Chemical Engineering Science 64 (2009) 1036 -- 1045

Contents lists available at ScienceDirect

Chemical Engineering Science

CFD modelling of the fast pyrolysis of biomass in fluidised bed reactors. Part B Heat, momentum and mass transport in bubbling fluidised beds
K. Papadikis a , S. Gu b, , A.V. Bridgwater a
a b

School of Engineering and Applied Science, Aston University, Aston Triangle, Birmingham B4 7ET, UK School of Engineering Sciences, University of Southampton, Highfield, Southampton SO17 1BJ, UK

A R T I C L E

I N F O

A B S T R A C T

Article history: Received 3 June 2008 Received in revised form 10 September 2008 Accepted 7 November 2008 Available online 19 November 2008 Keywords: CFD Fluidised bed Fast pyrolysis Heat transfer Particle modelling

The fluidparticle interaction inside a 150 g/h fluidised bed reactor is modelled. The biomass particle is injected into the fluidised bed and the heat, momentum and mass transport from the fluidising gas and fluidised sand is modelled. The Eulerian approach is used to model the bubbling behaviour of the sand, which is treated as a continuum. Heat transfer from the bubbling bed to the discrete biomass particle, as well as biomass reaction kinetics are modelled according to the literature. The particle motion inside the reactor is computed using drag laws, dependent on the local volume fraction of each phase. FLUENT 6.2 has been used as the modelling framework of the simulations with the whole pyrolysis model incorporated in the form of user-defined function (UDF). The study completes the fast pyrolysis modelling in bubbling fluidised bed reactors. 2008 Elsevier Ltd. All rights reserved.

1. Introduction Pyrolysis is thermal degradation of organic materials in absence of oxygen. The primary pyrolysis products of biomass particles are usually referred to as condensable (tars) and noncondensable volatiles and char. The condensable volatiles (tars) are also often classified as liquids, and the noncondensable volatiles as gases mainly CO, CO2 , H2 and C1 C2 hydrocarbons (Di Blasi, 2000, 1996). The liquids are frequently subdivided into water and organics. Secondary reactions of the tars produced result in the formation of secondary degradation products, and generally reduced organics yields. The extent of secondary cracking depends in particular on the hot vapour residence time. Fluidised beds are the most widely used type of reactor for fast pyrolysis, as they offer a number of advantages, such as high heat transfer rates and good temperature control. The hydrodynamics of fluidised beds have been widely investigated, both experimentally and numerically, to allow validation of the model results. To date most of the computational research interest has been focused on the simulation of the fluidised bed hydrodynamics, using either the Eulerian (continuum) (Pain et al., 2002) or the Lagrangian (discrete element) (Apte et al., 2003; Kafui et al., 2002) model, as well as novel models like the one developed by Bokkers et al. (2006) which is based on the modelling of the larger bubbles as discrete elements

Corresponding author. Tel.: +44 023 8059 8520; fax: +44 023 8059 3230. E-mail address: [email protected] (S. Gu).

that are tracked individually during their rise through the emulsion phase, which is considered as a continuum. Due to the significant increase in computing power of recent years, these models have now made computational modelling of multiphase granular flows possible, though it is still very challenging, particularly so for industrial scale reactor units. The Eulerian formulation of the granular medium, using the kinetic theory of granular flows, has made the realisation of fluidised bed simulations, less computationally intensive. The particulate phase is treated as a continuum with an effective viscosity, and thus the method is also called two-fluid approach. The drag force on a single spherical particle has been widely studied e.g. Clift et al. (1978) and Bird et al. (2002). However, when a single particle moves in a dispersed two-phase flow, the drag force is affected by the surrounding particles. Correlations for calculating the momentum exchange coefficient of gassolid systems have been reported in the literature, such as the models of Syamlal and O'Brien (1989), Gidaspow (1994), and Wen and Yu (1966). Compared to two-phase flows that have been widely studied (Ishii and Zuber, 1978; Ishii and Chawla, 1979; Tomiyama, 1998; Kolev, 2005), the information for bubble three-phase flows is limited. The calculation of drag forces on particles that are part of a solid/liquid/gas mixture is a more complicated case and certain assumptions have to be made. Kolev (2005) analyses the bubble three-phase flow by making the assumption that the solid particle is carried by the liquid or the gas alone, depending on the volume fraction of each one of the continuous phases. His approach is adopted for the implementation of the model described in this paper, concerning

0009-2509/$ - see front matter 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.ces.2008.11.007

