See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/222834939

Modelling wood combustion under fixed bed conditions☆

Article  in  Fuel · April 2003

DOI: 10.1016/S0016-2361(02)00296-X

CITATIONS READS

101 85

3 authors, including:

Bernhard Peters
University of Luxembourg
129 PUBLICATIONS   1,188 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

www.xdem.de. View project

Blast Furnace modelling View project

All content following this page was uploaded by Bernhard Peters on 25 July 2015.

The user has requested enhancement of the downloaded file.

 Fuel 82 (2003) 729–738
www.fuelfirst.com

Modelling wood combustion under fixed bed conditionsq

Christian Brucha, Bernhard Petersb,*, Thomas Nussbaumerc
a
Department of Energy Technology, ETH Zurich, CH 8092 Zurich, Switzerland
b
Institute for Nuclear Technology and Energy Technology, Karlsruhe Research Centre, P.O. Box 3640, Karlsruhe D-76021, Germany
c
Verenum Research, Langmauerstr. 109, CH 8006 Zurich, Switzerland
Available online 22 October 2002

Abstract
A computer model describing the conversion of wood under packed-bed conditions is presented. The packed bed is considered to be an
arrangement of a finite number of particles, typically sized between 5 and 25 mm, with a void space left between them. Each particle is
undergoing a thermal conversion process, which is described by a one-dimensional and transient model.
Within the single-particle model, heating, drying, pyrolysis, gasification and combustion are considered, whereby each particle exchanges
energy due to conduction and radiation with its neighbours. Because of the one-dimensional discretization of the particles, heat transfer and
mass transfer is taken into account explicitly. Therefore, no macrokinetic data are needed within the model. For ease of implementation and
access, kinetic data and property data are stored in a database. The global conversion of the packed bed is represented by the contributions of
single particles, where each particle is coupled to the surrounding gas phase by heat and mass transfer. For gas phase flow through the porous
bed, the conservation equations for mass, momentum and energy are solved on a Cartesian mesh by a Finite Volume method.
Experiments have been performed to validate the single particle model for the conversion of beech wood during pyrolysis and char
combustion. Agreement between experimental and predictions obtained by the model is very satisfactory. However, for wet wood, changes
in structure seem to enhance the heat transfer to the solid which is not yet covered in the model.
q 2002 Elsevier Science Ltd. All rights reserved.
Keywords: Packed bed; Modelling; Wood combustion

1. Introduction commercial CFD codes can be applied, the prediction of solid

fuel conversion into raw gas in the packed bed is difficult.
An increasing use of renewable energy sources, such as Therefore, a model is required which can predict the
wood, is an important issue in a future energy market to solve influence of operational parameters on solid fuel conver-
the CO2 problem. Most automatic furnaces used in wood sion, in such a way that the boundary conditions to gas phase
combustion are understoker and grate furnaces, in which the simulation by a CFD code can be calculated.
fuel is converted in a packed bed and the combustion process Several approaches predicting the solids conversion rate in
packed beds, or the species distribution in the raw gas are
is completed in the furnace chamber. In recent years,
presented in the literature. Ford et al. [1] described a model of
progress has been achieved in the optimisation of furnaces
the conversion of coal on a grate which is based on the
with respect to reduced emissions and higher efficiencies.
assumption of a combustion regime fully controlled by
However, further improvement requires detailed knowledge
boundary layer diffusion. No further description of the flow in
of the combustion process in the furnace.
the bed is taken into account. To predict the velocity of the
For the optimisation of flow conditions in combustion
burning front in a bed of wood particles, Fatehi and Kaviany
chambers, computational fluid dynamics (CFD) have
[2] developed a model based on equilibrium assumptions.
become a useful tool, employed together with experiments. Raupenstrauch [3] coupled a model to calculate solids
While for investigation of the flow conditions in the gas phase conversion by a CFD code which predicted the steady state
flow through the porous bed. Particle mass is coupled to its
* Corresponding author. Tel.: þ 49-7247-823491; fax: þ 49-7247-
824837.
size by a simplified approach but no discretization of the
E-mail address: bernhard.peters@avl.com (B. Peters). particle’s dimension is applied. A transient heterogeneous
q
Published first on the web via Fuelfirst.com—http://www.fuelfirst.com model of packed beds was developed by Hartner [4], with
0016-2361/03/$ - see front matter q 2002 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 1 6 - 2 3 6 1 ( 0 2 ) 0 0 2 9 6 - X
730 C. Bruch et al. / Fuel 82 (2003) 729–738

