Chemical Engineering Science 62 (2007) 550 – 560

www.elsevier.com/locate/ces

Dynamics of furnace processes in a CFB boiler

Andreas Johansson a,b , Filip Johnsson a,∗ , Fredrik Niklasson a , Lars-Erik Åmand a
a Department of Energy Conversion, Chalmers University of Technology, SE-412 96 Göteborg, Sweden
b University College of BorAAs, SE-501 90 BorAAs, Sweden

Available online 14 September 2006

Abstract
With the aim of understanding the dynamics of combustion, this work examines simultaneous fluctuations in fluid dynamic parameters and
gas composition measured in a CFB furnace operated with coal as a fuel. The fluid dynamic parameters investigated are pressure and air flow
entering the furnace. Gas composition was recorded by a zirconia-cell probe and a gas suction probe connected to a mass spectrometer having
a high time resolution (10 Hz). The principal fluctuations detected are around 1 Hz and below 0.3 Hz. The fluctuations below 0.3 Hz mostly
originate from variations in the fuel-feed rate. These variations create periods of reducing conditions caused by a momentarily high fuel input
accompanied by a pressure rise in the furnace and a reduction of the air feed, which occurs concurrent with the release of an enhanced quantity
of volatiles. Modelled pressure fluctuations based on the relation between volatile release and pressure in the furnace give similar pressure
fluctuations as the measured pressure fluctuations, with respect to amplitude and characteristic time scale of the fluctuations. There is also a
correlation between reducing conditions and the concentration of hydrocarbons.
䉷 2006 Elsevier Ltd. All rights reserved.

Keywords: Circulating fluidized bed; Combustion; Gas composition; Dynamics; Hydrodynamics; Fluidization