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

1037

the injection of a single biomass particle inside the bubbling bed of the reactor. For the momentum transport analysis and diagrams the reader is referred to the first part of this study by Papadikis et al. (2008). The bed to surface heat transfer has been widely studied through the years for different cases, using different approaches (Kunii and Levenspiel, 1991; Andeen and Glicksman, 1976). Boterill (1975) and Yates (1983) studied the heat transfer between a bubbling fluidised bed of smaller particles to a fixed cooling tube or a stationary wall. Agarwal (1991) studied the heat transfer to a large freely moving particle in gas fluidised bed of smaller particles and Parmar and Hayhurst (2002) tried to measure the heat transfer coefficient for freely moving phosphor bronze spheres (diam. 28 mm) around a bed of hot sand fluidised by air. Also, studies on the influence of bed particle size to the heat transfer coefficient have been performed (Collier et al., 2004) and a Nusselt number based on the thermal conductivity of the gas was derived for (U Um f ), while for (U > Um f ) the Nusselt number appeared to have a constant value. Numerical investigations have also been performed using the Eulerian approach (Schmidt and Renz, 1999, 2000) and the results showed a strong coupling between the local solid volume distribution and the heat transfer coefficients. They also showed a good agreement with the penetration theory (Kuipers et al., 1992) as well as the influence of the rising bubbles. In the current study the heat, mass and momentum transport from the bubbling bed and through a solid biomass particle is modelled. The paper comprises the second part of the computational model for the fast pyrolysis of biomass and extends the first part of this study by incorporating the heat transfer model and reaction kinetics of biomass in a user-defined function (UDF). In the first part the fluidised bed hydrodynamics were modelled using the Eulerian model included in FLUENT's code and the UDF used included the momentum transport equations, analysing the forces exerted on the discrete biomass particle when the latter is a part of a solid/liquid/gas mixture. The second part extends the existing momentum transport code and analyses the heat transfer from the fluidised bed as well as the biomass degradation based on reaction kinetics from the literature. The reaction mechanism used takes into account the intra-particle secondary reactions which result in increased char and gas formation and reduced bio-oil yields. The current study (Parts A and B) is the first work of this kind. It combines the bed hydrodynamics and fast pyrolysis process in a unified CFD model. It also provides a good insight to the heat and mass transfer in fast pyrolysis, and their intricate interaction with reaction kinetics and phase transition phenomena. 2. Model description The 150 g/h fast pyrolysis lab scale reactor of Aston University is illustrated in Fig. 1. The shape of the reactor is circular with dimensions as shown in Fig. 1 and its walls are heated in the area where the sand bed stands. Nitrogen flows through a porous plate at the bottom of the reactor at a velocity of U0 = 0.3 m/s. The superficial velocity is approximately 4 times greater than the minimum fluidising velocity Umf of the reactor, which is typically around 0.08 m/s using a sand bed with average particle diameter of 440 m Geldart (1973) B Group. The particle is injected at the centre of the sand bed which has been previously fluidised for 1 s. Momentum is transferred from the bubbling bed to the biomass particle as well as from the formed bubbles inside the bed. According to Bridgwater (1999), the most appropriate biomass particles sizes for liquid fuel production lie in the range of 0.010.6 102 m with temperature between 700 and 800 K. The studied biomass particle is chosen to be 0.05 102 m

Reactor Width 40 mm

Reactor Outlet (Pressure Outlet) Reactor Height 260 mm

Static Sand Bed Height 80 mm (Wall Thermal Boundary Condition)

Nitrogen Flow (Velocity Inlet)

in diameter, which is more or less the size of the particles, due to feeding problems, for a small rig like the one studied in this paper. Bigger rigs and commercial plants use larger particles in the range of 25 mm. The scope of the simulation is to determine the correct heat and momentum transport inside the reactor and inside the biomass particle. When the particle is injected inside the reactor, it can either be inside a bubble or inside the packed bed. The code will be able to identify the regime of interest, depending on the local volume fraction of the two continuous phases, and calculate the correct drag, buoyancy and virtual mass forces according to the state, as well as the correct bed to surface heat transfer coefficient. Regarding the extensive study of momentum transport from the bubbling bed to the particle the reader is referred to the first part of this study (Papadikis et al., 2008). Depending on the regime of interest the heat transfer on the surface of the particle will be computed according to the literature. When the particle is carried only by the fluidising gas the wellknown RanzMarshall (1952a,b) correlation is used, while, when the particle is found inside the bubbling bed the heat transfer coefficient is calculated according to the findings of Kuipers et al. (1992). The reaction kinetics is based on a two-stage, semi-global mechanism, with kinetic constants suitable for wood pyrolysis according to Chan et al. (1985) for the primary stage and Liden et al. (1988) and Di Blasi (1993) for the secondary reactions. This scheme has been chosen for this study because it can predict the correct behaviour of wood pyrolysis including the dependence of the product yields on temperature (Di Blasi, 1996, 1998; Chan et al., 1985). 3. Computational model The equations of motion that describe the hydrodynamics of the fluidised bed which are already included in FLUENT's code, together with the UDF extension regarding the momentum transport from the fluidised bed, are extensively analysed in the first part of this study (Papadikis et al., 2008). The equations that describe the additional extension of the UDF to simulate the heat transfer and reaction kinetics of biomass pyrolysis follow.

1038

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

k3

CHAR k4 GAS

dV Char

k2 WOOD k1 GAS
Fig. 2. Two-stage, semi-global model.

TAR

TAR

k5 CHAR

k=0

rk

k=N

dr

Biomass

Fig. 3. (Right) Particle discretisation and discrete volume generation, (left) char formation during pyrolysis.

