total heat transfer coefficient varies be-
tween 200 and 320 W/m2/K which is close to the range found by
Papadikis et al. [32] for a slightly smaller but otherwise similar
reactor. The total heat transfer coefficient also shows that high
heat transfer is possible in the splash zone. Here the high velocity
of gas results in high convective heat transfer, due to rupture of
bubbles.
Fig. 10 shows the heat transfer with the temperature difference
factored in. Some biomass goes directly upwards from the feeding
line and enters the splash zone. A low temperature contributes to a
very fast actual heat transfer. Biomass entering this region is thus
not necessarily negative and Fig. 9 together with Fig. 10 shows that
overall heat transfer actually is highest here.
4.4. Biomass conversion, residence time and Species
The total product formation rate was, in the case of ‘‘Without
tar cracking’’ 6.63 104 kg/s which means some accumulation
of biomass occurs during the beginning of the simulation. The li-
quid, gas and water that escape the outlet of the reactor amounts
to 3.99 104 kg/s (see Table C in Appendix A for specific compo-
nents). This is the basis for ‘‘With tar cracking’’ added with the char
formation; from both primary and secondary pyrolysis reactions
integrated over the whole domain. The total product formation
rate for the case ‘‘With tar cracking’’ is 4.66 104 kg/s. With the
unreacted biomass factored in the total output rate is
4.86 104
kg/s with specific solid components shown inTable D. This means some time still remains for the output rate
to become equal to the input rate 5.56 104 kg/s and as a result
of the cases ‘‘With tar cracking’’ and ‘‘With tar cracking incl. unre-
acted biomass’’ should be treated as indicative.
The unreacted biomass is shown in Table D, in Appendix A,
amounts to a small percentage of the fed biomass. The percentages
of Cellulose, Hemicellulose and Lignin in the unreacted biomass are
48.7%, 23.2% and 28.0% respectively while in the biomass it is
0
5000
10000
15000
20000
25000
30000
0 0.3 0.6 0.9
H e a t t r a n s f e
r , W / m
H
a
e
W
/
m
Reactor eactor
height, meight m
Gas-biomass
Sand-biomass
Total
Fig. 10. Shows the heat transfer coefficient multiplied with the temperature
difference between gas-biomass and sand-biomass, at 10.75 s.
Fig. 11. Path lines along the velocity field of gas, at 10.75 s.
1 6
1 0 9
1 0 8
7 3
5 1
2 8
1 7
1 2
1 8
7 7 4
2 2 1 5
2 6 ( t > 5 s )
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 4.5 5.0
0
0.05
0.1
0.15
0.2
0.25
0.3
0
20
40
60
80
100
120
R e s i d e n c e t i m e d i s t r i b u t i o n ( R T D )
N u m b e r o f p a t h l i n e s
Time, s
Fig. 12. The number of path lines, following the velocity field of the gas phase, in
each time span represented by bars. The fitted gas residence time distribution isrepresented by a curve with axis on the right.
712 P. Mellin et al./ Fuel 117 (2014) 704–715
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 10/12
42.2%, 31.0% and 26.8%. While this could change, it is interestingthat Hemicellulose reacts to a higher degree compared to Cellulose
and Lignin. This could be due to Hemicellulose beginning to react
at lower temperatures (according to Ref. [5]) and thus getting a
head start.
In order to get a sense of the residence time, mass-less particle
tracking along the velocity field of the gas phase was made. An
example is shown in Fig. 11, where 21 mass-less particles were
released from the inlet. In total 480 were released at different
points in time from the biomass inlet. 468 reached the outlet and
the distribution is shown in Fig. 12, with the required time for
reaching the outlet registered on the horizontal axis. The average
gas residence time is 1.8 s with around 4% (16 path lines) residing
in the reactor less than 1 s. The velocity field for the biomass has
not yet stabilized enough and actual particle tracking may be nec-essary to evaluate the precise particle residence time distribution.
We can however safely assume that the average particle residencetime is higher than the average gas residence time, 1.8 s, since the
biomass particles collides with the sand particles and backflow is
possible along the flight upwards. This gives the particles time to
react but a 100% conversion within the framework of the model
cannot be expected or is not even reasonable. Within the model
framework a 100% conversion would mean that the char fraction
entirely consists of carbon and ash.