Nomenclature G diffusion coefficient

1P porosity (–)
A Area (m2)
l thermal conductivity (W/mK)
ai Interpolation function coefficients
m dynamic viscosity (kg/m s)
c gas concentration (mol/m3)
r density (kg/m3)
cv specific heat (J/kg K)
v source term
D diffusion coefficient (m2/s)
v averaged source term
Ea activation energy (kJ/mol)
kl average value in porous media
Fi!j view factor
k0 frequency factor (1/s)
K permeability (m2) Indices
m
_ mass flow (kg/s) b bed
n shape factor (0 for plate, 1 for cylinder, 2 for c contact
sphere) ch chemical reactions
q_ heat flux (W/m2) cond conduction
r_ reaction rate (mol/m3 s) conv convection
R gas constant (J/mol K) E east face of volume element
t time (s) eff effective
Dt outer integration step (s) h heating
T temperature (K) i inner
u gas velocity in the particle (m/s) pa particle
u mass of water/mass of dry fuel rad radiation
v gas velocity in the bed (m/s) s value at the surface
Yi species mass fraction th thermal
a heat transfer coefficient (W/m2 K) w dry wood
b mass transfer coefficient (m/s) 1 value in the bulk flow
g contact angle (8)

the solid fuel reactor discretized in one-dimension. Beckmann valid. Therefore the aim of the present work is to develop a
and Scholz [5] divided the bed of a grate furnace into various general model covering the entire conversion process of
zones, each of them represented by a perfectly stirred reactor biomass fuels, from drying to char conversion, which can be
to describe char conversion on a grate. Mixing due to motion used to determine rate limiting conditions and potentials for
on the grate can be accounted for by the mass exchange optimisation in packed-bed combustion. Modelling the
between the reactors. Based on laboratory reactor exper- packed bed as a porous media, as often presented in
iments Kuo [6] applied empirical correlations predicting literature, macrokinetic or empirical data is required to
typical design parameters, such as grate burning rate and capture the influence of particle size on the conversion.
volume load for the combustion of wood in packed beds. Goh Thus, to reduce the need of empirical data, a different
et al. [7] developed a model based on different zones for approach needs to be chosen here.
virgin, dried and pyrolysed fuel and ash to calculate the
boundary conditions for a CFD code. Describing the fuel
conversion requires macrokinetic data. Investigating simul- 2. Modelling thermal conversion of a packed-bed
taneous drying and pyrolysis of a packed bed of wood
particles, Saastamoinen and Haukka [8] presented a one- 2.1. Single-particle model
dimensional model, where drying and pyrolysis are solved on
a single particle with the vaporisation taking place on a In wood furnaces, the packed bed consists of particles of a
shrinking core. For comparisons with experimental results at large variety of sizes, typically sized between 5 and 25 mm
a laboratory furnace, Shin and Choi [9] recently discussed a for automatic furnaces. During the thermal conversion of
one-dimensional transient model for the conversion of waste wood, different processes such as drying, pyrolysis and char
particles on a grate. Using a finite volume approach global conversion, occur which may overlap to some extent within a
kinetic data are applied to solids conversion in a volume particle depending, e.g. on fuel type, particle size or heating
element assuming kinetic control of the process. rate. It is not feasible to determine macrokinetic data for all
Most of the models are restricted to a certain part of the the different conditions which can prevail within a packed
conversion process or require empirical information, thus bed. To apply kinetic data derived from dust samples under
limiting application to a specific case for which the data are the assumption of kinetic control of the conversion process to
 C. Bruch et al. / Fuel 82 (2003) 729–738 731

larger particles, heat and mass transfer within the particle Boundary conditions. Boundary conditions are required to
need to be taken into account [10]. Therefore, spatial complete the system of equations. At the centre of the fuel
discretization is required, when describing the conversion particle, all gradients are set equal to zero. At the particle
process of a single fuel particle. However, as the formulation surface, heat is transferred between the fuel particle and the
for a single particle is part of an overall packed-bed model, surrounding gas phase under the impact of convection.
the numerical effort which can be spent on this subprocess is Moreover heat transfer to and from the particle occurs as a
limited. As a compromise between computation time and result of conduction and radiation between the solid particles,
accuracy of description, a transient one-dimensional model is which will be discussed below. Thus, the energy at the outer
employed to predict the conversion process in each particle. shell can be balanced according to Eq. (3).
In the literature, models are discussed for simulating one or 
two sub-processes such as pyrolysis or char combustion of ›kTl 
2leff ¼ aðkTR l 2 T1 Þ þ q_ rad þ q_ cond ð3Þ
single wood particles among the overall process [11 – 14]. ›r r¼R
Here, the whole process, including drying, pyrolysis, char Within the packed bed, the heat transfer coefficient is based on
combustion and gasification, is incorporated in the authors’ an empirical correlation accounting for an increase in heat
model. The governing equations describing the conservation transfer relative to a single particle in a gas flow according to
of a scalar f versus the radius of a solid particle are Ref. [14]. Thus, calculation of the Nusselt number in the bed
implemented in a general form with the exponent n in Eq. (1), Nub bases on the Nusselt number of a single particle in a gas
representing the formula for an infinite plate, an infinite flow Nupa, as stated in Eq. (4)
cylinder and a sphere for n ¼ 0; 1; 2:

 Nub ¼ 1 þ 1:5ð1 2 1P ÞNupa ð4Þ
› 1 › ›
krlkfl ¼ n Geff rn krlkfl 2 rn krlukfl þ v ð1Þ
›t r ›r ›r Calculated values of the heat transfer coefficient based on this
Each property value in a volume element of the porous approach are in good agreement with experimental values
particle is calculated as an average over the values of the measured in a packed bed of wood particles [15], even if the
species currently present in the control volume. The Reynolds numbers in packed bed combustion are quite low.
equations are discretized in space by a finite volume approach The boundary conditions for mass transfer at the particle
with the values for energy and species averaged over a control surface are given by
volume assuming thermal equilibrium between the solid and 
the gas phases within the porous particle. The processes in the ›kci l 
2Deff ¼ bi ðkcR;i l 2 c1;i Þ ð5Þ
particle are influenced by diffusion and convection associated ›r r¼R
with chemical reactions. During drying and pyrolysis, the
where kcR,il refers to the molar concentration of the species i
convective terms play an important role, because of the
at the surface and c1;i indicates concentration in the bulk
humidity and the volatiles content of the wood, whereas char
flow. The calculation of the mass transfer coefficient bases
combustion and gasification are determined by diffusion. For
on the Sherwood number calculated in the same way as the
the effective thermal conductivity the contribution of
Nusselt number in Eq. (4). Gas flow out of the particle
radiation in the pores of the particle according to Ref. [12]
reduces the transfer coefficients; thus, Stefan’s correction is
is included additionally to the property data.
introduced to the calculation of the coefficients.
The solution of the momentum equation to calculate the
Chemical reactions. The efficient description of the
flow velocity of the gases in the porous particle would exceed
chemical source term requires a general approach, able to
the numerical effort acceptable for a single particle as part of a
treat pyrolysis, char conversion or homogeneous reactions.
packed-bed model. Therefore, a simplified approach predict-
Each conversion process consists of one or more reactions
ing the flow velocity is applied. The flow within the particle is
describing the degradation from the educts to the products.
assumed to be determined by the gas production rate due to
Each reaction then is represented by a maximum number of
solids conversion. Thus, the continuity equation can be applied
three educts and three products, with the temperature
to predict the velocity, always assuming an outward flow from
dependence of the rate for reaction i described by the
the particle. Under the assumption that the change of the
Arrhenius equation in the following form, Eq. (6).
source terms vG due to gas production at different positions in

 3
the particle is much slower than the propagation of the pressure E Y
wave in the solid, the flow velocity on the boundary face of a r_i ¼ k0 exp 2 a c ð6Þ
RT k¼1 i;k
volume element can be calculated by Eq. (2), where vG,j stands
for the gas source terms in the inner elements with respect to The computer program allows different models for the
the boundary face of the volume element i. conversion of wood or char to be selected from a database
X
n [16] supplying the corresponding set of chemical reactions
vG;j which occur in a particle. This approach allows the influence
j¼1
ui ¼ ð2Þ of different models and parameters on the conversion to be
Ai;E krG;i l investigated easily. The database is linked to the code,
732 C. Bruch et al. / Fuel 82 (2003) 729–738