3.1. Reaction kinetics The reaction kinetics of biomass pyrolysis is modelled using a two-stage, semi-global model (Chan et al., 1985; Di Blasi, 1996). The mechanism is illustrated in Fig. 2. The mechanism utilises the Arrhenius equation which is defined as Ki = Ai exp(Ei /RT). (1) penetration theory can be applied with the following mixture properties (Kuipers et al., 1992): hpen = with km = g kg + (1 g )ks and ( Cp )m = g g Cp,g + (1 g ) s Cp,s . (10) (9) km ( Cp )m t (8)

The values of the kinetic parameters were obtained by Chan et al. (1985) for the primary pyrolysis products, while the fourth and fifth reaction from Liden et al. (1988) and Di Blasi (1993), respectively. The model associates the reaction kinetics mechanism with the discrete biomass particle injected in the fluidised bed. The particle's properties change according to the reaction mechanism due to the phase transition phenomena. Momentum and heat transfer on the particle are calculated according to the new condition in the UDF, and the variables associated with it updated in each time-step. Intraparticle secondary reactions due to the catalytic effect of char are taken into account, resulting on secondary vapour cracking. 3.2. Heat transfer The heat conduction along the radius of the particle is calculated by solving the heat diffusion equation for an isotropic particle

j jT j 1 j ( Cp eff T) = keff r2 + ( H) . jt jr jt r 2 jr
The boundary condition at the surface of the particle is given by h(T Ts ) = keff

The density of the particle is calculated as the sum of all the discrete densities that have been produced from the discretisation of the particle as it is illustrated in Fig. 3. The radius of the particle is discretised to N number of grid points numbered from k = 0 to N, where 0 is the centre of the particle, generating N discrete volumes. The density distribution along the radius of the particle is calculated according to the discrete masses of the solid phases (wood and char) that correspond to the specific discrete volume. The discrete masses of the solid phases are calculated using a linear approximation between two neighbouring points that form a discrete volume, according to the mass fraction in time of the specific phase. The discrete volumes are given by
3 3 dV k = 4 (rk rk1 ) for k = 1 . . . N, 3

(2)

(11)

where (3) rk1 = rk dr. The discrete mass dmk is calculated by dmk = ( kw + k1w ) + ( kc + k1c ) mp 2 N for k = 1 . . . N. (13) (12)

jT jr

r=R

and at the centre of the particle

jT jr

= 0.
r=0

(4)

The effective thermal conductivity keff and effective specific heat capacity Cpeff are given by keff = kc + |kw kc | w , Cpeff = Cpc + |Cpw Cpc | w . (5) (6)

Therefore, the discrete particle densities d p along its radius are k given by dmk d p = k dV k and the total average particle density is
pav =

(14)

The heat transfer coefficient is evaluated from the well-known Ranz and Marshall (1952a,b) correlation, when the particle is carried only by the fluidising gas Nu = hdp 1/2 = 2.0 + 0.6 Red Pr1/3 . k (7)

1 N

N k=1

d p . k

(15)

4. Results and discussions Table 1 shows the simulation parameters. FLUENT was used as the basic modelling platform to simulate the fluidised bed hydrodynamics. However, all the results for momentum transport (discussed

However, assuming that conductive heat transfer is dominant in the near-particle-region, when the particle is inside the bed, the

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045 Table 1 Simulation parameters Property Biomass density, w Biomass particle diameter, dp Biomass specific heat capacity, Cpw Char specific heat capacity, Cpc Biomass thermal conductivity, kw Char thermal conductivity, kc Superficial velocity, U0 Gas density, g Gas viscosity, g Gas specific heat capacity, Cp,g Gas thermal conductivity, kg Solids particle density, s Sand specific heat capacity, Cp,s Sand thermal conductivity, ks Mean solids particle diameter, ds Restitution coefficient, ess Initial solids packing, s Static bed height Bed width Heat of reaction Value 700 kg/m3 0.05 102 m 1500 J/kg K 1100 J/kg K 0.105 W/m K 0.071 W/m K 0.3 m/s 0.456 kg/m3 3.44 105 kg/m s 1091.6 J/kg K 0.0563 W/m K 2500 kg/m3 835 J/kg K 0.35 W/m K 440 m 0.9 0.63 0.08 m 0.04 m H = 255 kJ/kg Comment Wood Fixed Wood Char Wood Char 4Umf Nitrogen (773 K) Nitrogen (773 K) Nitrogen (773 K) Nitrogen (773 K) Sand Fixed Fixed Uniformdistribution Value in literature Fixed value Fixed value Fixed value Babu and Chaurasia (2003)

1039

Define standard values and constants Injection time? Yes Particle injection at t = 1 second Matrix initiation for temperature, products radial distribution No