The reaction pathways for the first Case is given in percent
in Fig. 3. LVG is highly favored in this case as almost all the cel-
lulose decomposes along this pathway. LignO decomposes very
slowly compared to LignH as most of the LignOH are formed
from LignH while more LignO is fed to the system. Conse-
quently, some unreacted LignO is still inside the reactor and
at 10.75 s, it amounts to 2.07
10
4
kg of LignO, while only0.04 104 kg of LignH.
Fig. 13. Shows some selected species in profile along the Z = 0 plane, at 10.75 s.
Table A
Primary pyrolysis reactions (Reaction 1-15) from [5] with reaction heat from [8] and drying (Reaction 16) from [6].
Reaction A (s1) E
(kJ mol1)
Dh
(kJ kg1)
1 Cell? CellA 8 1013a 192.5a 447.7
2 Cell? 5H2O + 6Char 8 107 125.5 1087.8
3 CellA? LVG 4Ta 41.8a 732.2
4 Ce llA? 0.95HAA + 0.25Glyoxal + 0.2Acetaldehyd + 0.25HMFU + 0.2Acetone + 0.16CO2 + 0.23CO + 0.9H2O + 0.1CH4 + 0.61Char 1 109a 133.9 899.6
5 H Cell? 0.4HCell1 + 0.6HCell2 1 1010a 129.7a 548.1
6 HCell1? 0.75H2 + 0.8CO2 + 1.4CO + 0.5Formaldehyde 3 109a 113.0a 447.7
7 HCell1? Xylan 3Ta 46.0a 707.1
8 HCell2? CO2 + 0.5CH4 + 0.25C2H4 + 0.8CO + 0.8H2 + 0.7Formaldehyde + 0.25Methanol + 0.125Ethanol + 0.125H 2O + Char 1 1010 138.1 259.4
9 L ig nC? 0.35LignCC + 0.1 pCoumaryl + 0.08Phenol + 0.41C2H4 + H2O + 0.495CH4 + 0.32CO + CO + H2 + 5.735Char 4 1015 202.9 602.5
10 LignH? LignOH + Acetone 2 1013 156.9 523.0
11 L ig nO? LignOH + CO2 1 109 106.7 510.4
12 LignCC? 0.3pCoumaryl + 0.2Phenol + 0.35Acrylic-acid + 0.7H2O + 0.65CH4 + 0.6C2H4 + 1.8CO + H2 + 6.4Char 5 106 131.8 288.7
13 LignOH? Lign + H2O + Methanol + 0.45CH4 + 0.2C2H4 + 2CO + 0.7H2 + 4.15Char 3 108 125.5 100.4
14 L ig n? Lumped-phenol 8Ta 50.2a 577.4
15 Lign? H2O + 2CO + 0.2Formaldehyde + 0.4Methanol + 0.2Acetaldehyd + 0.2Acetone + 0.6CH 4 + 0.65C2H4 + 0.5H2 + 5.5Char 1.2 109a 125.5a209.2
16 H2O(l)? H2O(g) 5.3 1010 88 2260.0
a Modified for high temperature according to Ref. [6].
P. Mellin et al. / Fuel 117 (2014) 704–715 713
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 11/12
Table B
Secondary pyrolysis reactions [6].
Reaction A (s1) E (kJ mol1) Dha (kJ kg1)
1 HMFU? 3CO + 1.5C2H4 4.28 106 108.0 642.7
2 Acetone? 0.5CO2 + 0.5H2 + 1.25C2H4 4.28 106 108.0 1878.2
3 pCoumaryl? CO2 + 2.5C2H4 + 3Char 4.28 106 108.0 359.6
4 Phenol? 0.5CO2 + 1.5C2H4 + 2.5Char 4.28 106 108.0 143.1
5 Xylan? 2CO2 + H2 + 1.5C2H4 4.28 106 108.0 563.0
6 LVG? 2.5CO2 + 1.5H2 + 1.75C2H4 4.28 106 108.0 1701.67 HAA? 2CO + 2H2 4.28 106 108.0 3562.7
8 Glyoxal? 2CO + H2 4.28 106 108.0 156.6
9 Lumped-phenol? 2CO2 + 3C2H4 + 3Char 4.28 106 108.0 693.8
10 Acrylic-acid? CO2 + C2H4 4.28 106 108.0 912.9
a By balance.
