Sei sulla pagina 1di 41

Modeling of Bubble Flow in Fluidized Bed

Combustion Units
Master of Science Thesis


JOHANNA OLSSON


Department of Energy and Environment
Division of Energy Technology
CHALMERS UNIVERSITY OF TECHNOLOGY
Gteborg, Sweden, 2008
Report No. T2008-316
































Modeling of Bubble Flow in Fluidized Bed Combustion Units
Johanna Olsson

Johanna Olsson, 2008

Examensarbete vid Chalmers Tekniska Hgskola



Department of Energy and Environment
Division of Energy Technology
Chalmers University of Technology
SE-41296 Gteborg (Sweden)
Telephone: +46 (0)31 7721000

Tryckstlle
Gteborg, 2008
Modeling of Bubble Flow in Fluidized Bed Combustion Units

Johanna Olsson

Department of Energy and Environment
Division of Energy technology
Chalmers University of Technology

Abstract
The performance of a fluidized bed combustion (FBC) unit depends on the fuel distribution in
the boiler, which is a complex process. This work includes a survey of previous research and
available models for solids mixing in fluidized beds. The available semi-empirical models
used to describe the solids mixing and the modeling from first principle through
computational fluid dynamics (CFD) both give an unsatisfactory description of the fuel
distribution in large industrial FBC units. To bridge the gap between these two approaches, a
model based on plausible physical mechanisms is needed.

Recognizing the importance of bubbles in a fluidized bed, the aim of this work is the
development and implementation of a dynamic bubble flow model. In this work it is
recognized that the fluidization regime, hence the bubble flow pattern, depends on the ratio of
the pressure drop across the air distributor to that across the bed, not on the individual values.
This ratio determines the relative parts of gas flow in the throughflow phase and the emulsion-
only phase.

The first-step one-dimensional model implemented for the bubble flow in a fluidized bed
under an exploding bubble regime is unique in that it accounts for the influence of the ratio on
the local fluctuations in superficial gas velocity and pressure. It is shown that a high pressure
drop ratio results in severe fluctuations in the system, while a lower ratio results in more
moderate fluctuations, which is in agreement with experimental observations. Moderate
fluctuations indicate that the superficial gas velocities in the respective phases do not differ
much: the bubbles are smaller and the bed is probably not operated in the exploding bubble
regime.

Due to the restriction of the model, it is not possible to perform realistic simulations of cases
in which the pressure drop across the distributor dominates strongly over the pressure drop
across the bed.

Keywords: Fluidized bed combustion units; Modeling; Mixing; Bubble flow pattern

Table of Contents
1. INTRODUCTION .................................................................................................... 1
Scope and aim .................................................................................................................................................... 1
2. BACKGROUND ..................................................................................................... 3
2.1 Fundamentals of fluidization ......................................................................................................................... 3
Minimum fluidization and terminal velocity ...................................................................................................... 3
Pressure drop ...................................................................................................................................................... 4
Geldarts classification of particles .................................................................................................................... 4
The combustion application ............................................................................................................................... 5
2.2 Bubbles in fluidized beds ................................................................................................................................ 6
The two phase theory of fluidized beds .............................................................................................................. 6
Bubble shape, size and rise velocity ................................................................................................................... 7
Davidsons bubble model ................................................................................................................................... 7
Coalescence ........................................................................................................................................................ 8
The dynamic behavior of the gas fluidized bed and fluidization regimes .......................................................... 8
2.3 Solids transport mechanisms ....................................................................................................................... 10
Wake transport and gulf streaming................................................................................................................... 10
Splash zone ....................................................................................................................................................... 11
2.4 Models of gas flow and solids mixing .......................................................................................................... 11
The Counter-Current Back Mixing (CCBM) model ........................................................................................ 12
The Brownian dispersion model ....................................................................................................................... 12
CFD simulations ............................................................................................................................................... 13
3. THEORY ............................................................................................................... 14
3.1 The bubble flow model.................................................................................................................................. 14
3.2 Computational flow scheme ......................................................................................................................... 19
4. RESULTS AND DISCUSSION ............................................................................. 21
4.1 Reference case ............................................................................................................................................... 21
4.2 The influence of the characteristic constant of the air distributor and the bed height ........................... 23
5. CONCLUSIONS ................................................................................................... 28
6. FURTHER WORK ................................................................................................ 29
REFERENCES ......................................................................................................... 30

1
1. Introduction
The use of fluidization technology was first developed for chemical reactors, but its many
beneficial properties have also made it interesting for combustion applications. Although the
process was developed in the 1940s and introduced in combustion systems in the 1960s, the
processes taking place inside the bed are not yet fully understood or explicitly described.
During the past few decades much research was conducted in this area, multiple experimental
studies were made, and in recent years numerical simulations using computational fluid
dynamics (CFD) were also conducted. The findings of these decades have been summarized
in several articles, Lim et al. (1995) being one of them.

However, most of the research on solids mixing and fuel distribution does not have its
experimental basis in conditions corresponding to large-scale fluidized boiler units, as
discussed in Pallars (2005). The boiler dimensions, fluidization gas velocity, bed height-to-
width ratio and the ratio of the pressure drops across the distributor to that of the bed are only
a few of the characteristics that set fluidized bed combustion units for solid fuel apart from
other applications. One of the main features determining the performance of a fluidized boiler
is the fuel distribution in the bed, which is among the more complex phenomena to describe.

In a fluidized bed boiler, fuel makes up only a small percent of the total bed material (Lim and
Agarwal, 1994), which might suggest that if the motion of the bed material and its interaction
with the fuel particles were known, fuel distribution could be determined. However, obtaining
a quantitative description of the fuel distribution, with satisfactory agreement with
experimental data, is more complex than indicated, since solids mixing is coupled to gas
mixing, and vice versa, as schematized in Figure 1. Solids are present in three regions of a
circulating fluidized bed (CFB) riser; the bottom bed, the splash zone and the transport zone,
as shown by Johnsson and Leckner (1995). The movement of solids in all three regions of a
CFB riser is the result of the interaction between particles and gas. This interaction exhibit a
special behavior in the bottom bed, as bubbles and channeling of gas appear under the
operational conditions common to boilers. These phenomena distort the flow fields
throughout the riser, for both bubbling (BFB) and circulating beds (CFB).

Scope and aim
The ultimate goal of this work is to develop a model describing the fuel distribution (marked
with squared background in Figure1) in a gas fluidized bed boiler as a function of operating
conditions that are measurable and controllable. By recognizing the importance of bubbles in
fluidized beds, the aim of this work is to develop, as the first step, a bubble flow model
(marked with dotted background in Figure1). Semi-empirical models for solids mixing
proposed in the literature should be used with care outside of the domain for which they were
developed. A macroscopic model of the fuel mixing, based on plausible underlying physical
mechanisms, describing both the gas and solids flow patterns in freely bubbling beds, would
bridge the existing gap between empirically-fitted models and the under-developed CFD
modeling.

This work begins with a survey of previous research and available models in the literature for
the solids mixing; it then proceeds to present a first-step bubble flow model for large-scale
fluidized bed combustion (FBC) units, both bubbling and circulating. Since the mechanisms
governing the dynamics of a fluidized bed are not yet fully understood, the model will need,

2
at present, to rely on experimental data for the fitting of parameters. Care should be taken so
that the model is subject to empirical fitting as little as possible.


Figure 1: Mapping of the physical mechanisms governing the fuel distribution


Fuel distribution
Gas - Solids flow pattern Fuel properties
Transport zone
(only in CFB)
Bottom bed Splash zone
Particle
motion
Gas
velocity
field
Bubble flow
pattern
Particle
motion
Gas
velocity
field
Particle
motion
Gas
velocity
field

3
2. Background
Fluidization is the operation by which solid particles are transformed into a fluidlike state
through suspension in a gas or a liquid. (Kunii and Levenspiel, 1991.)