1. Introduction
During fluidized-bed combustion, constant fuel supply over
time and even distribution of fuel over the furnace cross-section
are desired to avoid unwanted variations in gas concentration
within the furnace. This may be difficult to ensure, particularly
during combustion of fuel with a significant content of volatiles,
because of rapid devolatilization compared to dispersion in
a fluidized bed. It has been reported that large fluctuations
between reducing and oxidizing conditions occur during oper-
ation with high-volatile fuels in fluidized beds (e.g. Niklasson
et al., 2003). Such variations in gas composition retards con-
version of gaseous fuel components and displaces it to down-
stream regions of the furnace, which may influence burn-up
and cause undesirable emissions.
In addition to mixing problems related to release of volatiles,
the dense fluidized bed present in the bottom part of many
designs of CFB boilers could be a source of the changes
between reducing and oxidizing conditions. The pressure in
such
∗ Corresponding author. Tel.: +46 31772 1449; fax: +46 31772 3592.
E-mail address: filip.johnsson@me.chalmers.se (F. Johnsson).
0009-2509/$ - see front matter 䉷 2006 Elsevier Ltd. All rights reserved.
doi:10.1016/j.ces.2006.09.020
a bed oscillates due to the bubble flow through the bed which
may affect combustion. Imperfections associated with large-
scale combustion conditions, such as small variations in opera-
tion, maldistribution of the fuel over the furnace’s cross-section
and irregularities in the external solids flow could also be the
origin of fluctuations in the gas composition. Pressure variations
in the furnace can propagate through the air distributor down
into the air-feed system and cause fluctuations in the air feed
(e.g. Svensson et al., 1996; Johnsson et al., 2002; Sasic et al.,
2004), especially at a low pressure drop across the air distrib-
utor, typically found in CFB boilers.
To optimize the performance of boilers and to develop com-
bustion models, it is essential to better understand the origin
of the fluctuations in the furnace and the air-feed system. This
requires simultaneous information on the fluid dynamic and
combustion processes. The related research is hitherto limited
to a few works. Stubington and Chan (1990) tried to find a cor-
relation between bubbles and high oxygen concentration in a
bed, and Lin et al. (2004) investigated the impact of uniformity
of fluidization on the generation of organic pollutants (BTEXs
and PAHs). However, both these investigations were concerned
with bubbling beds and furthermore, the expected correlations
 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560
were not found in any of these works. Jang et al. (2003)
observed that the properties of pressure fluctuations, such as
mean pressure and standard deviation, varied depending on
the composition of the fuel, but those quantities were not
connected to fluctuations in gas concentration.
The present work is an experimental study on the link
between fluid dynamic processes and combustion in the fur-
nace of a 12 MWth CFB boiler. The study covers also the
influence of the air and the fuel-feed system. The parameters
measured include fluctuations of pressure at various locations
in the furnace and in the air-feed system, representing the fluid
dynamics, and simultaneous variations in local gas concentra-
tion, representing combustion processes. In addition, variations
in the fuel-feed rate were measured and a conceptual model of
the relation between release of volatiles and in-furnace pres-
sure fluctuations is given. A comparison is made with results
from similar measurements in a 235 MWe CFB boiler.
2. Experiments
Fig. 1 shows the measurement positions in the 12 MWth CFB
boiler and the positions of the input of air and fuel. The fur-
nace has a cross-section of 1.7 × 1.4 m and a height of 13 m
and is further described by Leckner et al. (1991). The part
of the fuel-feed system investigated in this work consists of
a conveyer (scrape feeder) containing densely spaced scrapes,
which transport the fuel from the fuel hoppers to a cell feeder
and then to the fuel-feed chute, where the fuel is blown into
Fig. 1. Examples of power spectral densities at various positions in the furnace in the reference case and the case without a bottom bed.
551
the furnace, assisted by air to improve dispersion. The conveyer
is inclined by around 40◦ as the hopper is situated below the
cell feeder. At normal load the interval between discharges of
scrapes is around 1 s. The cell feeder, consisting of four cells
and rotating with 36 revolutions per minute, distributes the fuel
of each scrape into the chute before the discharge of the next
scrape into the cell feeder.
The origins of the fluctuations were investigated by alter-
ing the solids inventory, corresponding to the different furnace
pressure drops: 7.5, 3.0 and 2.0 kPa. Except for the furnace
pressure drop all other parameters were kept constant. In the
reference case the solids inventory corresponded to the normal
operating conditions of the boiler (i.e., with a total pressure
drop over the furnace of 7.5 kPa). For the case with the lowest
furnace pressure drop (lowest solids inventory) measured, ver-
tical pressure drop profiles indicated that there was hardly any
bottom bed present in the furnace (which is the case for nor-
mal operating conditions). The boiler was operated at the loads
listed in Table 1. In all cases the amount of excess air was 15%.
Silica sand was used as bed material, except in the case with-
out a bottom bed (2 kPa) where the bed material was changed
to limestone. The fuel was for all cases bituminous coal con-
taining 29% volatiles, 8% moisture, and 9% ash (calculated on
coal as delivered). The fuel was ground and the size of the par-
ticles was less than 20 mm (size fractions of some samples are
given in Table 2).
The fuel feed was stopped during 6 min in the reference case
as well as in the case with low bed height to investigate its
552 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560

Table 1
Operating conditions

Case Boiler load (MW) Pressure drop (kPa) Primary/total air* (dimensionless) Temperature (◦ C) Particle size (mm)

Reference 6 7.5 0.6 850 0.3

Low bed 6 3.0 0.6 850 0.3
No bed 7.5 2.0 0.6 890 0.8
∗ The total air flow is 2.7 kg/s.