Table C
Product yields for specific components, with formation rate from primary pyrolysis and outflow rate for the case ‘‘With tar cracking’’.
Specie Formula Included in Formation rate of primary pyrolysisa Outflow rateb
(kg/s) (wt%) (kg/s) (wt%)
Methane CH4 Gas yield 1.38E-05 2.08 1.12E-05 2.40
Carbon-monoxide CO Gas yield 5.18E-05 7.81 4.42E-05 9.49
Carbon-dioxide CO2 Gas yield 4.52E-05 6.82 4.53E-05 9.72Hydrogen H2 Gas yield 2.87E-06 0.43 2.57E-06 0.55
Water, gas H2O Liquid yield 7.60E-05 11.46 7.40E-05 15.88
Nitrogen N2 – – – 6.39E-04 -
Formaldehyde CH2O Liquid yield 2.01E-05 3.03 1.46E-05 3.13
Acetaldehyde CH3HCO Liquid yield 1.82E-06 0.27 9.37E-07 0.20
Methanol CH3OH Liquid yield 9.59E-06 1.45 7.07E-06 1.52
Glyoxal C2H2O2 Liquid yield 2.97E-06 0.45 1.39E-06 0.30
Ethylene C2H4 Gas yield 1.31E-05 1.98 1.73E-05 3.71
Hydroxyacetaldehyde (HAA) C2H4O2 Liquid yield 1.17E-05 1.76 5.46E-06 1.17
Ethanol C2H5OH Liquid yield 5.85E-06 0.88 4.57E-06 0.98
Acrylic-acid C3H4O2 Liquid yield 1.17E-05 1.76 1.87E-09 0.00
Acetone C3H6O Liquid yield 5.80E-06 0.87 6.41E-06 1.38
Xylan C5H8O4 Liquid yield 3.97E-05 5.99 4.75E-05 10.19
Levoglucosan (LVG) C6H10O5 Liquid yield 2.64E-04 39.82 1.03E-04 22.10
Phenol C6H5OH Liquid yield 4.38E-06 0.66 3.27E-06 0.70
Hydroxymethylfurfural (HMFU) C6H6O3 Liquid yield 6.44E-06 0.97 3.01E-06 0.65
pCoumaryl C9H10O2 Liquid yield 8.74E-06 1.32 6.52E-06 1.40
Lumped-phenol C11H12O4 Liquid yield 2.89E-07 0.04 6.41E-08 0.01
Char C Char yield 6.72E-05 10.13 6.76E-05c 14.51
a Used for the case ‘‘Without tar cracking’’.b Used for the case ‘‘With tar cracking’’.c Value used for the case ‘‘With tar cracking’’, not the actual outflow rate.
Table D
All solid species with feeding and outflow rate.
Specie Formula Feeding rate Outflow ratea
kg/s wt% kg/s wt%
Cellulose (Cell) C6H10O5 2.11E-04 38.01 2.05E-06 10.76
Activated cellulose (CellA) C6H10O5 - - 7.22E-06 37.95Hemicellulose (HCell) C5H8O4 1.55E-04 27.81 1.81E-10 0.00
Hemicellulose 1 (HCell1) C5H8O4 – - 5.80E-07 3.05
Hemicellulose 2 (HCell2) C5H8O4 – - 3.84E-06 20.18
C-rich lignin (LignC) C15H14O4 9.04E-05 16.28 2.57E-06 13.52
H-rich lignin (LignH) C16H10O6(OCH3)4 8.06E-06 1.45 1.34E-08 0.07
O-rich lignin (LignO) C17H13O4(OCH3)5 3.54E-05 6.37 1.81E-08 0.10
C-rich lignin (LignCC) C15H14O4 – – 7.99E-07 4.20
OH-rich lignin (LignOH) C19H22O8 – – 1.85E-06 9.75
Lignin (Lign) C11H12O4 – – 8.10E-08 0.43
Water, liquid H2O 5.44E-05 9.8 8.85E-12 0.00
Ash SiO2 1.56E-06 0.28 1.08E-07b –
Char C – – 1.02E-06b –
a Used for the case ‘‘With tar cracking incl. unreacted biomass’’.b Values not used.