allowing the data to be changed without modifications of the the interpolation function with the new radius. Thus, the
computer code describing the actual problem. whole range of char combustion from reacting core mode to
As has been discussed in the literature, the drying process shrinking core mode is covered by the model. During drying
in wood can also be described by a heterogeneous reaction and pyrolysis the diameter is assumed to remain constant.
[3,17]. However, comparisons between experimental and Mathematical solution. In discretizing the governing
calculated data showed, that the pseudokinetic data would equations by a finite volume approach, the partial
have to be adapted to the given boundary conditions, which differential Eq. (1) are transferred into a system of ordinary
excludes this model from general use. Therefore, drying is differential equations (ODE). Because of the different time
solved via a constant evaporation temperature, on the scales, the resulting ODE system is stiff, requiring a
assumption that, above this temperature, all energy will be sophisticated integration method to assure stability and
consumed by the drying process as long as water remains in accuracy. As a result of several tests [19] a backward
a control volume of the particle. This approach guarantees differentiation formulae (BDF) proved to be reliable in this
that no water evaporates below the evaporation temperature instance. Due to the fact that the single particle model is
and temperature in the drying section remains constant. This combined with an overlying bed model the integration time
matches the experimental observation made by Ref. [18] for step is divided into outer steps for which the submodels,
wet coal particles. single particle and gas phase, are calculated consecutively
Heterogeneous reactions during char conversion can be and inner time steps of the BDF scheme for the particle
represented in the form of Eq. (6) or as intrinsic rates by model. This reduces the computational effort by shifting
replacing the molar concentration of the solid species by the calculation procedures to the explicit loop which do not
specific surface. Whether or not a reaction rate is handled as change the values of the single particle significantly within
an intrinsic rate is indicated by the database; so, no changes the inner steps. Here, flow velocities of the gases in the
need to be made in the computation code. More detailed particle, property data, the drying process, and particle
models for the description of heterogeneous reactions would shrinking are calculated within the explicit loop. This
exceed the numerical effort which is acceptable for the enables the explicit formulation of the ODE problem over
single-particle model. Moreover, more detailed models lead each outer time step which helps to increase stability and the
to a higher number of parameters which are either unknown convergence rate of the integration process. Moreover, this
or uncertain for char particles from wood pyrolysis, thus the approach allows to handle the delta function introduced into
accuracy of description would not be increased necessarily. the system of differential equations to calculate the drying
Homogeneous reactions within the particle are neglected rate at a constant evaporation temperature, as mentioned
for the cases presented here. Calculations showed, that in above.
the diameter and temperature ranges of the single particle
experiments, less than 10% of the primary tars are converted
in the particle by secondary reactions and that char 2.2. Packed-bed model
combustion is limited by the amount of oxygen on the
particle surface. A packed bed in the current application is considered to
Particle shrinking. While wood contains only a negli- consist of a finite number of particles with the above-
gible amount of ash, fuel particles shrink during char mentioned particle model applied to each particle. The
combustion, which represents a moving boundary for the particle is coupled to the surrounding gas phase by source
solution of the transport equations. Therefore, they are terms, e.g. heat and mass transfer. For the presented case,
solved in a dimensionless form, allowing an adaptation of particles are assumed spherical as regarding discretization
the discretization mesh onto the actual particle size. An and solution of the transport equations.
interpolation function is used to approximate the discrete Due to the arrangement of particles in a packed bed, they
solution values to obtain the continuous solution versus form a void space, through which the combustion air flows.
particle radius. The distribution of the different scalars fi is The flow through the bed of particles can be described as
characterised by a front moving through the particle. Thus, flow through a porous media with the solid volume taken up
the following form of the interpolation function is suggested by the particles and determining the locally varying
to capture the steep gradients porosity. The position of the solid particles may change
during the conversion process, because of depletion of solid
fi ðrÞ ¼ a0 þ a1 ½2 þ tanhða2 r þ a3 Þ 2 tanhða2 r þ a3 Þ and shrinkage of the particles. Even if coupling to a motion
model based on an overall sum total of forces [20] is
þ a4 ½2 þ tanhða5 r 2 a6 Þ 2 tanhða5 r þ a6 Þ ð7Þ
intended for the future, a simplified approach is used here.
The coefficients ai for the interpolation function (7) are Assuming a constant porosity and a certain number of
calculated by a least-squares method. If shrinking occurs neighbour particles, the bed mesh is filled up with the solid
after an integration step due to depletion of the solid in an particles, until the given porosity is reached for the bed
outer control volume, the mesh is adapted to the new radius cells. If shrinking of the particles occurs, the distribution of
and the discrete values on the mesh nodes are calculated by particles in the bed is recomputed. This implies a decreasing
 C. Bruch et al. / Fuel 82 (2003) 729–738 733

bed height with further shrinkage or depletion of the effective thermal conductivity is excluded, because the
particles and increasing porosity of the upper fluid cells. packed bed is not assumed as a homogeneous porous body.
A finite volume approach using a Cartesian mesh is In the current application the bed process is resolved by
applied to discretize the transport equations for species, single particle processes, which requires keeping track of
momentum and energy to which the solid phase contributes particle cell and particle –particle relations. It is required to
source terms due to heat and mass transfer. Within the voids predict interaction between the solid and the gas phases and
of the bed, diffusion processes are neglected relative to the the solid particles, respectively. As wood contains only a
convective transport. Thus the change of the mass fraction small amount of ash, the solid will be fully depleted without
Yi for the species i can be determined by the divergence of any remaining residue. A particle diameter of 0.5 mm is
the species field and the source term due to the interaction chosen to be the minimum particle diameter appearing in the
with the solid phase and reactions in the void space of the bed. If a particle reaches the minimum diameter it is
bed as follows Eq. (8). The source term vm is averaged over removed from the computation.
the outer integration step which is chosen for the flow field The connectivity between neighbour particles in a bed
model. allows the heat flux due to conduction between two particles
›rG Yi   to be calculated in the following form Eq. (13).
þ 7 rG vG ¼ vm ð8Þ
›t 1 ›T 1 TN1 2 TN2
q_ 12 ¼ 2 ¼2 ð13Þ
Because of the coupling between pressure field and velocity 1 1 ›r 1 1 DrN1 2 DrN2
þ þ
field, the gas phase requires simultaneous solution of the l1 l2 l1 l2
mass and momentum equations to satisfy both balances. An The temperature gradient between two particles is approxi-
approach based on the well-known SIMPLER algorithm mated by the temperature difference between the outer shell
[21] is applied to solve the momentum Eq. (9) in the porous values of the particles and the distance of the discretization
bed. For the pressure field the ideal gas law is assumed. The points as indicated in Fig. 1b. As shown in Fig. 1c the
pressure loss is estimated by Darcy’s law, based on the contact area is assumed quadratic, which is determined by
phase averaged velocity. the contact angle g. Hence, the contact area is given by
› m
ðr v Þ þ 7ðrG vG vG Þ ¼ 27p 2 vG ð9Þ Ac ¼ 1
ððR1 tan gÞ2 þ ðR2 tan gÞ2 Þ ð14Þ
›t G G K 2