2.1 Fundamentals of fluidization
The first large-scaled, commercially significant use of fluidized beds was for the gasification
of powdered coal, invented in 1922. The early use of fluidized beds for catalytic cracking
processes expanded to applications outside of the petroleum industry in the 1940s, at which
point the use and development of the fluidization technology was limited by a lack of
understanding of the processes actually taking place inside the bed. In hope of finding a
combustion system suitable for low grade coal and oil fuels that could not be burnt in
conventional furnaces, fluidized bed combustion (FBC) was developed in the early 1960s, but
it was not commercialized until the early 1980s. At present, fluidization technology is widely
used in a broad range of applications, e.g. chemical and catalytic reactors, combustion and
gasification of solid fuel, and the drying of granular solids; it is now of interest to extend its
operational ranges and areas of application.
Minimum fluidization and terminal velocity
A gas introduced at a low flow rate, as an upward flow from the bottom of a fixed bed,
percolates through the voids in the bed. At a high enough gas flow rate, the gas friction force
developed counterbalances the gravity force of the particles. At this point, called the
minimum fluidization point, particles are suspended in the gas flow and the gas-solid
suspension acquires properties similar to those of a fluid, i.e. the bed becomes fluidized.
Minimum fluidization conditions are characterized by a certain gas velocity, u
mf,
and bed
voidage (gas volume fraction),
mf
.

At gas velocities beyond the minimum fluidization velocity, also known as minimum
bubbling velocity, u
mb
, is reached, at which gas bubbles and channeling of the gas flow
appear, as illustrated in Figure 2. At higher gas velocities, typical for bubbling fluidized beds
(BFB), agitation becomes more violent. As the gas velocity is increased further and bubbles
grow larger, the bed surface becomes strongly fluctuating and difficult to determine. Finally,
at gas velocities exceeding the terminal velocity of the solids, u
t
, bed material is carried out of
the riser by the gas. To ensure a roughly constant solids inventory in the riser, entrained
particles have to be externally recirculated by means of a cyclone or replaced. Under
conditions of significant external recirculation of solids, a circulating fluidized bed (CFB) is
established, as shown in Figure 2.


4

Figure 2: (a) Fixed bed: u < u
mf
; (b) Bubbling fluidized bed: u > u
mb
; and (c) Circulating fluidized bed: u > u
t


Pressure drop
In particulate media, the frictional pressure drop can be described by the Ergun equation:

2
3 2 3
2
) 1 ( 75 . 1 ) 1 ( 150
u
d
u
d L
P
p p
c
c
c
c
+

=
A
. (1)

The pressure drop is described by two terms representing the viscous and the inertial or
kinetic energy losses. According to the Ergun equation a pressure drop depends not only on
the superficial gas velocity but also on particle size and bed voidage. At minimum
fluidization, the pressure drop across the bed equals the static pressure of the bed. Once the
bed is fluidized, the time-averaged pressure drop is nearly independent of the gas flow rate.
This constancy in the pressure drop, as explained by Kunii and Levenspiel (1991), occurs
when the gas-solid phase is well-aerated and can deform easily without appreciable resistance,
thus the viscous pressure drop is rather small.
Geldarts classification of particles
The properties of the particles that make up the bed affect the fluidized bed behavior.
Therefore, particles have been classified by Geldart (1973) into four groups, which are
characterized by solids-gas density difference and mean particle size, as seen in Figure 3.


5

Figure 3: Geldarts classification of particles under ambient conditions (taken from Kunii and Levenspiel, 1991)

Most of the particles in fluidized beds used for commercial scale combustion of solid fuels
belong to group B which contains most of the material of medium-range size and density,
i.e. sand. For beds consisting of group B particles, the minimum bubbling and minimum
fluidization velocities coincide (Geldart, 1973). Most of the literature on the basic properties
of fluidized beds is derived from tall and narrow beds with particles belonging to group A or
small group B particles.
The combustion application
The original aim of developing fluidized bed combustion (FBC) was to obtain a technique
which could efficiently convert low grade coal and oil. In addition, FBC shows high fuel
flexibility, and biomass and waste are becoming the usual fuel in FBC units. Furthermore,
efficient co-combustion of different fuel types is one of the main advantages FBC offers.
Also, the ability to use bed additives to control environmentally hazardous emissions makes
FBC performance relatively insensitive to the quality of the fuel.

The bed is fluidized by primary air supplied through the air distributor at the bottom of the
riser. This primary air represents the main oxygen source for the combustion process. Fuel is
usually fed into the boiler from a chute. As fuel particles enter the furnace they heat up;
moisture and volatiles are released leaving solid char particles. Volatiles are consumed in the
freeboard or the cyclone, while char particles remain mostly in the bed.

The bed material in a fluidized bed combustor acts as a thermal flywheel, providing stability
to the system and making it less sensitive to changes in the fuel heating value. Violent
bubbling and rapid mixing in a fluidized bed leads to a good gas-solids contact, as well as a
rather homogeneous temperature field in the dense bed, which eases the control of the process
(Kunii and Levenspiel, 1991). Only a few percent (typically from 0.1 to 5%) of the bed
material of a fluidized bed for solid fuel combustion consists of fuel (Lim and
Agarwal, 1994).

Since combustion rates of both char and volatiles increase with oxygen concentration, it is of
the utmost importance that good gas-solids contact and gas mixing exist, i.e. that the bed is
well mixed. Fuel mixing is therefore a key phenomenon in the performance of FBC units. In

6
the vertical direction, good fuel mixing is important to ensure sufficient contact time between
fuel and oxygen. In the horizontal direction, a homogeneous fuel distribution is desirable for
both the burnout of the fuel and the optimization of the excess air ratio, thus minimizing the
operational cost. The extent to which the mixing of fuel is sufficient depends on the fuel
conversion time and the characteristic mixing length, as discussed in Pallars (2007).

Bubbles and gas channeling (i.e. gas flowing through the bed without interacting with the gas-
solids emulsion) affect the FBC efficiency negatively. Fuel distribution in the bed depends
mainly on the motion of the bed materials, but the gas flow also influences the distribution.
As commented by Lim and Agarwal (1994), a good understanding of the dynamic behavior of
the bed is therefore crucial for the optimization of the design of the air and fuel feed systems.

The aspect (height-to-width) ratio of the bottom bed in FBC units is often very low, i.e. the
bed is shallow. The bed temperature is often kept close to 850C. Due to operational costs, the
pressure drop over the distributor is often kept as small as possible; the distributor type most
commonly used in commercial large-scale beds is the tuyere or bubble cap type. Boilers are
usually operated at atmospheric or slightly below atmospheric pressure, and the fluidization
gas velocity is often one or two orders of magnitude higher than the minimum fluidization
velocity.

2.2 Bubbles in fluidized beds
Knowledge of the general behavior of a fluidized bed is insufficient for some purposes, e.g.
reaction kinetics and heat transfer depend on details of the gas-solids interaction in the bed.
Hence, a satisfactory treatment of these phenomena requires a reasonable model representing
the gas flow through the bed and its interaction with bed material. As a consequence, the
bubble size, rise velocity, shape, distribution, frequency and flow patterns are of key interest.

In the early stages of fluidization research, the analogy of fluidized bed bubbles with bubbles
in liquids was used. Despite several similarities between these, a major difference is the fact
that no mass transfer with the surrounding exists for bubbles in a liquid, whereas a bubble in a
fluidized bed is a local high-voidage region exchanging gas with its surrounding, as explained
by Davidson (1985). Nevertheless, the above-cited analogy combined with experiments on
single bubbles has been used as basis for several concepts and models developed for bubbling
beds.
The two phase theory of fluidized beds
Toomey and Johnstone (1952) postulated what is known as the simple two-phase theory of
fluidization, which proposes that all gas flow in excess of that required for minimum
fluidization flows through the bed as bubbles. Thus, the bed can be divided into two phases: a
particulate phase (dense or emulsion phase) kept at minimum fluidization conditions and a
solids-free phase (void phase or bubble).

Although this theory has been frequently used, there is substantial evidence that it
overestimates the visible bubble flow. There has been considerable controversy over the
reason for this discrepancy (see Davidson, 1985, and Kunii and Levenspiel, 1991); the lack of
a throughflow term is the most widely accepted reason at present. Throughflow refers to gas
flowing through the bubble at a rate higher than that of the bubble itself. As discussed in Gera
and Gautam (1994), this flow is important in stabilizing the upper surface of bubbles, but is
difficult to measure in three-dimensional beds. A semi-empirical model for the division of the

7
gas flow, which accounts for throughflow, often called the modified two-phase flow theory,
was proposed by Johnsson et al. (1991).
Bubble shape, size and rise velocity
Observation of single bubbles injected into a bed kept at incipient fluidization provides
fundamental qualitative information about the character of the bubble flow. Bubbles in beds
of particles belonging to groups A and B in Geldarts classification are typically spherical
cap-shaped, as shown in Figure 4. As described in Kunii and Levenspiel (1991), this is often
ignored and bubbles are approximated through a sphere. The part of the sphere not included in
the actual bubble is often called the wake. In beds with gas flow rates above that of minimum
fluidization, bubbles are distorted in ways which are hard to predict.