Calculation of effective values, (i.e thermal conductivity, specific heat capacity etc.) t > 1 second

in Part A), heat transfer as well as reaction kinetics associated with phase transition phenomena in fast pyrolysis were achieved by a self-developed UDF. 4.1. Simulation sequence Fig. 4 illustrates the flow-chart of the extension code for FLUENT and it is called at the end of each time-step. If the simulation reaches the injection time, the UDF initialises all the fields and matrices that are going to be used for the calculations that follow. When the injection time is exceeded (indicated by No in the flow-chart) then the discrete phase variables used by FLUENT in its MACROS are being replaced by the ones calculated by the code. The new values are calculated by defining scalar updates, since the cell and the thread pointers in the specific MACRO correspond to the cell that the particle is currently in. Therefore, the local property values of the sand and nitrogen can be easily accessed by separating the solution domains for each phase in the UDF. The UDF is splitted in multiple parts in order to achieve more efficient code structure as well as convenient future modifications. The first part calculates the properties of the particle (density, effective thermal conductivity, effective specific heat capacity, etc.). Then, identifies the regime that the particle finds itself in (the particle can be either in the freeboard or inside a bubble or the sand bed). The code then moves into the discretisation of the particle in a number of radially distributed nodal points defined by the user. The heat transfer section calculates the radially distributed heat transfer and product yield distribution during the pyrolysis of the particle. A small loop is used to calculate the average properties of the reacted particle (mass, density, temperature), as well as the mass, momentum and energy sources of the produced tars. Finally, the forces exerted on the particle are calculated and its velocity is defined for the next time-step. 4.2. Bed hydrodynamics and position of particle Fig. 5 illustrates the hydrodynamics of the fluidised bed at different simulation times with the particle position indicated by the black spot inside the contours. Until the 1st second of the simulation the bed is fluidised without the biomass particle injected in it. Since the sand has gained some velocity in the x and y direction, the biomass particle is injected and momentum is transferred from the

Identification of regime of interest for momentum, heat transfer (Particle is carried either by the sand or nitrogen).

Particle discritisation Radial temperature, pyrolysis products distribution

Average particle mass, density and temperature calculation

Definition of tar mass, momentum and energy sources

Drag, virtual mass gravity and boyancy force calculation according to regime of interest

End
Fig. 4. User defined function flow-chart.

fluidised sand to the particle. The velocity of the particle is calculated by integrating in time the equation of motion for discrete particles, as described in the first part of this study, and its new position is illustrated by the black spot inside the reactor. The particle's position in the reactor is a result of the heat transfer and phase change due to reaction effects. Different heat transfer rates will result in different biomass degradation rates and consequently different particle properties in time. The density drop of the particle (Fig. 6) will differ and the drag and virtual mass forces exerted on the particle will significantly change. The model can predict the particle position inside the reactor, as it is subjected to pyrolysis, taking all of these effects into account.

1040

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

6.29e-01 5.98e-01 5.67e-01 5.35e-01 5.04e-01 4.72e-01 4.41e-01 4.09e-01 3.78e-01 3.46e-01 3.15e-01 2.83e-01 2.52e-01 2.20e-01 1.89e-01 1.57e-01 1.26e-01 9.44e-02 6.29e-02 3.15e-02 0.00e+00

0.0

0.5

1.0

1.5 2.0 Time (S)

2.5

3.0

3.5

Fig. 5. Fluidised bed hydrodynamics with biomass particle position.

750 700 650 600 550 500 450 400 350 300 1 1.5 2 2.5 3 3.5

0.8 0.6
xcomponent ycomponent

Particle velocity (m/s)

0.4 0.2 0 0.2 0.4 0.6 0.8 1 1.5 2 2.5 3 3.5

Density (kg/m3)

Time (s)
Fig. 6. Average particle density.

Time (s)
Fig. 7. Particle velocity.

4.3. Particle dynamics The particle's velocity components are shown in Fig. 7. The particle cannot escape from the reactor in 2.5 s of simulation because the drag force from nitrogen in the freeboard is much lower than gravity. Its density has dropped to 300 kg/m3 (Fig. 6 at 2.5 s) and its terminal velocity is 0.75 m/s. As Fig. 8 shows, the average y-velocity component of nitrogen on the freeboard of the reactor ranges from 0.4 to 0.5 m/s. This means that nitrogen cannot provide the necessary drag to carry out of the reactor the pyrolysed particle and further degradation is necessary to reduce its density to the required levels. However, the assumption that the particle retains its initial volume is crucial. Shrinkage usually occurs during pyrolysis as well as char attrition would have a significant impact on the size of the particle. Also, the fact that the reactor was designed to pyrolyse cellulose particles is another factor. Cellulose pyrolysis is described by the BroidoShafizadeh scheme, where char yields are much lower ( 45%) and the density drop of the particle much higher. Thus, entrainment of char out of the reactor would be easier. 4.4. Evolution of volatiles Fig. 9 shows the evolution of tar during biomass degradation. The sandtar interaction is simulated using the Gidaspow drag function, while the nitrogentar interaction is modelled using the drag correlation of MorsiAlexander (1972), both included on FLUENT's main code. Fig. 10 shows the tar velocity distribution inside the bed. Modelling the flow of the evolved tars one can determine its residence time, which is a significant factor regarding the final bio-oil yield since tars are catalysed by char, reducing bio-oil conversion efficiency. In relation to biomass particle dynamics, the model can provide a good guide for reactor design especially on the aspect of charvolatiles interaction. 4.5. Heat transfer The radial temperature distribution is shown in Fig. 11. The surface plot shows the different temperature profiles inside the particle for different times during pyrolysis. The particle reaches the reactor temperature in 0.6 s and the exothermic nature of the reaction makes the internal points of the particle to reach a slightly higher temperature than that of the reactor. However, this raise in temperature is in the order of 0.52 K and it cannot be easily noticed in Fig. 11. Therefore, it can be considered insignificant and does not affect the product yields. The heat transfer coefficient (Fig. 12) ranges from 190 to 475 W/m2 K (average of 350 W/m2 K) which is a typical range