Due to the dominance of convective transport, conduction However, for solid fuels with low thermal conductivity,
within the gas phase is neglected in the energy Eq. (10). such as wood, conduction between solid particles only
Source terms occur because of the convective heat and mass contributes to a small extent to the heat transfer in a packed
transfer between the solid and the gas phase and chemical bed.
reactions within the voids. Heat transfer due to radiation between the solid particles
is a complex interaction of absorption, reflection and
›
ðr c TÞ þ 7ðrG vG cv TÞ emission. Each particle emits a certain amount of heat to
›t G v its surroundings depending on its surface temperature, area
¼ 27pvG þ vcon þ vch þ vth ð10Þ and emission coefficient. The radiative heat loss of each

The source term due to convection between the solid phase

and the gas phase is calculated as the sum total of
convection terms over all n particles in a cell i Eq. (11),
where the transfer coefficients are computed as stated above.
Xn
1 ðDt
vcon ¼ ak Ak ðT 2 TG;i Þdt ð11Þ
k¼1
Dt 0 G;k

The sensible heat which is transported with the mass of m

species between n solid particles in a volume element and
the surrounding gas phase is calculated according to Eq.
(12)
Xn X m
1 ð
vth ¼ _ j;k TS;k Þdt
ðcv m ð12Þ
k¼1 j¼1
Dt

Moreover, heat transfer due to conduction and radiation

between the different particles occurs within a packed bed.
This interaction has to be accounted for as a source term in
the particle model. An approach to treat conduction by an Fig. 1. Modelling of heat conduction in a packed bed of solid particles.
734 C. Bruch et al. / Fuel 82 (2003) 729–738

Fig. 2. Modelling of radiation in a packed bed of solid particles.

particle is distributed between n neighbour particles Fig. 4 shows comparisons of experimental and predicted
according to the surface area weighted ratio, results for pyrolysis and char combustion of a single
Ai particle. Mass loss during pyrolysis was calculated by a
Fi!j ¼ X
n ð15Þ simple single step pyrolysis model. The property data and
Aj kinetic data can be taken from Table 1. As the temperature
j¼1 during experiments is quite low while the oxygen content in
In order to account for reflection and transfer to particles not the surrounding gas phase is high, only char conversion due
in direct contact with the emitting particle, the emanating to reaction with oxygen is taken into account in the model.
radiation contributes to a radiation field related to the mesh The temperature does not reach a sufficiently high value for
of the flow field model, as shown in Fig. 2. The radiation gasification reactions with water vapour of carbon dioxide
terms of all particles in the bed result in a global radiation to start. Kinetic data on char combustion with oxygen are
field to which the particles are exposed. given as intrinsic rate and can also be taken from Table 1.
The radiation field also accounts for radiative heat Mass loss curves during pyrolysis are shown in Fig. 4a
transfer between the bed surface and the surroundings. Due for different diameters of the dry beech wood particles at a
to short ray travelling distances in the voids of the packed reaction temperature of Th ¼ 743 K. With increasing
bed, adsorption, emission and scattering of the gas phase is diameter of the particles, the mass loss rate decreases
neglected. slightly. This indicates that, under the given conditions, the
progression of pyrolysis depends on propagation of the
heating front into the solid. The dependence on internal
3. Results and discussion thermal resistance is captured well by the one-dimensional

To study conversion of a single wood particle, a test

reactor was built up, Fig. 3. The cylindrical reactor, with an
inner diameter of 50 mm, consists of a heated (reaction) zone
(1) and a cooled part (2), where the particle rests until the
reactor is heated to the desired temperature. When the
temperature is reached, the particle is exposed to the reaction
zone. The reactor temperature was kept constant during an
experiment. To prevent oxygen from penetrating into the
reactor during pyrolysis experiments, a nitrogen flow
through the reactor is established (3). In char combustion,
pure air was used as the carrier gas through the reactor. The
particle is attached to a sample holder suspended from a
precision balance (4). For each experiment, the mass loss of
the particle and the evolution of the main gas species are
recorded. Beech wood spheres of different diameters, with
initial moisture content of 0, 30 and 67%, were used.
Experiments were performed at various temperatures,
diameters and moisture contents [22]. The mixture of carrier
gas and production gas is extracted from the bottom of the Fig. 3. Laboratory reactor for the conversion of a single fuel particle (1-
reactor through a gas cleaning section (5) to prevent tar from heating, 2-cooling chamber, 3-carrier gas, 4-precision balance, 5-gas
entering the measuring devices. cleaning section).
 C. Bruch et al. / Fuel 82 (2003) 729–738 735