Figure 4: Bubble shape as observed in particles belonging to groups A and B in Geldarts classification (taken
from Kunii & Levenspiel, 1991)

Bubble size is non-uniform throughout the bed and bubbles grow as they raise through the
bed, which makes the determination of the bubble size difficult. Various empirical and semi-
empirical correlations have been proposed to determine the mean bubble size for bubbles in
freely bubbling beds. However, as discussed by Davidson (1985), these correlations are based
on particular data from relatively small beds, which means they do not include all parameters
affecting the variables to model.

Based on experimental bubble observation in fluidized beds, the following correlation was
presented by Clift and Grace (1985) for the bubble rise velocity and is often used for bubbles
in any kind of fluidized bed:

( )
6
1
2
1
79 . 0 711 . 0
b e br
V g gD u = = . (2)

In practice, interactions between adjacent bubbles influence bubble properties such as shape,
size and rise velocity. Grace and Harrison (1969) showed that bubbles in a swarm rise more
rapidly than a single bubble, due to interaction with neighboring bubbles.
Davidsons bubble model
The Davidson model, presented in Davidson (1985), is based on experimental results derived
from observation of single bubbles injected in beds kept at incipient fluidization. The

8
equations constituting the model are based on the assumptions that bubbles in a fluidized bed
follow the same behavior as in an incompressible liquid of low viscosity and that the gas flow
is described by the potential flow theory. The model accounts for the movement of both the
gas and solids as well as the pressure distribution around rising bubbles.

Davidsons bubble model makes a distinction between two types of bubbles.
- Slow or cloudless bubbles (u
br
< u
f
): The emulsion gas rises faster than the bubble,
thus taking a shortcut through the bubble on its way through the bed.
- Fast or clouded bubbles (u
br
> u
f
): The bubble rises faster than the emulsion gas and
the gas leaving the top of the bubble is consequently swept around and returns into the
bubble without mixing with the main stream of fluidizing gas.

However, the assumption made in the Davidson model of potential flow for the gas phase is
only valid for conditions under which the viscous terms govern the pressure field in front of
the inertial terms, i.e. at low gas velocities. At the higher gas velocities typical for operating
fluidized bed units of most types, the model description of the pressure drop is not accurate.
Despite this, the simple approach and correctness of Davidson model under certain conditions
have made it the starting point of many other models for bubbling beds, as discussed in Kunii
and Levenspiel (1991).
Coalescence
As bubbles rise through the bed, they grow partly by means of coalescence (the process by
which two bubbles merges into a larger one). Coalescence occurs typically when a trailing
bubble catches up with a leading bubble. As explained in Davidson (1985), the trailing bubble
accelerates as it reaches the wake of the leading bubble and is drawn into the leading bubble.
As discussed in Kunii and Levenspiel (1991), bubbles in a freely bubbling bed are considered
to be small at the distributor, coalesce rapidly close to the bottom of the bed and grow more
slowly as they rise through the bed.

The general prediction of bubble size as a function of height is one of the ultimate aims of any
investigation of bubble coalescence in a fluidized bed. The various analytical models used to
describe the motion of interacting bubbles are based on the treatment of single bubbles in
liquid, derived by Davies and Taylor (1950). This problem is complex to treat theoretically
and analytically: no adequate general description of the coalescence process has been
developed so far.
The dynamic behavior of the gas fluidized bed and fluidization regimes
Although the bubble flow pattern is strongly dynamic, most research has been devoted to the
gross behavior of the fluidizing bed, i.e. the time averaged behavior. Bubbles give rise to local
differences in gas flow rate, which result in pressure fluctuations, as discussed in Davidson
(1985). Hence, the distribution of gas bubbles in space, time and size is of interest in any
application. The spatial distribution of bubbles formed at the gas distributor is controlled to a
large extent by the distributor. The gas flow through each nozzle or orifice in the gas
distributor depends both on the local pressure difference between the plenum and the bed
bottom and on the flow resistance of the distributor.

The fluidization regimes of a freely bubbling bed have been mapped by Svensson et al.
(1996) in terms of bubble behavior and pressure fluctuations. When operated at velocities
typical for fluidized beds, both bubbling and circulating, the fluidization regimes known as
the single and exploding bubble regime (shown in Figures 5a and 5c) can be observed in the

9
bottom bed. The single bubble regime occurs for low gas velocities; for higher gas velocities
the bed enters the exploding bubble regime. These regimes are characterized by a few large
bubbles which cause severe pressure fluctuations and are typical for beds with low-pressure
drop gas distributors. As shown in Figure 5b, the multiple-bubble regime presents numerous
smaller bubbles, which results in smoother fluidization and damped fluctuations.


Figure 5a: Single bubble regime (taken from Sasic et al. (2005) and Pallars et al. (2006))


Figure 5b: Multiple bubble regime (taken from Sasic et al. (2005) and Pallars et al. (2006))


Figure 5c: Exploding bubble regime (taken from Sasic et al. (2005) and Pallars et al. (2006))


10
2.3 Solids transport mechanisms
Experiments have shown that solids mixing is severely limited at minimum fluidization but
increases notably at slightly higher gas flow rates. Bubbles induce the solids mixing by
stirring the bed material as they are rising through the dense phase. As discussed by Niklasson
et al. (2002) among others, it is now commonly accepted that solids mixing in the bottom
region of a gas fluidized bed is mainly determined by the bubble flow pattern. For large units
with shallow beds, such as fluidized bed combustors, the lateral solids mixing has a major
influence on the performance of the boiler, as explained in Pallars et al. (2007). The main
transport mechanisms identified for the bottom region of fluidized beds, as illustrated in
Figure6, are listed below.


Figure 6: A representative diagram of solids transport taking place in the bottom bed and splash zone (taken
from Pallars, 2005)
Wake transport and gulf streaming
Gas drawn into the wake of the bubble drags solids with it, as schematized in Figure 4. The
solids are lifted with the bubble rise and dispersed along the path of the bubble. As discussed
in Shi and Fan (1984), this is the primary physical mechanism by which solids are mixed
axially in a bubbling bed, with little influence on the horizontal transport of solids. Numerous
experiments have been dedicated to determine the influence of the wake transport and its
contribution to the solids mixing.

It has been observed in large industrial fluidized-bed units that bubbles tend to follow a stable
spatial distribution with preferred bubble paths, as discussed in Werther (1977), determined
by the wall effects in combination with bed and air distributor properties. This bubble flow
pattern leads to a characteristic circulation pattern of solids in which they flow upwards in
regions where bubbles are frequent and downwards in regions where bubbles are less
frequent. The combined effects of gross circulation and small scale local mixing leads to a
notable axial mixing. The rate of lateral mixing is much less pronounced, particularly in
shallow beds where the lateral influence of rising bubbles is weak, as described in Lim et al.
(1995).

11
Splash zone
As bubbles reach the bed surface and erupt, solids from the bubble wake and roof are ejected
into the freeboard. As explained in Niklasson et al. (2002), these ejected solids either can be
entrained by the gas stream and leave the riser or fall back to the bed surface and rejoin the
emulsion. The lateral solids mixing is strongly dependent upon this bubble-induced scattering
of solids at the top of the bed surface. Indeed, the presence of vertically aligned baffle plates
has been shown to significantly decrease the lateral solids mixing, as discussed in Shi and Fan
(1984).

2.4 Models of gas flow and solids mixing
As discussed in Verloop (1968) and later summarized in Lim et al. (1993), various models
have been proposed in the literature to describe the solids mixing in fluidized beds. The
earliest models are of limited use, since they describe the solids mixing by combining
different flow systems (such as perfectly mixed, plug flow, short circuiting), thus neglecting
the physical behavior of the bed. Later, more refined models for solids mixing (and
occasionally also gas flow) are briefly introduced and commented on below.