Table 2
Particle size distribution and mass for some individual scrapes/batches of the
fuel feed
Scrapes Mass Particle size distribution (%)
(g) > 8 mm > 4 mm > 1 mm
1 334 35 61 82
2 21 58 78 87
3 814 21 48 75
influence on the fluctuations. The impact of the fuel-feed rate
was studied separately. The fuel was then by-passed the fur-
nace, close to the position where the conveyer discharges the
fuel into the furnace, and directed into a container resting on a
scale. Variations in fuel-feed rate with time was obtained from
variations in the weight of samples collected by the container.
The total time for the fuel-feed measurement was 6 min. By
operating the conveyer manually it was possible to collect sam-
ples of fuel carried by single scrapes. These were weighed and
sieved. All samples were separated into 16 fractions with diam-
eters varying from 0.016 m to 45 m. Afterwards each fraction
was weighed by a precise scale (with an accuracy of ±0.05 g).
The size distributions of the withdrawn samples are used as
inputs to the modelling of mass release.
Pressure fluctuations were measured against the atmosphere
through pressure taps mounted in four positions on the wall: in
the windbox (pwb ), just above (0.03 m) the air nozzles (pf b ),
at 0.99 m above the nozzles (p0.99 ) and close to the inlet of
the cyclone (pci ). Simultaneously, the air flows (primary, sec-
ondary and total) were obtained from venturi meters together
with the oxygen concentration in the flue gas duct and the
signals from probes inserted in the furnace. The latter devices
include zirconia-cell probes, a pressure probe and a gas suction-
probe connected to a mass spectrometer. All signals were
sampled at a frequency of 40 Hz. The response time of the pres-
sure transducers including tap and tubing is less than 0.01 s,
i.e., considerably shorter than the fluctuations studied in this
work. The mass spectrometer was originally designed to mea-
sure constituents of exhaust gas in real time (in the automotive
industry) and has a high time resolution. Its performance has
been reported by Dearth (1999). Although the mass spectrom-
eter allows on-line time-resolved quantification of several gas
compounds, in the present tests it was calibrated to record only
one gas component in order to increase time resolution up to
10 Hz (benzene is reported here). The gas suction probe and
the gas conveying line to the gas analysers were maintained
at a temperature high enough to avoid condensation of wa-
ter vapour (i.e., > 100 ◦ C) and equipped with a ceramic filter
to prevent entrainment of solids. The gas suction probe was
located at 3.7 m during all measurements. The zirconia-cell
< 1 mm
probe, previously described by Lyngfelt et al. (1993), indicates
18 whether the environment at the probe tip is oxidizing or reduc-
13 ing (peaks in the signal of the zirconia cell represent reducing
25 conditions). The pressure probe applied is the same as was the
previously denoted “momentum probe” (Zhang et al., 1997).
It allows the addition of purge-air and has probe tips that end
in vertical pipes oriented in opposite directions (each tip has a
length of 0.02 m). During the present tests, the pressure was
measured in each pipe against atmosphere. Results from the
measurement with the pressure probe are shown in Figs. 2a
and b. Pressure signals in figures other than Fig. 2 derive from
wall mounted pressure taps.
The zirconia-cell and pressure probe measurements were
mainly performed at three heights: 0.57, 3.7 and 7.9 m above
the primary-air nozzles (in the bottom of the furnace). For the
simultaneous zirconia cell and gas suction measurements at
3.7 m, the probes were inserted so that the tips of the probes
were in contact in the centre of the furnace. Then the probes
were withdrawn to make the distance between the tips 100 mm.
In some measurements two zirconia-cell probes were inserted
at the same height with a horizontal distance between the tips
of 1 m, located at a distance of 0.20 m from the furnace wall.
To prevent air in-leakage a wet cloth was applied around the
probes in the measurement ports.
The signal of the zirconia cell has steep characteristics and
the signal is therefore more or less binary, showing either ox-
idizing or reducing conditions. In order to better visualize the
co-existence of fluid dynamic (pressure) and combustion pa-
rameters, the pressure signals in the furnace are also converted
to a binary form. The pressure in the bottom of the furnace
and the total airflow (Figs. 11 and 12) were transformed into a
binary signal p̂, applying
 
p(t) − p
p̂ = 0.5 ∗ +1 , (1)
|p(t) − p  |
where p  is the value above which the signal is set to 1 and p(t)
is the signal. The value p  was chosen to achieve the highest
possible correlation with the zirconia-cell signal.
The interaction between the input airflow and the pressure
fluctuations in the furnace was recorded also in a 235 MWe
 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560 553

0.6 h = 0.0 m

Normalized cross-correlation [-]

h = 1.0 m
0.4 h = 7.9 m
h = 10.9 m

0.2

-0.2

-0.4
-3 -2 -1 0 1 2 3
(a) Time lag [s]

0.6
Normalized cross-correlation [-]

h = 0.0 m
h = 0.55m
0.4 h = 1.0 m
h = 10.9m
0.2

-0.2

-0.4
-3 -2 -1 0 1 2 3
Time lag [s]

Fig. 2. Cross-correlation between: (a) the pressure in the windbox and the pressure at various heights in the furnace, (b) the pressure obtained by the upward
pointing tip of the pressure probe inserted at h = 0.57 m and the pressure at different positions in and above the dense bed (h = 0.55 m refers to the downward
pointing probe tip).