Fig. 4. Comparison of theoretical (solid lines) and experimental (symbols) data for a spherical beech wood particle in case of pyrolysis of dry wood (a,b),
pyrolysis of wet wood (c) and combustion of wood char in air (d).

approach. Thus, the calculated values match the exper- shown in Fig. 4c. The drying period of the wet wood is
imental results with sufficient accuracy. described well by the model based on a constant evaporation
For different reactor temperatures at a constant particle temperature, resulting in a steep drying front moving into
diameter of dpa ¼ 17 mm, comparisons between exper- the particle at a face velocity of roughly 2.0 £ 1025 and
imental data and calculated values are shown in Fig. 4b. The 3.5 £ 1025 m/s at a water content of 67 and 33%.
theoretical values are in good agreement with the exper- Comparison with a drying model based on a heterogeneous
imental data at reactor temperatures of Th ¼ 743 and reaction with pseudokinetic data taken from Ref. [17],
Th ¼ 773 K. At Th ¼ 673 K, calculations show an earlier showed an overestimation of the drying time and an
start of pyrolysis than in the experiments. The difference overlapping with the pyrolysis process, which cannot be
between the experimental and the calculated values can be seen from the experiments. Assuming a constant evapor-
attributed to the drop in reactor temperature when the (cold) ation temperature, drying and pyrolysis take place consecu-
particle was transferred into the reaction zone. At higher tively for the given particle diameters. However, for both
temperatures, this drop was balanced out faster by the water contents, the rate of pyrolysis is underestimated,
energy stored in reactor material.
Table 1
Even if a simple pyrolysis model is used, the agreement Property and kinetic data for wood and wood char used in calculation runs
between theoretical and experimental data is satisfying.
Thus, applying kinetic data derived at dust samples to larger Beech wood Wood char
particles seems to be feasible by including description of the
cp (W/m2 K) 1500 þ T [12] 420:0 þ 2:09T þ 6:85 £ 1024 T 2 [12]
heat transfer. For property and kinetic data taken from
l (W/m K) 0.35 [12] 0.1046 [12]
literature no adaptation was necessary to match the results krl (kg/m3) 750 [24] 200 [23]
of dry wood particles. k0 (1/s), (m/s) 1.35 £ 109 [25] 3.01 £ 102 [26]
Pyrolysis of wet wood particles with 33 and 67% water Ea (kJ/mol) 123.1 [25] 149.38 [26]
Ai (m2/kg) – 2.71 £ 105 [27]
content, respectively, and a particle diameter of
dpa ¼ 17 mm at a reactor temperature Th ¼ 743 K is Intrinsic rate.
736 C. Bruch et al. / Fuel 82 (2003) 729–738

indicating that the calculated temperatures at the end of the derive a relationship between water content, temperature,
drying period are too low compared to the experimental particle size and changes in the solid matrix.
values. Even though agreement between experimental and For char combustion at Th ¼ 773 K in pure air, mass loss
theoretical data is satisfying for pyrolysis of dry wood, the curves for two different diameters of the particles are
same property and kinetic data cannot be applied to wet presented in Fig. 4d. In the computation runs, char is
wood. This may be explained by fractures and fissures assumed to consist of pure carbon. The char was generated
within the solid which appear as a result of rapid by pyrolysis of the beech wood particles. Even at a
evaporation during the drying period. In case of the relatively low temperature, char combustion is completely
experiments with dry wood samples, the virgin wood was boundary layer controlled, indicating the high reactivity
dried at 120 8C for 24 h. The lower heating rate compared to which is typical of wood char. Under these conditions, char
the pyrolysis of wet wood does not enhance changes in the conversion can be described correctly by a shrinking-core
solid matrix. Thus, pyrolysis cannot be treated separately model, resulting in a linear decrease of the particle radius
from the drying process before. The difference between with conversion, which was observed also for the calcu-
experimental and theoretical mass loss histories increases lations with the one-dimensional model. Agreement
with increasing water content. The change in the solid between the calculated and the experimental data is very
matrix will be influenced by several parameters, such as good, showing that the mass transfer coefficient and reaction
particle size, water content and temperature. This indicates rate are described sufficiently accurate. Due to the boundary
that within a packed-bed model, which should be applied to layer control of the process, the uncertainty of the kinetic
thermal conversion of wet biomass, a possibility to consider data has no influence on the conversion of the solid. For char
this influence is required. Due to the varying boundary particles derived with wood at different water contents of
conditions each particle may experience in a packed bed, the virgin material, no major difference in mass loss rates
separate treatment of each fuel particle as realised in the was observed during the experiments for the same particle
given model allows to take these parameters into account. diameter and reactor temperature as consequence of the
However, further experimental investigation is needed to boundary layer control of the process. Thus, for modelling