The K-L bed model or The bubbling bed model
The K-L bed model developed by Kunii and Levenspiel (1991), whose principle is illustrated
in Figure 7, is based on the Davidson bubble model; it gives the gas flow and pressure profile
around the bubble and differentiates between slow cloudless bubbles and fast clouded
bubbles. The model also takes into account the contribution of the wakes to the flow pattern
of both gas and solids.


Figure 7: The bubbling bed model (taken from Kunii and Levenspiel, 1991)

All rising bubbles are assumed to drag behind them a wake with solids which are
continuously exchanged with solids from the emulsion. At the top of the bed, wake particles
rejoin the downflowing emulsion, where the relative gas-solid velocity is assumed to be that
of minimum fluidization. The bubble rise velocity is determined by Eq. 2 given in Section 2.2.
The emulsion phase is assumed to remain at minimum fluidization.


12
While the K-L model does account for the importance of the bubble size under the flow
conditions for both gas and solids in the bed, it is based on the Davidson model; it is therefore
unsuitable for systems operated under conditions far from incipient fluidization.
The Counter-Current Back Mixing (CCBM) model
The counter-current back mixing (CCBM) model proposed by van Deemter (1961) and later
refined by Gwyn et al. (1970) is based on considering the varying flow characteristics for
different bed regions. Van Deemter suggested that the bed could be divided into two phases: a
dense phase consisting of particulate aggregates and a dispersed phase with high vertical gas
velocity. The mass exchange between the two phases is enabled through the break-up and
formation of particulate aggregates, and the exchange between the regions is described by an
empirically-determined exchange coefficient. To account for the horizontal movement of
solids, CCBM models are often combined with a dispersion or diffusive model. Abanades et
al. (2001) included the lateral displacement of solids in a CCBM model structure by
connecting multiple vertical mixing columns.

Since the two-phase division in the CCBM model, there have been models dividing the flow
into more phases, but the development and utilization of the multi-phase CCBM model has
been crippled by the increased need for empirical coefficients which are difficult to measure,
as discussed in Lim et al. (1995). Among the available models at present, CCBM models are
generally believed to give the best description of solids mixing in bubbling beds, despite the
fact that other possible mixing mechanisms, such as solids splashing at the bed surface and
turbulent mixing near the distributor, are not accounted for.

The division of the bed into layers with different flow characteristics is represented by
counter-current back mixing models. The use of CCBM models is inhibited by the lack of
experimental data needed to determine some parameters in the model setup. None of these
models gives a description of the fuel mixing in satisfactory agreement with experimental data
under conditions generally applicable to fluidized bed combustors.
The Brownian dispersion model
As discussed in Kunii and Levenspiel (1991) solid mixing in a fluidized bed occurs as the
sum of two flow mechanisms: a convective and a random (diffusive) one. However, the
convective contribution from the circulatory motion is usually combined with the diffusion,
and both are represented by a dispersion coefficient. Horizontal solids mixing on a
macroscopic scale is often simplified by describing the process in terms of Brownian
diffusion transport:

S C D
t
C
si
s
+ V =
c
c
2
(3)


where C is the solids concentration, and D
si
is the dispersion coefficient which depends on
operational parameters and unit geometry and varies with location, both horizontally and
vertically. All transport processes are embedded in the dispersion coefficient which is
generally averaged over the cross-sectional area and, therefore, denoted as the effective
dispersion coefficient.

Several techniques have been used to determine the rate of solids transport in a fluidized bed.
The results of these investigations are often reported in the form of a dispersion coefficient.
As shown by Niklasson et al. (2002), the high scattering in reported values for the dispersion

13
coefficient is an indication that the affect of the different physical mechanisms underlying the
lateral dispersion vary from study to study. As a consequence, the correlations derived for the
dispersion coefficient found in the literature obviously show a large scatter. This observation
should invoke caution when applying expressions for the dispersion coefficient derived from
a unit operating under conditions other than those modeled.

The original development of the dispersion model, using experimental data from tall, narrow
beds with Geldarts group A and small group B particles, focused on the axial mixing.
However, as explained in Abanades et al. (2001), the dispersion model can not describe the
vertical movement of solids in large-scale fluidized bed units (i.e. combustors) with
satisfactory agreement.

Characterization of the solids mixing as a diffusive process or by a mixing coefficient is an
easy-to-implement method, but its use is questionable since the local dynamics of the bed are
neglected.
CFD simulations
The use of numerical simulations from first principles, computational fluid dynamics (CFD),
is thought to be the future tool for modeling fluidized units, but so far mostly systems of
limited particle numbers and uniform particle size and density have been evaluated due to
limited computational power.



14
3. Theory
The performance of a combustion unit is strongly dependent on the fuel distribution in the
bed, which is why accurate predictions of this through modeling is a key aim in the
understanding and optimization of the combustion process.

The desired model should, for easy implementation, describe the fuel distribution on the basis
of variables that can be measured and/or controlled during operation:
- the properties of the bed material (density and size distribution),
- the bed dimensions,
- the characteristic P-u curve of the air distributor, and
- the operating conditions (fluidization velocity, temperature, etc).

3.1 The bubble flow model
The gas-solid flow in the bottom bed is governed by the presence of bubbles, which gives the
flow picture a dynamic character and strongly affect the pressure profile. Thus, modeling has
to be carried out on a dynamic basis for an accurate description of the bottom bed. The events
to be modeled are: bubble formation, growth, detachment and uprise through the bottom bed.

Once the bed dimensions and bed material properties are established the complete bubble
flow model should properly predict:
- bubble characteristics (position, size, spatial distribution, frequency, density, rise velocity
and direction),
- dynamical pressure field, and
- superficial gas velocity field.

As discussed in Section 2.1, the single and exploding bubble regimes are the most common in
fluidized bed boilers. In the exploding bubble regime the bubble can remain attached to the
distributor until it breaks the bed surface, while in the single bubble regime the bubble often
detaches from the distributor and travels through the bed. Since stable preferred bubble paths
can be readily established under these regimes, a single preferred bubble path and its adjacent
emulsion is considered to be a representative section of the bed in this work.

At present, the bubble is assumed for simplicity to be rectangular and to grow only in the
vertical direction. As schematized in Fig. 8, the region including the bubble is termed the
throughflow phase and the adjacent emulsion is termed the emulsion-only phase.


15

Figure 8: A schematic representation of the modeled bed segment

The bubble flow model presented in this work is based on the following assumptions:
i. Solid particles are spherical and monosized;
ii. The air distributor is of porous type;
iii. The gas has no vertical velocity gradients and flows strictly vertically, i.e. there is no
exchange of mass between the throughflow and the emulsion-only phase;
iv. The emulsion voidage is a local function of the gas velocity;
v. The pressure drop in the emulsion follows the Ergun equation;
vi. The bed is operated in the exploding bubble regime, thus the bubble never detaches
from the air distributor plate;
vii. Bubbles are solids-free voids without properties of their own; and
viii. The bed height is assumed constant and equal to the height at minimum fluidization,
i.e. the increase in volume due to the expansion of the bubble is assumed to be
compensated for by the part of the bed excluded from the section of interest.

Gas flowing through the bubble stabilizes the bubble roof, and is the means by which the
bubble rises through the bed. Gas leaving through the bubble roof exerts a drag force on the
particles contained between the bubble roof and the bottom bed surface. A high enough gas
velocity causes these particles to move, allowing the bubble to ascend through the bed. Thus,
particles are mere passengers dragged along by the gas flow. Since the system is strongly
dynamic, a system of equations has to be solved at each time step to determine the evolution
of the bubble front position and the gas velocity through the two phases involved.

The gas flow through the two phases making up a bed segment is controlled by the pressure
drop across each phase. The impact of a bubble on the balance between the two phases can be
compared with the electrical circuit, schematized in Fig. 9 with the emulsion-only and the
throughflow phases set in parallel.

16

Figure 9: A representation of the flow resistances in a freely bubbling fluidized bed as a power circuit.

In the electrical analogy, gas velocity is likened to current intensity, pressure to voltage, and
mass flow resistance to electric resistance. Hence,

M
gas
E
R
P
U
R
V
I
A

A
= . (4)

As seen from Figure 9, two pressure loss terms exist for each of the phases: one through the
air distributor, R
Dist
, and one through the bed material, R
Bed
. Thus, flows through each one of
the two phases considered are arranged so that a same total pressure loss is obtained for both.