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

1041

1.60e+00 1.50e+00 1.40e+00 1.30e+00 1.20e+00 1.10e+00 1.00e+00 9.05e-01 8.07e-01 7.08e-01 6.09e-01 5.10e-01 4.12e-01 3.13e-01 2.14e-01 1.16e-01 1.68e-02 -8.19e-02 -1.81e-01 -2.79e-01 -3.78e-01 -4.77e-01 -5.76e-01 -6.74e-01 -7.73e-01 -8.72e-01

1.59e+00 1.49e+00 1.40e+00 1.30e+00 1.20e+00 1.10e+00 1.01e+00 9.09e-01 8.11e-01 7.14e-01 6.16e-01 5.19e-01 4.21e-01 3.24e-01 2.26e-01 1.29e-01 3.10e-02 -6.65e-02 -1.64e-01 -2.62e-01 -3.59e-01 -4.57e-01 -5.54e-01 -6.52e-01 -7.49e-01 -8.47e-01

1.28e+00 1.20e+00 1.11e+00 1.03e+00 9.40e-01 8.54e-01 7.68e-01 6.82e-01 5.96e-01 5.10e-01 4.24e-01 3.38e-01 2.53e-01 1.67e-01 8.07e-02 -5.17e-03 -9.11e-02 -1.77e-01 -2.63e-01 -3.49e-01 -4.35e-01 -5.21e-01 -6.06e-01 -6.92e-01 -7.78e-01 -8.64e-01

1.48e+00 1.39e+00 1.30e+00 1.21e+00 1.12e+00 1.03e+00 9.34e-01 8.43e-01 7.51e-01 6.60e-01 5.68e-01 4.77e-01 3.85e-01 2.93e-01 2.02e-01 1.10e-01 1.86e-02 -7.30e-02 -1.65e-01 -2.56e-01 -3.48e-01 -4.39e-01 -5.31e-01 -6.22e-01 -7.14e-01 -8.06e-01

2.28e+00 2.15e+00 2.03e+00 1.90e+00 1.77e+00 1.64e+00 1.52e+00 1.39e+00 1.26e+00 1.13e+00 1.00e+00 8.77e-01 7.50e-01 6.22e-01 4.94e-01 3.67e-01 2.39e-01 1.11e-01 -1.64e-02 -1.44e-01 -2.72e-01 -3.99e-01 -5.27e-01 -6.55e-01 -7.82e-01 -9.10e-01

1.61e+00 1.51e+00 1.41e+00 1.32e+00 1.22e+00 1.13e+00 1.03e+00 9.33e-01 8.36e-01 7.40e-01 6.44e-01 5.47e-01 4.51e-01 3.54e-01 2.58e-01 1.62e-01 6.54e-02 -3.10e-02 -1.27e-01 -2.24e-01 -3.20e-01 -4.16e-01 -5.13e-01 -6.09e-01 -7.06e-01 -8.02e-01

1.0

1.5

2.0

2.5

3.0

3.5

Fig. 8. Contours of y-velocity component of nitrogen.

2.70e-22 2.57e-22 2.43e-22 2.30e-22 2.16e-22 2.03e-22 1.89e-22 1.76e-22 1.62e-22 1.49e-22 1.35e-22 1.22e-22 1.08e-22 9.46e-23 8.11e-23 6.76e-23 5.41e-23 4.05e-23 2.70e-23 1.35e-23 0.00e+00

5.20e-05 4.94e-05 4.68e-05 4.42e-05 4.16e-05 3.90e-05 3.64e-05 3.38e-05 3.12e-05 2.86e-05 2.60e-05 2.34e-05 2.08e-05 1.82e-05 1.56e-05 1.30e-05 1.04e-05 7.81e-06 5.20e-06 2.60e-06 0.00e+00

2.40e-04 2.28e-04 2.16e-04 2.04e-04 1.92e-04 1.80e-04 1.68e-04 1.56e-04 1.44e-04 1.32e-04 1.20e-04 1.08e-04 9.60e-05 8.40e-05 7.20e-05 6.00e-05 4.80e-05 3.60e-05 2.40e-05 1.20e-05 0.00e+00

5.05e-04 4.80e-04 4.55e-04 4.29e-04 4.04e-04 3.79e-04 3.54e-04 3.28e-04 3.03e-04 2.78e-04 2.53e-04 2.27e-04 2.02e-04 1.77e-04 1.52e-04 1.26e-04 1.01e-04 7.58e-05 5.05e-05 2.53e-05 0.00e+00