Fig. 5. Bed mass and height versus conversion time and mass loss history of monitor particles for a bed of beech wood particles with a heat flux of q_ ¼ 50 and
q_ ¼ 100 kW=m2 introduced from the top.
 C. Bruch et al. / Fuel 82 (2003) 729–738 737

Fig. 6. Bed mass versus conversion time and average conversion rate versus primary air flow for a bed of beech wood particles exposed to a heat flux of
q_ ¼ 100 kW=m2 :

conversion of a packed bed of char particles with air under height has a pyrolysis period roughly twice as long as for the
the given conditions the heat and mass transfer within each upper particle. Char conversion does not start until bed
solid particle need not to be described separately allowing height is reduced to half of its initial value, when the particle
the application of simpler models such as given by Ref. [1]. is fully exposed to the heat flux from above, leading to a
However, in a model which can be applied to the whole char conversion rate comparable to the particle originally
conversion process of biomass, transport processes within located at the top of the bed Fig. 6.
the particles cannot be neglected. For the parameters selected, combustion takes place
With the same property data and kinetic data as for the mainly in the upper section and the temperature in the lower
single-particle problem discussed above, results for a one- layers of the bed does not reach a sufficiently high level for
dimensional pile of beech wood particles are shown in char combustion to start. This results in a linear decrease of
Fig. 5. The bed height is chosen to be 30 cm, and the initial bed height with time which is qualitatively in good
particle diameter is dpa ¼ 15 mm for all particles. Primary agreement with experimental results of Ref. [9].
air at an inlet velocity of 0.2 m/s is supplied at the bottom,
and a radiative heat flux of 50 and 100 kW/m2, respectively,
is applied to the top of the bed. Fig. 5 shows height and mass 4. Conclusions
of the bed versus the conversion time and mass loss histories
for two particles in the bed at a heat flux of 100 kW/m2. The A numerical approach to predict the thermal conversion
two particles are located in the upper cell and the middle cell of a packed bed of wood has been presented. Within the
of the bed mesh to monitor progression of the reaction front. method, the packed bed consists of a finite number of
Bed mass decreases linearly with time, resulting in a particles, whose individual processes are resolved by a
grate combustion rate of 75 kg/h m2 and 95 kg/h m2 for a particle model, applicable to each particle of a packed bed.
heat flux of 50 and 100 kW/m2, respectively. Even if only It comprises heating, drying, pyrolysis, gasification and
kinetic data derived from dust samples is used in the combustion on a one-dimensional and transient basis in
calculation, the values correspond to experimental results conjunction with the relevant reactions.
given by Ref. [6] for the combustion of wood particles on a However, the approach allows the need for macrokinetic
grate. Due to the linear decrease of mass the overall data to be reduced and provides a deeper insight into the
conversion times of particles at different locations within processes within a packed bed during combustion of a solid
the bed is comparable. This can also be seen from the mass fuel. Comparisons show good agreement between exper-
loss histories of wood and char of the two monitor particles. imental and theoretical results for the single particle model,
The particle initially located at the top of the bed is indicating that the one-dimensional approach captures the
pyrolysed immediately, due to the high heat flux in the main rate determining characteristics. Pyrolysis of wet
upper part of the bed. Depletion of the char material starts particles at high heating rates requires further investigation
after the particle has been fully pyrolysed. Thus no into the structural changes of the solid matrix.
overlapping of the processes can be observed for these The particle processes are coupled to the gas flow in the
parameters. Limitation of the process by boundary layer voids of a packed bed through heat and mass transfer. The
diffusion causes the major part of the overall conversion latter is described by a model to predict the flow of
time to be consumed by char combustion. Because of the combustion gas in the void space of a packed bed by solving
lower heating rate the particle located at half of the bed the conservation equations for mass, momentum and
738 C. Bruch et al. / Fuel 82 (2003) 729–738