If the pressure drop across the bed is much larger than that across the distributor, that is R
Dist

can be more or less disregarded, the presence of a bubble significantly alters the pressure
balance between the two phases, notably increasing the flow in the throughflow phase.
Conversly, if the pressure drop across the distributor is larger than the drop across the bed, the
presence of a bubble does not alter the pressure balance much, and the division of the gas
between both phases remains relatively unaffected.

When the pressure drop across the bed is dominant, bubbles grow larger since most of the gas
seeks out the low resistance path which bubbles represent. When the pressure drop across the
distributor dominates, smaller and more homogeneously distributed bubbles are formed in the
bed, which does not induce any stable preferred bubble paths:

>>
A
A
1
dist
bed
P
P
Large bubbles forming stable preferred bubble paths
<<
A
A
1
dist
bed
P
P
Small, homogeneously-distributed bubbles

P
ple

P
atm

R
Bed

Eq. 10
R
Bed

Eq. 10
R
Dist

Eq. 9
Throughflow
phase
Emulsion-only
phase
R
Dist
Eq. 9
u
th
u
e


17
A widely used correlation, given by Wen et al. (1966), to determine the particle Reynolds
number at which minimum fluidization occurs is:

1 2
2
1 ,
Re C Ar C C
mf p
+ = (5)

where
( )
2
3
g
g s g p
g d
Ar


= . (6)

Several pairs of the values for C
1
and C
2
have been proposed in the literature. Grace (1986)
suggested the use of C
1
= 27.2 and C
2
= 0.0408. The minimum fluidization velocity, u
mf
, can
be obtained from the Reynolds number expression:

g p
g mf p
mf
d
u


,
Re
= . (7)

The void fraction under minimum fluidization conditions,
mf
, is calculated with the Ergun
equation:

( ) ( )
2
,
3
,
2
Re
1 75 . 1
Re
1 150
mf p
mf
mf
mf p
mf
mf
Ar
c
c
c
c
+

= . (8)

The pressure drop through the distributor plate is generally determined by the characteristic
curve of the distributor:

2
i Dist
ku P = A . (9)

Since the bubble is assumed to be solids free, the gas passes through the bubble without any
pressure drop. The pressure drop across the emulsion is described by the Ergun equation:
( ) ( )
|
|
.
|

\
|

+

= A
2
3 2 3
2
1 75 . 1 1 150
i
p
i
p
Bed
u
d
u
d
L P
c
c
c
c
(10)

where
0 B
h L = for the emulsion-only phase and
x h L
B
=
0
for the throughflow phase.

Unrealistically high pressure drops are obtained at high gas velocities with Eq. 10 if particles
are assumed to be homogeneously distributed. Thus, particles in the emulsion are thought to
have irregular mesoscale structures, i.e. to form clusters, which results in effective larger
particles (see Section 3.3). The clusters are assumed to be spherical and are characterized by a
particle diameter with the same density as a single particle. The excess pressure in the plenum
can then be described by combining Eq. 9 and Eq. 10:

Bed dist ple
P P i P A + A = ) ( . (11)

The pressure inside the bubble exerts a force on the emulsion above the bubble, as shown in
Fig.10. Thus, at a given time step, the displacement of the bubble front can be determined by

18
Newtons second law. Since the gas leaving through the bubble roof flows through the
emulsion above the bubble, the excess pressure in the bubble, P
bub
, can be described by the
Ergun equation (Eq. 10 with L = hb0-x), as show in Eq. iii in the derivation immediately
below.


mg F
g
= (i)
A P F
atm atm
= (ii)
( )A P P A P F
Erg atm Bub Bub
A + = = (iii)

Thus, Newtons second law gives:
ma A P mg F F F F
Erg ple atm g
= A = + = E (12)


Figure 10: Force balance on the emulsion above the bubble in the throughflow phase

A further derivation of Eq. 12 yields:

( ) ( )
( )( ) g a u
d
u
d
s i
p
i
p
+ =

c
c
c
c
c
1
1 75 . 1 1 150
2
3 2 3
2
. (13)

The local emulsion voidage has been studied with optical probes in a cold fluidized bed, as
detailed in Appendix A. Withthe experimental data obtained in this study, a correlation
describing the local voidage in the emulsion as a function of the gas velocity can be derived:

37325 . 0 2104 . 0 + =
i
u c . (14)

The pressure in the plenum is controlled by an air fan regulating system which is modeled
according to:

( ) ( ) ( ) ( ) 1 1 A + = i u u t i P i P
m wanted ple ple
o . (15)

The rate of change in pressure is governed by the fan variable, , and is determined so that the
time-averaged gas velocity corresponds to the value defined as input (see Section 3.3 for
details). Although the rate of change should be restricted by the mechanical capacity of the air
systems to react to changes, this is not accounted for in the model at present time. The phase-
averaged velocity, u
m
, used in Eq. 15, u
m
, is defined as:

( )
B E B B m
i u i u i u + = 1 ) ( ) ( ) ( (16)

where
B
is the fraction of cross-sectional area covered by the bubble.

The horizontal size of the bubble is fitted to predict the average bubble density as calculated
from published experimental data on Chalmers CFB boiler, which is roughly around = 0.27,
as measured by Johnsson and Leckner (1995).

F
g

F
Bub

F
atm


19
3.2 Computational flow scheme
The model requires the set of input data specified in Table 1 and derives the output data
specified in Table 2:

Table 1: Input data required by the model
Solids density
s

Particle diameter d
p

Operating temperature T
Bed height h
B0

Average operating superficial gas velocity u
wanted

Characteristic constant of air distributor k
Average bubble density
B


The calculation of the bubble rise transient is time-discretized and implemented through an
iteration-based computational structure, which is given in Fig.11a. The dynamic loop
containing the system of equations solved at each time step is depicted by Fig.11b. As seen in
Fig.11a, three self-contained iterations are needed to reach a solution.


Figure 11a: Computational flow scheme Figure 11b: Dynamic loop

gas U
bed P A
B o
Input
(table 1)
Output
(table 2)
gas
wanted
U
u
o o =
mf
bed
cluster cluster
P
P
d d
A
A
=
B
B
B B
o
o
=
Dynamic loop
(fig.11b)
( ) ( ) 1 = i x i x
Eq. 11
and
Eq. 15
u
i

Eq. 13
a
i = i + 1
( ) ( ) t i t i t A + = 1
X
x(i)=X

20
Table 2: Output data from the model
Bubble front position x(t)
Bubble cross-sectional fraction
B

q (see Chapter 4)
Dist
Bed
P
P
A
A

Superficial gas velocity in the through-flow phase U
Th
(t)
Superficial gas velocity in the emulsion-only phase U
E
(t)
Pressure in the plenum P
ple
(t)
Pressure drop across the bed P
ThBed
(t), P
EBed
(t)
Pressure drop across the distributor plate P
ThDist
(t), P
EDist
(t)

Since minimum fluidization conditions have to be fulfilled before the bubbling begins, these
are chosen as initial bed conditions. Initial values and the time step used in the present
calculations are given in Table 3. The computer code used (implemented in Matlab 7.3.0) is
available in Appendix B.

Table 3: Initial parameter values and time step used in the present calculations
Regulating system constant
0
3.5e6 Pa/m
Bubble fraction
B0
0.8
Time step t 0.1e-3 s



21
4. Results and discussion
The present model describes the behavior of a growing bubble in the bottom bed of a
fluidized bed combustion unit, for which all results from the model are given as a function of
time. Once the bubble has reached the bed surface, the bubble structure is assumed to collapse
(as seen from experiments) and the gas velocity in the bed drop drastically, defluidizing the
entire segment, which represents the end of the dynamic cycle. At present, the cycle is
incomplete since the period between short-circuit, when the bubble breaks the bed surface,
and the refluidization of the bed segment is not included in the model. Two parameters
influencing the dynamics in an FBC unit are the bed height, h
B0
, and the characteristic
constant of the air distributor, k. A reference case is chosen as the basis for a later comparison
in which the influence of these two parameters is investigated.

4.1 Reference case
Care should be taken when determining the input data, i.e. the operating conditions, to avoid
exceeding the parameter range in which the model is valid. The conditions for the reference
case, given in Table 4, are therefore chosen to represent typical operational conditions for a
circulating FBC unit.