3.64e-04 3.46e-04 3.28e-04 3.09e-04 2.91e-04 2.73e-04 2.55e-04 2.37e-04 2.18e-04 2.00e-04 1.82e-04 1.64e-04 1.64e-04 1.27e-04 1.09e-04 9.10e-05 7.28e-05 5.46e-05 3.64e-05 1.82e-05 0.00e+00
1

3.44e-04 3.26e-04 3.09e-04 2.92e-04 2.75e-04 2.58e-04 2.41e-04 2.23e-04 2.06e-04 1.89e-04 1.72e-04 1.55e-04 1.37e-04 1.20e-04 1.03e-04 8.59e-05 6.87e-05 5.15e-05 3.44e-05 1.72e-05 0.00e+001

1.0

1.5

2.0 Time (S)

2.5

3.0

3.5

for fluidised bed reactors of Geldart B group particles (Basu, 2006). The model takes into account the degradation of the particle in the calculation of the convective and conductive heat transfer in each time-step. Due to phase change phenomena the properties of the particle change and the heat transferred is significantly affected as it is indicated in Eqs. (2) and (3). The radial temperature profile, however, seems to be relatively flat. This is due to the small biomass particle size which for an average heat transfer coefficient of h 350 W/m2 K has a Biot number of Bi 0.28. This results in almost radially uniform product yields as it is shown on the surface plots of Fig. 13. Biomass degradation starts at about 400 C, and as Fig. 11 shows the temperature gradient at that instant is not significant. The particle has reached an almost uniform temperature,

so the product distribution is radially uniform. The case of course would not be the same if a larger diameter particle was examined, where the Biot number is greater and the temperature gradient more significant. The UDF can be used to model such a case and gives the opportunity to the user to define the number of nodal points along the radius of the particle. 4.6. Product yields The simulation time of the particle is 2.5 s but we can still see 30% of wood unreacted (Fig. 13). The reason for simulating 2.5 s is because usual residence times for fast pyrolysis are close to 2 s. The particle needs almost 0.6 s to reach the reactor temperature, so the

1042

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

3.27e+00 3.11e+00 2.94e+00 2.78e+00 2.61e+00 2.45e+00 2.29e+00 2.12e+00 1.96e+00 1.80e+00 1.63e+00 1.47e+00 1.31e+00 1.14e+00 9.81e-01 8.17e-01 6.54e-01 4.90e-01 3.27e-01 1.63e-01 0.00e+00

2.5

3.0 Time (S)

3.5

Fig. 10. Vectors of velocity magnitude of tar inside the bed.

750

700 800 650 700 Temperature (K) 600 600 550 500 500 400 450 300 0 0.5 1 x 104 1.5 2 Radial distance (m) 2.5 1 1.2 Time (s) 1.4 1.6 350

400

Fig. 11. Radial temperature distribution.

particle is pyrolysed for almost 2 s. Also, the computational needs and outputs, led to this limited, but quite representative time of simulation. The final tar yield reached 45% by weight of wood, while the gases 10% and char 15%, respectively. The reason for the

relatively increased char formation and decreased tar production in the 70% of the wood that has reacted is the 0.6 s that the particle needs to be heated up to 773 K. At the early stages of the biomass injection the particle spends more than half a second in intermediate

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

1043

temperatures (303700 K) which favour the formation of char. In a process that lasts around to 2 s, this transitional stage from atmospheric to reactor temperature can significantly affect the final biooil yields. Typical fast pyrolysis yields on dry wood basis would recommend tar yields of 75%. However, the reaction kinetics scheme takes
500

Local heat transfer coefficient (W/m2K)

450 400 350 300 250 200 150 1 1.5 2 2.5 3 3.5

into account intra-particle tar cracking catalysed by char, which contributes to the higher amount of char and gas products. Also, complete degradation of the particle would result to higher amount of tar yields. According to Di Blasi (1998) the same reaction kinetics scheme, without taking into account the secondary reactions, would produce yields of 23% char, 13% gas and 64% tar, for complete (100%) biomass degradation at 773 K. Therefore, the yields produced in the simulation are pretty similar with those produced by the single particle models in the literature, since complete degradation of the particle would result to almost identical yields percentages. This indicates the correct implementation of the reaction kinetics in the UDF associated with the discrete particle. 4.7. Model advantages and drawbacks A complete fast pyrolysis model in bubbling fluidised bed reactors would have certain advantages over single particle models especially on the direction of reactor design. Simulating the complete process of fast pyrolysis, one can have a realistic view of what is happening inside the reactor. Reactor designs can be optimised for efficient char entrainment, something that would increase bio-oil yields. Tars are catalysed by char, lowering bio-oil yield efficiency. Bed hydrodynamics can be simulated for different reactor geometries and its effect on the heat transfer coefficient and biomass degradation can be quantified. Possible effects of the products to the bed hydrodynamics can also be noticed.

Time (s)
Fig. 12. Local heat transfer coefficient.