energy. Even if no empirical correlation or specific data for [11] Miller RS, Bellan J. Analysis of reaction products and conversion time
a packed bed such as an effective thermal conductivity in the in the pyrolysis of cellulose and wood particles. Combust Sci Technol
1996;119:331–73.
bed is used, agreement of the results with values of
[12] Grønli M. A theoretical and experimental study of the thermal
experimental studies from the literature is satisfactory. conversion of biomass. PhD Thesis. NTNU, Trondheim; 1996.
Thus, the approach of the model to divide the packed bed [13] Saastamoinen J, Impola R. Drying of biomass particles in fixed and
into its particles seems feasible, to capture the main rate moving beds. Proceedings of the 10th International Drying Sym-
limiting parameters, allowing to take particle diameters and posium, Krakow; 30 July–2 August, 1996.
properties into account. However, further research is [14] Gnielinski V. Berechnung des Wärme- und Stoffaustauschs in
durchströmten ruhenden Schüttungen. Verfahrenstechnik 1982;
required to validate, e.g. the description of the heat transfer
16(1).
within the bed. [15] Schröder E. Bestimmung des Druckverlustes und des Wärmeüber-
ganges von gasdurchströmten Feststoffschüttungen in der Pantha
Anlage. Report 6373, Research Centre Karlsruhe; 1999.
Acknowledgements [16] Bruch C, Peters B, Nussbaumer T. Modellierung der Festbettver-
brennung am Beispiel des Brennstoffes Holz. German Flame Days,
This study is funded by the Swiss Federal Office of Dresden; 14–15 September 1999. p. 139 –44.
[17] Chan WR, Kelbon M, Krieger BB. Modelling and experimental
Energy. verification of physical and chemical processes during pyrolysis of a
large biomass particle. Fuel 1985;64:1505 –13.
[18] Heidenreich C, Zhang D. Measuring the temperature response of large
References wet coal particles during heating. Fuel 1999;78:991– 4.
[19] Peters B, Bruch C. Evaluation of ODE-solvers for the prediction of
[1] Ford N, Cooke M, Sage P. Modelling of fixed bed combustion. Fuel thermal conversion of solid fuel particles. 16th IMACS World
Process Technol 1993;36:55–63. Congress, Lausanne; 21– 25 August 2000.
[2] Fatehi M, Kaviany M. Adiabatic reverse combustion in a packed bed. [20] Peters B. A detailed model for devolatilization and combustion of
Combust Flame 1994;99:1– 17. waste material in packed beds. Third European Conference on
[3] Raupenstrauch H. Ein Beitrag zur Computersimulation reagierender Industrial Furnaces and Boilers, Porto; 1995. p. 86–104.
Schüttschichten. PhD Thesis. Technical University Graz; 1990. [21] Patankar SV. Numerical heat transfer and fluid flow. New York:
[4] Hartner P. Entwicklung eines Computerprogramms zur eindimensio- McGraw-Hill; 1980.
nalen Simulation von heterogenen Festbett- und Vorschubreaktoren. [22] Schaeffer B, Wyrsch F. Diploma Thesis. ETH Zurich, Department of
PhD Thesis. Technical University Graz; 1996. Energy Technology; 2000.
[5] Beckmann M, Scholz R. Simplified mathematical model of combus- [23] Kansa EJ, Perlee HE, Chaiken RF. Mathematical model of wood
tion in stoker systems. Third Eur Conf Ind Furnaces Boilers 1995;18/ pyrolysis including internal forced convection. Combust Flame 1977;
21:61–70. 29:311–24.
[6] Kuo JT. Estimation of burning rates in solid waste combustion [24] Petek J. Experimentelle Untersuchung der Pyrolyse in inerter und
furnaces. Combust, Sci Technol 1998;137:1–29. reaktiver Atmosphäre unter den Bedingungen der Wurfbeschickung.
[7] Goh YR, Siddal RG, Nasserzadeh V, Zakaria R, Swithenbank J, PhD Thesis. Technical University Graz; 1998.
Lawrence D, Garrod N, Jones B. Mathematical modelling of the [25] Balci S, Dogu T, Yücel H. Pyrolysis kinetics of lignocellulosic
burning bed of a waste incinerator. J Inst Energy 1998;71:110– 8. materials. Ind Engng Chem Res 1993;32:2573– 9.
[8] Saastamoinen J, Haukka P. Simultaneous drying and pyrolysis in fixed
[26] Caram HS, Amundson NR. Diffusion and reaction in a stagnant
bed combustion of wet biomass. Proceedings of the 11th International
boundary layer about a carbon particle. Ind Engng Chem Fundam
Drying Symposium, Halkidiki; 19–22 August 1998. p. 1975–1982.
1977;16:171–81.
[9] Shin D, Choi S. The combustion of simulated waste particles in a fixed
[27] van den Aarsen FG, Beenackers A, van Swaaij W. Wood
bed. Combust Flame 2000;121:167–80.
pyrolysis and carbon dioxide char gasification kinetics in a
[10] Maschio G, Lucchesi A, Koufopanos C. Study of kinetic and transfer
fluidized bed. Fundam Thermo-chem Biomass Conversion 1985;
phenomena in the pyrolysis of biomass particles. Adv Thermochem
Biomass Conversion 1994;. conversion.

View publication stats