Table 4: Input data for the reference case
Solids density
s
2600 kg/m
3

Particle diameter d
p
0.4 mm
Temperature T 20C
Average gas velocity u
wanted
3 m/s
Average bubble density
B
0.27
Bed height h
B0
0.5 m
Characteristic constant of air distributor k 300 Pas
2
/m
2


The bubble front propagates through the bed until it reaches the bed surface, as shown in
Figure 12. The rise velocity is low during the initial formation of the bubble, but increases as
the bubble grows larger. As the bubble front approaches the bed surface the rise velocity
levels out and becomes more constant. While analyzing these results, it must be kept in mind
that only the bubble front is taken into account in this model, and that both the rise velocity
and bubble position might be significantly altered if the bubble detaches from the air
distributor.


22

Figure 12: Modeled evolution of the bubble front, h
B0
= 0.5m, k = 300 Pas
2
/m
2


The superficial gas velocity for each phase as well as the weighted average gas velocity is
shown in Figure 13. It is clear that the velocity in the throughflow phase is higher than in the
emulsion-only phase. This is a consequence of the introduction of a region without flow
resistance, i.e. a bubble. The high velocity indicates that a substantial amount of the gas flow
passes through the throughflow phase. As the bubble grows, the flow resistance in the bed
section of the passage is decreased accordingly, and an increasing amount of the gas flow
passes through the throughflow phase. Due to the corresponding decrease in gas flow through
the emulsion-only phase, the gas velocity in the emulsion-only phase decreases as the bubble
propagates through the bed. The velocity even drops below the minimum fluidization
velocity, u
mf
, as the bubble approaches the bed surface. Therefore, in the case studied, the
emulsion-only phase defluidizes before the bubble breaks the bed surface.


Figure 13: Modeled superficial gas velocities in the throughflow (U
TH
) and emulsion only phase (U
E
) and the
weighted average gas velocity (U
M
).


23
Since it is not possible to reach as high fluidizing gas velocities as desired in a FBC unit
without forming bubbles the pressure in the plenum has to overcome the weight of the
incipiently fluidized bed and the flow resistance of the distributor plate. Once the pressure
required to sustain the desired gas velocity is obtained, the pressure in the plenum drops off,
as shown in Figure 14. This explains why the gas velocity appears independent of the bubble
fronts position: A lower flow resistance in the bed section allows for a high gas flow to be
sustained with a lower pressure in the plenum.


Figure 14: Modeled pressure in the plenum, pressure drop across the bed (as expressed by Eq. 17) and time-
averaged pressure drop across the bed.

The pressure drop across the bed is the weighted sum of the pressure drop in both phases
expressed as:

( ) ( )
2 2
1
E B B B Ple Bed
u u k P P + = A . (17)

It can also be seen in the figure that the pressure in the plenum is actually higher than the
minimum requirement for forming a bubble. The rapid increase of pressure in the plenum is a
result of the present formulation of the fan regulating equation, Eq. 15.

The frequency of the pressure and gas velocity fluctuations is determined by the time it takes
for the bubble to break through the surface of the bed and short-circuit the system, while the
amplitude of the pressure fluctuation is the difference between the maximum value (occurring
at bubble formation) and the lowest value (occurring as the bubble approaches the surface).

4.2 The influence of the characteristic constant of the air distributor
and the bed height
The ratio of the pressure drop across the bed to that across the distributor determines the
division of gas flow between the emulsion-only phase and the throughflow phase (see Figure
8). In this work the variable, q, is defined as the ratio of the time-averaged weighted sums of
the pressure drops across the bed to those across the distributor:


24
Dist
Bed
P
P
q
A
A
= . (18)

This ratio, as given by Eq. 18, can be modified by changing the bed height, the characteristic
constant of the air distributor, or both. To investigate the impact of the bed height and the
characteristic constants of the air distributor represented by q, the simulation matrix shown in
Figure 15 was applied.


Figure 15: Simulation matrix; case numbers given in parenthesis.

The fact that, at present stage, the model does not include the detachment of the bubble from
the air distributor imposes the assumption that the bed is operated in the exploding bubble
regime, which is the regime for which this assumption is most realistic. This regime is
observed in systems where q >>1, which implies that the span of the simulation matrix is
limited by the inability of the model to handle cases for which the pressure drop across the
distributor dominates over the pressure drop across the bed, that is q << 1.

In the exploding bubble regime, the pressure drop across the distributor has to be low enough
to allow the flow through the distributor to be non-uniform. Thus, the higher the variable q,
the better the model applies to the simulated conditions. With q < 1, the flow through the
distributor becomes more uniform and a significant rearrangement of the gas flow has to
occur in the bed, which the model at present cannot handle, as discussed above. Both
conditions and results for the cases presented in the simulation matrix are given in Table 5.

Table 5: Conditions and results for the nine cases considered
Parameters Results

Case
h
B0

[m]
k
[kg/m
3
]

B
[-]
q
[-]
10
-6

[Pa/m]
d
cluster
[mm]
Fluctuation
frequency
[Hz]
Ref 0.5 300 0.8234 2.40 4.1 2.4 3.21
2 0.5 100 0.8195 7.02 3.5 2.3 3.38
3 0.5 500 0.8265 1.46 4.9 2.5 3.01
4 0.25 300 0.8245 1.22 3.6 2.5 4.12
5 0.25 100 0.8189 3.57 3.5 2.4 4.67
6 0.25 500 0.8197 0.74 4.7 2.8 3.48
7 1.0 300 0.8210 4.70 4.7 2.4 2.35
8 1.0 100 0.8203 13.99 4.1 2.3 2.42
9 1.0 500 0.8228 2.87 5.4 2.4 2.30
Reference case
k = 300 Pas
2
/m
2
h
B0
= 0.5 m
Constant k
(4) h
B0
= 0.25 m
(7) h
B0
= 1.0 m
Constant h
B0

(2) k = 100 Pas
2
/m
2

(3) k = 500 Pas
2
/m
2

Changing both h and k
(5) h
B0
= 0.25 m, k = 100 Pas
2
/m
2

(6) h
B0
= 0.25 m, k = 500 Pas
2
/m
2

(8) h
B0
= 1.0 m, k = 100 Pas
2
/m
2

(9) h
B0
= 1.0 m, k = 500 Pas
2
/m
2




25
The influence of the bed height on the dynamics of the bubble front position, i.e. the bubble
front velocity, is shown in Figure 16. When the bed height is decreased (compare the
reference case to Case 4), the bubble front propagates more slowly through the bed, i.e. longer
cycle times lower the fluctuation frequency. The same results can be reached by increasing
the characteristic constant for the distributor plate (compare Case 4 to Case 6). On the other
hand, a decrease in the characteristic constant for the distributor as well as a decreased bed
height would then increase the bubble front velocity (compare the reference case to Case 5).


Figure 16: Modeled evolution of bubble front position for the reference case (q = 2.4), Case 4 (q = 1.22), Case 5
(q = 3.57) and Case 6 (q = 0.74).

The effect off a further increase in bed height or decrease of the characteristic constant for air
distributor will where off when q increase far beyond unity. As the relative change in rise
velocity becomes less noticeable, further decreasing the characteristic constant or increasing
the bed height only adds to the operational costs in terms of fan power without further
modifying the gas flow pattern.

These results demonstrate that individual values of the bed height and the characteristic
constant for the air distributor are not important in themselves, but their ratio is. Henceforth,
only the pressure drop ratio, q, will be taken into account, rather than the individual values for
the two parameters investigated. Two systems are therefore expected to exhibit similar
dynamic properties, as long as they have similar values of q, as can be seen in Figure 17.


26

Figure 17: Modeled evolution of the bubble front position for the reference Case (q = 2.4), Case 5 (q = 3.57) and
Case 9 (q = 2.87)

The effect the variable q has on the superficial gas velocity through the throughflow phase is
too small to give a qualitative description of the influence the variable has on the throughflow
phase. Therefore, attention is given to the superficial gas velocity through the emulsion-only
phase, shown in Figure 18, where the influence of the variable q is much clearer.


Figure 18: Modeled superficial gas velocity through the emulsion phase for the reference case (q = 2.4), Case 7
(q = 4.70), Case 8 (q = 13.99) and Case 9 (q = 2.87).