Wood, % by weight of wood

100 90 80 70 60 50 40 30 1 1.5 2 2.5 3 0.5 3.5 0 1 1.5 2 2.5

Tar, % by weight of wood

50 40 30 20 10 0 0 1 2.5 3 3.5

x 104

x 104 Radial distance (m )

2 1

1.5

Time (s)

Radial distance (m)

Time (s)

Gas, % by weight of wood

8 6 4 2 0 0 1

Char, % by weight of char

10

20 15 10 5 0 0 1

104

2 1

1.5

2.5

104

2 1

1.5

2.5

3.5

Radial distance (m)

Time (s)

Radial distance (m)

Time (s)

1044

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

However, the computational cost of such a simulation is high. The extensive code development and the extremely high amount of memory allocation slows down the simulations significantly. Especially in three-dimensional cases the mass, momentum and energy sources of a 3rd Eulerian phase (in this case the tars) would make things even worse. The UDF contains numerical algorithms for the solution of the heat diffusion equation inside the particle. The code uses the flow time-step for the time marching of heat flux in the surface of the particle and conduction along its radius. Both flow solution (performed by FLUENT) and UDF algorithm solution have to converge in a synchronous way. This sometimes can be a reason for significant reduction of the time-step if the Courant number is high, which means extreme computational delays due to small time-steps. Post-processing of the UDF output file can also be a problem. Using small time-steps (in the order of 104 105 ) results in large output files with thousands of data, for simulations in the order of seconds. However, such a situation can be relatively easily resolved by including small data-save loops, specifying the number of timesteps to save. 5. Conclusions The fast pyrolysis of biomass in bubbling fluidised bed reactor was modelled. Several useful observation and conclusions were derived. Modelling of multiphase flows combined with discrete particle tracking can be possible by incorporating external a UDF in a commercial CFD software. FLUENT 6.2 was used as the modelling framework for the simulation with an extensive computational sub-routine to account for the heat and momentum transport from the fluidised bed to the biomass particle, as well as the prediction of the radial distribution of temperature and product yields. This was done by associating reaction kinetics of biomass with the properties of the moving discrete biomass particle. It was found that the model can predict the residence time of vapours and biomass particle. The properties of the pyrolysing particle can be predicted since the model accounts for its degradation which has an immediate effect on its trajectory inside the reactor. The model can be a useful guide for reactor design where char entrainment is of major importance. It can be easily modified and simplified for computational convenience if all the information of biomass pyrolysis is not needed (i.e. not to include evolution of tars if only char formation and entrainment is necessary, etc.). The first part (Papadikis et al., 2008) and the current paper complete the study on the numerical modelling of the fast pyrolysis of biomass in fluidised bed reactors. Improvements and applications will follow to further investigate the applicability and flexibility of the model in the design and scale-up of fluidised bed reactors for biomass pyrolysis. The follow-up development will also examine the influence of the reactor geometry on char entrainment and bed hydrodynamics. Notation Ai Bi Cp di E h k m Nu Pr pre-exponential factor, 1/s Biot number, dimensionless specific heat capacity, J/kg K diameter, m activation energy, J/mol convective heat transfer coefficient, W/m2 K thermal conductivity, W/m K mass, kg Nusselt number Prandtl number

r R Re t T U0 V

radial coordinate, m universal gas constant, J/mol K Reynolds number, dimensionless time, s temperature, K superficial gas velocity, m/s volume, kg/m3

Greek symbols H
i

mass fraction heat of reaction, J/kg density, kg/m3

Subscripts av c d eff g i k m mf p s w average char droplet effective gas general index radial position mixture minimum fluidisation particle solids wood