714 P. Mellin et al./ Fuel 117 (2014) 704–715
8/10/2019 1-s2.0-S0016236113008387-main (1)
http://slidepdf.com/reader/full/1-s20-s0016236113008387-main-1 12/12
Fig. 13 shows the molar concentration of some selected species
in the domain. The species propagates throughout the domain, a
higher concentration of species in the midsection indicates that
the results have not stabilized fully yet. An accumulation of bio-
mass is seen on the left hand side, opposite to the feeding line,
which results in concentrated gas release. This accumulation is also
visible in the volume fraction of the biomass, see Fig. 5. Other accu-
mulations are also visible in Fig. 5 but high heat transfer in combi-nation with an almost complete evaporation of water cause the
high pyrolysis reaction rate shown Fig. 13.
5. Conclusion
The main conclusions of this study can be summarized by the
following paragraphs.
The primary pyrolysis scheme used in this work is possible to
implement in a comprehensive CFD model. The thermal cracking
reactions are helpful but should be replaced or put in parallel with
other secondary reactions that can explain the formation of more
water.
The even distribution of biomass in the bed suggests good oper-
ational conditions. Pressure drop as compared with experiment is
accurate. Total heat transfer rate was found to be highest in the
splash zone.
The momentum of the sand is easily transferred to the biomass
particles and more so compared to the gas phase. The residence
time for the gas is 1.8 s on average and the biomass particles
should reside at least as long in the reactor, which is sufficient time
for conversion.
Evaluating the percentages of different pathways in the reaction
scheme, shows that almost all the cellulose is converted into levo-
glucosan. As indicated by the model results, the conversion of
hemicellulose is more complete than cellulose and lignin.
Acknowledgments
This work was supported by the Swedish National Infrastructurefor Computing (SNIC 001-11-26) via PDC. KIC Innoenergy and The
Swedish Energy Agency (Contract No. 33284-1) are acknowledged
for funding of the project. Preem, Boson Energy, Sveaskog and The
Division of Chemical Technology at KTH are acknowledged for
support in other ways. We are grateful to all above named
organization for the opportunity to carry out this research.
Appendix A
See Tables A–D.
References
[1] Dupont C, Chen L, Cances J, Commandre J, Cuoci A, Pierucci S, et al. Biomasspyrolysis: Kinetic modelling and experimental validation under hightemperature and flash heating rate conditions. J Anal Appl Pyrol2009;85:260–7.
[2] Papadikis K, Gerhauser H, Bridgwater AV, Gu S. CFD modelling of the fastpyrolysis of an in-flight cellulosic particle subjected to convective heattransfer. Biomass Bioenergy 2009;33:97–107.
[3] Mellin P, Zhang Q, Kantarelis E, Yang W. An Euler–Euler approach to modelingbiomass fast pyrolysis in fluidized-bed reactors – focusing on the gas phase.Appl Therm Eng 2013;58:344–53.
[4] Kantarelis E, Yang W, Blasiak W. Production of liquid feedstock from biomassvia steam pyrolysis in a fluidised bed reactor. Energ Fuel 2013;27:4748–59.
[5] Ranzi E, Faravelli T, Frassoldati A, Migliavacca G, Pierucci S, Sommariva S.Chemical kinetics of biomass pyrolysis. Energy Fuel 2008;22:4292–300 .
[6] Blondeau J, Jeanmart H. Biomass pyrolysis at high temperatures: prediction of gaseous species yields from an anisotropic particle. Biomass Bioenergy2012;41:107–21.
[7] ANSYS FLUENT 14.5, Theory Guide; ANSYS, Inc., 2012.[8] Calonaci M, Grana G, Hemings EB, Bozzano G, Dente M, Ranzi E.
Comprehensive kinetic modeling study of bio-oil formation from fastpyrolysis of biomass. Energy Fuel 2010;24:5727–34.
[9] Dufour A, Ouartassi B, Bounaceur R, Zoulalian A. Modelling intra-particlephenomena of biomass pyrolysis. Chem Eng Res Des 2011;89:2136–46.