Increasing q, by increasing bed height (compare the reference case to Case 7), decreasing the
characteristic constant of the air distributor compare Case 7 to Case 8), or both, results in a
decrease in superficial gas velocity through the emulsion-only phase. Decreasing q (compare
Case 7 to Case 9) has the opposite effect on the gas velocity. Regardless of the value of q, the
velocity in the emulsion-only phase drops below the minimum fluidization velocity, and the
emulsion-only phase is defluidized. In Figure 18 it can also be noted that the amplitude of the
velocity fluctuations in the emulsion-only phase increases when decreasing q (compare Case

27
8 to Case 9), while the fluctuation frequency is barely altered. Hence, the bed height can be
argued to determine the frequency while the characteristic constant influences the amplitude
of the fluctuations.

As shown in Figure 19, lowering q leads to reduced pressure in the plenum (compare the
reference case to Case 6). It can also be seen that the amplitude of the pressure fluctuations
decreases as q is lowered. This could, according to the discussion in Section 3.1, indicate that
bubbles are smaller, more frequent, and unable to form stable preferred bubble paths through
the bed.


Figure 19: Modeled pressure levels in the plenum position for the reference case (q = 2.4), Case 6 (q = 0.74) and
Case 8 (q = 14)

The tendency observed from the modeled fluctuations (both pressure levels in the plenum and
superficial gas velocity in the emulsion-only phase) seems to indicate that the lower the value
of q, the less severe the fluctuations, which is also in agreement with the observations
discussed in Section 2.2 and shown in Figures 18 and 19. This indicates that the superficial
gas velocities in the phases do not differ much, the bubbles are smaller and the bed is
probably not operated in the exploding bubble regime.

As noted above, since the application mapping of the model is restricted to the exploding
bubble regime, it is not possible to make realistic simulations of cases in which the pressure
drop across the distributor dominates strongly over the pressure drop across the bed.


28
5. Conclusions
A first-step one-dimensional model for the bubble flow in a fluidized bed operated in the
exploding bubble regime has been implemented. From stable preferred bubble paths which
commonly occur, a single such bubble path and its adjacent emulsion is assumed to be a
representative section of the bed. The gas flow is divided into a throughflow phase and
emulsion-only phase. It is found that the fluidization regime depends on the ratio of the
pressure drop across the air distributor plate to that across the bed. The model implemented is
unique in its capability to account for the influence of the air distributor characteristics and the
bed height on the dynamics of both superficial gas velocity and pressure fields.

It was shown that a high pressure drop ratio, q, results in severe fluctuations in the system,
while a lower q results in more moderate fluctuations, which is in agreement with the
observations discussed in Section 2.2 and shown in Figures 18 and 19. Moderate fluctuations
indicate that the superficial gas velocities in the phases do not differ much: the bubbles are
smaller and the bed is probably not operating in the exploding bubble regime. Due to the
restriction of the model, it is not possible to perform realistic simulations of cases in which the
pressure drop across the distributor dominates strongly over the pressure drop across the bed.



29
6. Further work
In the short term, in order to encompass other bottom bed regimes besides the exploding
bubble regime considered here, criteria and descriptions of the processes taking place when
the bubble detaches from the air distributor are needed. This, together with a refined modeling
of the fan inertia can open the way for comparison with experimental data on the bubble flow.

In a longer term, once the description of the bubble is completed, the model of the bubble
flow can be coupled to the movement of the solids (both mixing and segregation) to obtain a
model describing the solids (and thereby, fuel) mixing in the bottom region of a fluidized bed.


30
References
Abanades, J. C., Grasa, G. S., 2001. Modeling the Axial and Lateral Mixing of Solids in
Fluidized Beds. Ind. Eng. Chem. Res., 40, 5656-5665
Clift, R., Grace, J. R., 1985. In: Fluidization, 2
nd
edn., J. F. Davidson et al., eds., pp.79-80,
Academic Press, Inc., Orlando, Florida
Davidson, J. F., 1985, In: Fluidization 2
nd
edn., J. F. Davidson et al. eds., Academic Press,
Inc., Orlando, Florida
Deemter, van, J. J. 1961. Mixing and contacting in gas-solid fluidized beds. Chemical
Engineering Science, 18, pp. 143-154.
Geldart, D., 1973. Types of gas fluidization, Powder technology, 7, 5, 285-292.
Gera, D., Gautam, M., 1994. Variation of throughflow velocity in a 2-D rising bubble.
Powder Technology, 79, 257-263.
Grace, J. R., Harrison, D., 1969. The behaviour of freely bubbling fluidising beds. Chemical
Engineering Science, 24, pp. 497-508.
Grace, J. R., 1986. Contacting modes and behaviour classification of gas-solid and other two-
phase suspensions. Canadian Journal of Chemical Engineering, 64, 353-363.
Gwyn, J. E., Moser, J. H., Parker, W. A., 1970. A three-phase model for gas-fluidized beds.
Chem. Eng. Prog. Symp. Ser., 66, 19-27.
Johnsson, F., Andersson, S., Leckner, B., 1991. Expansion of a freely bubbling fluidized bed.
Powder Technology, 68, 117-123.
Kunii, D., Levenspiel, O., 1991. Fluidization engineering. Butterworth-Heinemann, Newton,
MA.
Lim, K. S., Gururajan, V. S., Agarwal, P. K., 1993. Mixing of homogeneous solids in
bubbling fluidized beds: Theoretical modelling and experimental investigation using
image analysis. Chemical Engineering Science, 48, 12, pp. 251-2265.
Lim, K. S., Agarwal, P. K., 1994. Circulatory motion of a large and lighter sphere in a
bubbling fluidized bed of smaller and heavier particles. Chemical Engineering Science,
49, 3, pp. 421-424.
Lim, K. S., Zhu, J. X., Grace, J. R., 1995. Hydrodynamics of gas-solid fluidization, Int. J.
Multiphase Flow, 21, Suppl. pp. 141-193.
Magnusson, A., Rundqvist, R., Almstedt, A.E., Johnsson, F. 2005. Dual fiber optical probe
measurements of solids volume fraction in a circulating fluidized bed. Powder
Technology, 151, pp. 19-26.
Niklasson, F., Thunman, H., Johnsson, F., Leckner, B., 2002. Estimation of Solids Mixing in
a Fluidized-Bed Combustor. Ind. Eng. Chem. Res., 41, 4663-4673.
Pallars D., 2005. Macroscopic Modeling of Fluid Dynamics in Large-Scale Circulating
Fluidized beds, Licenciate Thesis, Dept. of energy and Enviroment, Chalmers
University of Technology, Gothenburg, Sweden.
Pallars, D., Johansson, A., Johnsson, F. 2006. Interpretation of dynamics of pressure
measurements. Proc. of the 19
th
Int. Conf. on Fluidized Bed Combustion (Vienna,
Austria).
Pallars, D., Dez, P., Johnsson, F., 2007. Experimental Analysis of Fuel Mixing Patterns in
a Fluidized Bed. Proc of the 12
th
International Conference on Fluidization,
(Vancouver, Canada)
Sasic, S., Leckner, B., Johnsson, F., 2005. Parametric modeling of time series of pressure
fluctuations in gassolid fluidized beds. Chem. Eng. Sci., 60, 5069-5077
Shi, Y., Fan, L. T., 1984. Lateral Mixing of Solids in Batch Gas-Solids Fluidized Beds. Ind.
Eng. Chem. Process Des. Dev., 23, 337-341.

31
Svensson, A., Johnsson, F., Leckner, B., 1996. Bottom bed regimes in a circulating fluidized
bed boiler. Int. J. Multiphase Flow, 22, 6, pp. 1187-1204.
Toomey, R.D., Johnstone H.F., 1952. Chemical Engineering Progress, 48, 220
Verloop, J., Nie, L. H., Heertjes, P. M., 1968. The Residence Time of Solids in Gas-Fluidized
Beds. Powder Technology, 2, 1, 32-42.
Wen, C. Y., Yu, Y. H., 1966. A generalized method for predicting the minimum fluidization
velocity. AlChE Journal, 12, 610-612.
Werther, J., 1977. Bubble chains in large diameter gas fluidized beds. International Journal of
Multiphase Flow, 3(4), 367-381.



i






Appendix A

The use of optical probes for experimental
determination of the local bed porosity

ii
The influence of the superficial gas velocity on the local emulsion porosity has been measured
with an optical probe; see Magnusson et al. (2005). The resulting data, given in Table 6, were
correlated through linear regression to give an expression for the porosity as a function of the
gas velocity. The correlation was formulated expressly to ensure the correct prediction of the
point of minimum fluidization, as can be seen in Fig.A.1.