Acknowledgements The authors gratefully acknowledge the financial support from the UK Department for Innovation, Universities & Skills on the grants: the UK-China Fellowships for Excellence and ICUK Proof of Concept Grant Award. References
Agarwal, P.K., 1991. Transport phenomena in multi-particle systems-IV. Heat transfer to a large freely moving particle in gas fluidized bed of smaller particles. Chemical Engineering Science 46, 11151127. Andeen, B.R., Glicksman L.R., 1976. Heat transfer to horizontal tubes in shallow fluidized beds. ASMEpaper 76-HT-67. Apte, S.V., Mahesh, K., Lundgren, T., 2003. A Eulerian Lagrangian model to simulate two-phase/particulate flows. Annual Research Briefs 2003, Centre for Turbulence Research. Babu, B.V., Chaurasia, A.S., 2003. Modeling for pyrolysis of solid particle: kinetics and heat transfer effects. Energy Conversion and Management 44, 22512275. Basu, P., 2006. Combustion and Gasification in Fluidized Beds. Taylor and Francis Group, LLC. Bird, R.B., Stewart, W.E., Lightfoot, E.N., 2002. Transport Phenomena. second ed. Wiley, New York. Bokkers, G.A., Laverman, J.A., Van Sint Annaland, M., Kuipers, J.A.M., 2006. Modelling of large-scale dense gassolid bubbling fluidised beds using a novel discrete bubble model. Chemical Engineering Science 61, 55905602. Boterill, J.S.M., 1975. Fluid-bed Heat Transfer. Academic Press, New York. Bridgwater, A.V., 1999. Principles and practice of biomass fast pyrolysis processes for liquids. Journal of Analytical and Applied Pyrolysis 51, 322. Chan, W.R., Kelbon, M., Krieger, B.B., 1985. Modelling and experimental verification of physical and chemical processes during pyrolysis of large biomass particle. Fuel 64, 15051513. Clift, R., Weber, M.E., Grace, J.R., 1978. Bubbles, Drops, and Particles. Academic Press, New York. Collier, A.P., Hayhurst, A.N., Richardson, J.L., Scott, S.A., 2004. The heat transfer coefficient between a particle and a bed (packed or fluidised) of much larger particles. Chemical Engineering Science 59, 46134620. Di Blasi, C., 1993. Analysis of convection and secondary reaction effect effects within porous solid fuels undergoing pyrolysis. Combustion Science Technology 90, 315339. Di Blasi, C., 1996. Heat, momentum and mass transport through a shrinking biomass particle exposed to thermal radiation. Chemical Engineering Science 51, 11211132. Di Blasi, C., 1998. Comparison of semi-global mechanisms for primary pyrolysis of lignocellulosic fuels. Journal of Analytical and Applied Pyrolysis 47, 4364. Di Blasi, C., 2000. Modelling the fast pyrolysis of cellulosic particles in fluid-bed reactors. Chemical Engineering Science 55, 59996013.

K. Papadikis et al. / Chemical Engineering Science 64 (2009) 1036 -- 1045

1045

Geldart, D., 1973. Types of gas fluidization. Powder Technology 7, 285. Gidaspow, D., 1994. Multiphase Flow and Fluidization: Continuum and Kinetic Theory Descriptions. Academic Press, New York. Ishii, M., Chawla, T.C., 1979. Local drag laws in dispersed two-phase flow. NUREG/CR1230, ANL-79-105. Ishii, M., Zuber, N., 1978. Relative motion and interfacial drag coefficient in dispersed two-phase flows of bubbles, drops and particles. Paper 56 a, A.I.Ch.E. 71st Annual Meeting, Miami. Kafui, K.D., Thornton, C., Adams, M.J., 2002. Discrete particle-continuum fluid modelling of gassolid fluidised beds. Chemical Engineering Science 57, 2395 2410. Kolev, N.I., 2005. Multiphase Flow Dynamics 2, Thermal and Mechanical Interactions. second ed. Springer, Berlin. Kuipers, J.A.M., Prins, W., Van Swaaij, W.P.M., 1992. Numerical calculation of wallto-bed heat-transfer coefficients in gas-fluidized beds. American Institute of Chemical Engineers Journal 38, 10791091. Kunii, D., Levenspiel, O., 1991. Fluidization Engineering. second ed. Butterworth, Heinemann, Liden, A.G., Berruti, F., Scott, D.S., 1988. A kinetic model for the production of liquids from the flash pyrolysis of biomass. Chemical Engineering Communications 65, 207221. Morsi, S.A., Alexander, A.J., 1972. An investigation of particle trajectories in twophase flow systems. Journal of Fluid Mechanics 55, 193208. Pain, C.C., Mansoorzadeh, S., Gomes, J.L.M., de Oliveira, C.R.E., 2002. A numerical investigation of bubbling gassolid fluidized bed dynamics in 2-D geometries. Powder Technology 128 (1), 5677.

Papadikis, K., Bridgwater, A.V., Gu, S., 2008. CFD modelling of the fast pyrolysis of biomass in fluidised bed reactors. Part A: Eulerian computation of momentum transport in bubbling fluidised beds. Chemical Engineering Science 63, 42184227. Parmar, M.S., Hayhurst, A.N., 2002. The heat transfer coefficient for a freely moving sphere in a bubbling fluidised bed. Chemical Engineering Science 57, 34853494. Ranz, W.E., Marshall, W.R., 1952a. Evaporation from drops, part I. Chemical Engineering Progress 48, 141146. Ranz, W.E., Marshall, W.R., 1952b. Evaporation from drops, part II. Chemical Engineering Progress 48, 173180. Schmidt, A., Renz, U., 1999. Eulerian computation of heat transfer in fluidized beds. Chemical Engineering Science 54, 55155522. Schmidt, A., Renz, U., 2000. Numerical prediction of heat transfer in fluidized beds by a kinetic theory of granular flows. International Journal of Thermal Sciences 39, 871885. Syamlal, M., O'Brien, T.J., 1989. Computer simulation of bubbles in a fluidized bed. A.I.Ch.E. Symposium Series 85, 2231. Tomiyama, A., 1998. Struggle with computational bubble dynamics. In: Third International Conference on Multiphase Flow, ICMF 98, Lyon, France. Wen, C.-Y., Yu, Y.H., 1966. Mechanics of fluidization. Chemical Engineering Progress Symposium Series 62, 100111. Yates, J.G., 1983. Fundamentals of Fluidised-bed Chemical Processes. Butterworths, London.