[10] Fagbemi L, Khezami L, Capart R. Pyrolysis products from different biomasses:application to the thermal cracking of tar. Appl Energy 2001;69:293–306 .
[11] Hoekstra E, Westerhof RJM, Brilman W, et al. Heterogenous and homogenousreactions of pyrolysis vapors from pine wood. AIChE J 2012;58:2830–42 .
[12] Park WC, Atreya A, Baum HR. Experimental and theoretical investigation of heat and mass transfer processes during wood pyrolysis. Combust Flame2010;157:481–94.
[13] Gunn DJ. Transfer of heat or mass to particles in fixed and fluidized beds. Int JHeat Mass Transfer 1978;21:467–76.
[14] Mickley HS, Fairbanks DF. Mechanism of heat transfer to fluidised beds. AIChE J 1955;1:374–84.
[15] Papadikis K, Gu S, Bridgwater AV. CFD modelling of the fast pyrolysis of biomass in fluidised bed reactors. Part B: heat, momentum and mass transportin bubbling fluidised beds. Chem Eng Sci 2008;64:1036–45.
[16] Ranz WE, Marshall WR. Evaporation from drops: Part I. Chem Eng Prog1952;48:141–6.
[17] Ranz WE, Marshall WR. Evaporation from drops: Part II. Chem Eng Prog1952;48:173–80.
[18] Syamlal M, Rogers W, O’Brien TJ. MFIX documentation: theoryguide. Morgantown, WV: U.S. Department of Energy; 1993 .
[19] Morsi SA, Alexander AJ. An investigation of particle trajectories in two-phaseflow systems. J Fluid Mech 1972;55:193–208.
[20] Syamlal M. The particle–particle drag term in a multiparticle model of fluidization. VA: Springfield; 1987.
[21] Lun CKK, Savage SB, Jeffrey DJ, Chepurnity N. Kinetic theories for granularflow: inelastic particles in couette flow and slightly inelastic particles in ageneral flow field. J Fluid Mech 1984;140:223–56.
[22] Gidaspow D, Bezburuah R, Ding J. Hydrodynamics of circulating fluidized beds,kinetic theory approach. In: Paper presented at: proceedings of the 7thengineering foundation conference on fluidization, 1992.
[23] Brown PN, Byrne GD, Hindmarsh AC. VODE: a variable-coefficient ODE solver.
Siam J Sci Stat Comp 1989;10:1038–51.[24] Alexander T. Water reuse in paper mills: measurements and control problems
in biological treatment, licentiate thesis. Lund Institute of Technology; 2003 .[25] Boroson ML, Howard JB, Longwell JP, Peters WA. Product yields and kinetics
from the vapor phase cracking of wood pyrolysis tars. AIChE J 1989;35:120–8 .[26] Morf P, Hasler P, Nussbaumer T. Mechanisms and kinetics of homogeneous
secondary reactions of tar from continuous pyrolysis of wood chips. Fuel2002;81:843–53.
[27] Zhang X, Yang W, Blasiak W. Thermal decomposition mechanism of levoglucosan during cellulose pyrolysis. J Anal Appl Pyrol 2012;96:110–9 .
[28] Faix O, Jakab E, Till F, Székely T. Study on low mass thermal degradationproducts of milled wood lignins by thermogravimetry-mass-spectrometry.Wood Sci Technol 1988;22:323–34.
[29] Patwardhan PR, Satrio JA, Brown RC, Shanks BH. Product distribution from fastpyrolysis of glucose-based carbohydrates. J Anal Appl Pyrol 2009;86:323–30.
[30] Lin YC, Cho J, Tompsett GA, Westmoreland PR, Huber GW. Pyrolysis kineticsand mechanism of cellulose. J Phys Chem C 2009;113:20097–107.
[31] Shen DK, Gu S, Bridgwater AV. Study on the pyrolytic behaviour of xylan-basedhemicellulose using TG–FTIR and Py–GC–FTIR. J Anal Appl Pyrol
2010;87:199–206.[32] Papadikis K, Gu S, Bridgwater AV. CFD modelling of the fast pyrolysis of
biomass in fluidised bed reactors: modelling the impact of biomass shrinkage.Chem Eng J 2009;149:417–27.
P. Mellin et al. / Fuel 117 (2014) 704–715 715