Cold Bed
y = 0,2104x + 0,3733
R
2
= 0,5302
0,35
0,40
0,45
0,50
0,55
0,60
0,65
0,70
0,75
0,10 0,30 0,50 0,70 0,90 1,10 1,30
Superficial gas velocity
E
p
s
i
l
o
n

Figure A.1. Measured porosity values (Epsilon) as a function of the superficial gas velocity. The solid line
indicates the correlation adopted.

Particles used in the experiment were typical silica sand particles used as bed material in FBC
units. As the gas velocity increases the correlation predicts unreliable values for the porosity,
hence the correlation is assumed to be valid only in the range of gas velocities for which the
particles can be assumed to expand homogeneously. When the gas reaches such a high
velocity that the formation of cluster starts, the porosity is assumed to become independent of
the velocity.

Table 6. Data shown in Figure A.1.
Gas velocity
Epsilon
(measured)
Epsilon
(correlated)
0.157 0.40632 0.40632
1.180 0.680 0.6214
0.910 0.700 0.5646
1.017 0.703 0.5871
0.917 0.693 0.5660
0.923 0.657 0.5674
0.900 0.610 0.5625
0.960 0.587 0.5751
0.957 0.590 0.5744
0.997 0.563 0.5829
u
mf
, e
mf


iii
0.867 0.570 0.5555
0.907 0.540 0.5639
0.840 0.553 0.5499
0.913 0.523 0.5653
0.840 0.527 0.5499
0.837 0.527 0.5492
0.787 0.517 0.5387
0.813 0.527 0.5443
0.783 0.513 0.5380
0.780 0.520 0.5373
0.710 0.520 0.5226
0.757 0.517 0.5324
0.747 0.513 0.5303
0.740 0.507 0.5289
0.693 0.503 0.5191
0.680 0.497 0.5163
0.700 0.493 0.5205
0.733 0.490 0.5275
0.767 0.490 0.5345
0.700 0.490 0.5205
0.707 0.490 0.5219
0.637 0.490 0.5071
0.693 0.490 0.5191
0.647 0.483 0.5093
0.637 0.477 0.5071
0.610 0.467 0.5015
0.667 0.463 0.5135
0.743 0.463 0.5296
0.730 0.463 0.5268



i







Appendix B

Computer code

ii

function [ratio,Dp,Alpha,Emax,Xpos,Tid,U,XiB,P,Eps]=bubbleflowmodel;
% The program determine the dynamic superficial gas velocity and the bubble
% front position.

Tic; clear all; close all;
global Repmf Ar i hb0 x k P_ply umf u1 u2 my_air rho_air rho_s dp g e1 e2
emax

%% Constants
g = 9.82; % [N/kg]
rho_s = 2600; % [kg/m3]
dp = 0.4e-3; % [m]
my_air = 1.81e-5; % [Pa*s]
rho_air = 1.189; % [kg/m3]
hb0 = 0.5; % bed height at minimum fluidization [m]
k = 300; % characteristic constant for the air distributor
u_wanted = 3; % operating fluidization gas velocity
deltat = 1e-4; % time step
alpha = 3.5e6; % initial guess for the fan constant
DBgoal = 0.27; % average bubble density

%% Minimum fluidization, Arkimedes and Reynolds number
K1 = 27.2;
K2 = 0.0408;
Ar = dp^3*rho_air*(rho_s-rho_air)*g/my_air^2;
Repmf = sqrt(Ar*K2+K1^2)-K1;
umf = Repmf*my_air/(dp*rho_air);
e_mf = fzero(@sandra,[0.3 0.6]);
P_ply(1) = rho_s*(1-e_mf)*hb0*g + k*umf^2;
Ptarget = hb0*(1-e_mf)*rho_s*g; %Time-average pressure drop across the bed

%% for t>0 %%
XiBub = 0.80; % Bubble fraction
err=1; dp=[dp,dp]; deltaB=DBgoal;
while err > 0.01
XiBub = DBgoal/deltaB*XiBub;
XiEmul = 1-XiBub;
Gast = 1; DP = Ptarget;
while Gast > 0.01
if Gast > 1
dp(1) = dp(1)*2;
else
dp(1) = dp(1)*DP/Ptarget;
end
emax(1) = ptry(my_air,rho_air,dp(1));
emax(2) = ptry(my_air,rho_air,dp(2));
Grind = 1; Umean = u_wanted;
while Grind > 0.01
i=1; t=0; x=0; u1=umf; u2=umf; um=umf; Hast=0;
P_ply=P_ply(1); e1=e_mf; e2=e_mf;
alpha = alpha*u_wanted/Umean;
while x(i)<hb0
i = i+1;
t(i) = t(i-1)+deltat;
x(i) = x(i-1);
P_ply(i) = P_ply(i-1) + alpha*deltat*(u_wanted - um(i-1));
Skillnad = 1;
while Skillnad > 1e-5

iii
u1(i) = fsolve(@pelle, u1(i-1),
optimset('Display','off'));
Gamma = 150*(1-e1(i))^2*my_air/e1(i)^3/(dp(1))^2;
Psi = 1.75*(1-e1(i))*rho_air/e1(i)^3/(dp(1));
Acc = (Gamma*u1(i)+Psi*u1(i)^2)/rho_s/(1-e1(i))-g;
X = x(i-1) + Hast(i-1)*deltat + Acc*deltat^2/2;
Skillnad = abs(x(i) - X);
x(i) = X;
end
if x(i)<=-1e-4 % safety measure when ratio is below one
disp('Negative X!'); keyboard
end
Hast(i) = 2/deltat*(x(i)-x(i-1))-Hast(i-1);
u2(i) = fsolve(@palle,u2(i-1),optimset('Display','off'));
um(i) = u1(i)*XiBub + u2(i)*XiEmul;
end
Umean = mean(um);
Grind = abs(Umean-u_wanted)/u_wanted;
end
Pdist = k*(u1.^2*XiBub + u2.^2*XiEmul);
Pbed = P_ply - Pdist; DP = mean(Pbed);
Gast = abs(DP-Ptarget)/Ptarget;
end
deltaB = mean(XiBub*x/hb0);
err = abs(deltaB-DBgoal)/DBgoal;
end
toc
Ptarget = Ptarget*ones(1,length(t)); ratio = mean(Pbed)/mean(Pdist);
Dp=dp; Alpha=alpha; Emax=emax; Xpos=x; Tid=t; XiB=XiBub;
U=[u1; u2; um]; P=[P_ply; Pbed; Ptarget]; Eps=[e1; e2];

%% --- Subfunctions --- %%
function bd = sandra(e)
global Repmf Ar
bd = (1.75*Repmf^2/(e^3) + 150*(1-e)*Repmf/(e^3)-Ar);

function bb = pelle(urban)
global i hb0 x k P_ply dp e1 my_air rho_air emax
e_f = 0.2104*urban+0.37325;
e1(i) = min(e_f,emax(1));
A = 150*(1-e1(i))^2*my_air/e1(i)^3/dp(1)^2;
B = 1.75*(1-e1(i))*rho_air/e1(i)^3/dp(1);
bb = (hb0-x(i))*(A*urban+B*urban^2)+k*urban^2-P_ply(i);

function bc = palle(urban)
global i hb0 k P_ply dp my_air rho_air e2 emax
e_f = 0.2104*urban+0.37325;
e2(i) = min(e_f,emax(2));
A = 150*(1-e2(i))^2*my_air/e2(i)^3/dp(2)^2;
B = 1.75*(1-e2(i))*rho_air/e2(i)^3/dp(2);
bc = hb0*(A*urban+B*urban^2)+k*urban^2-P_ply(i);

function [Emax]= ptry(my_air,rho_air,dp)
Ugas = linspace(0,5,1e5);
E = 0.2104*Ugas+0.37325;
A = 150*(1-E).^2*my_air./E.^3./dp.^2;
B = 1.75*(1-E)*rho_air./E.^3./dp;
DPDX = A.*Ugas + B.*Ugas.^2;
toppen = max(DPDX);
i = find(DPDX == toppen);
Emax = E(i(1));

Potrebbero piacerti anche