CFB boiler, located in Bogatynia, Poland (see Goidich and by the common empirical relation:
Hyppänen, 2001 for further information about the boiler).
There, measurements were conducted with a sampling fre- tdev (Rp ) = a · Rpb , (2)
quency of 200 Hz during one operational load and with brown
where Rp is the radius of the fuel particle and a and b are
coal as fuel. The air-feed system of the 235 MWe unit consists
empirical coefficients. For bituminous coal, a = 16.500 and
of two separate air ducts, ending in three windboxes. The fluc-
b = 1.5 (Stubington and Moss, 1995; Winter, 1995; Ross et al.,
tuations in airflow supplied to the furnace were recorded by a
2000). In addition, according to Thunman and Leckner (2001),
Prandtl tube inserted into one of the air-feed ducts. A pressure
the mass release over time for a certain size fraction can be
sensor was connected to the Prandtl tube, giving a voltage sig-
calculated as
nal that is linearly proportional to the dynamic pressure in the
 
duct. The pressure in the furnace was measured simultaneously (1 − ) 3/b
by taps mounted at the wall 0.25 m above the air distribu- x(, Rp ) = , 0  tdev (Rp ), (3)
tdev (Rp )
tor. The results of these measurements are shown in Figs. 4b
and 14. All other measurement data given in the paper are where  is the time from the start of devolatilization. The size
from the 12 MWth boiler. distribution of the fuel was recorded in 13 scrapes/batches and
the mass flow of fuel was measured during 8 min (480 batches
in total). The measured fuel mass in each of the 480 batches
3. Modelling
was used as input to the modelling. They were given the same
average size distribution with corresponding mass release of
The low-frequency pressure fluctuations are modelled based
volatiles. To simplify the analyses, the cumulative release of
on the assumption that the pressure in the furnace is propor-
volatiles with time for a given size distribution was fitted to a
tional to the gas produced in the furnace. Although simple, the
simplified distribution:
model should be detailed enough for testing the hypothesis on
variations in the release of volatiles being a source to fluctua- y(t) = c · (1 − exp(−t d · e)), (4)
tions in the furnace pressure. The model requires fuel proper-
ties, size distribution of the fuel and mass flow of fuel as inputs. where the constants c, d and e are found by fitting the curve
The time for release of volatiles and water vapour is estimated y(t) to the curve formed by Eq. (3). Average values of d (0.66)
554 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560
and e (0.74) for the 13 recorded batches are used for all batches,
whereas c is given by the mass of volatiles and water vapour
in each batch. Differentiating Eq. (4) with time yields that the
mass release over time for a given size distribution can be
described as
jy
= c · exp(−t d · e) · (−d · t d−1 · e). (5)
jt
Even though the fuel enters the cell feeder in batches it will be
slightly distributed over time before it reaches the furnace (due
to fuel chute, additional air, etc.). This distribution is assumed
to follow
X(t) = (t/ts )0.5 · (1 − (t/ts ))0.5 , 0 t ts .
The time needed for all fuel particles to reach the furnace is set
to 1 s (i.e., ts = 1).
If the amount of solids in the freeboard is regarded as con-
stant, the pressure drop between the dense bed and the outlet
of the combustor can be calculated as
p = 1
2 · air · C 2 · ,
where  is an equivalent resistance coefficient for the gas parti-
cle flow, here denoted as friction coefficient. f is the density
of air at 850 ◦ C and C is the average gas velocity, which is
calculated as
ṁair + ṁf
C= . (8)
air · AF
If the moisture and the volatiles are assumed to be released
in the splash zone, this release yields the gas velocity from
Eq. (8) using the fuel feed as input (ṁf ). Thus, the variation
in gas velocity is obtained from corresponding variations in the
amount of fuel-feed based on the above described weighting
of the content of individual scrapes by collecting samples of
fuel carried by single scrapes in the fuel conveyer. Combining
Eqs. (7) and (8) with the measured pressure drop between the
1 m height and the cyclone inlet (0.6 kPa) gives that the resulting
friction coefficient will be 280 for the gas velocity employed.
If acceleration effects are neglected, the friction coefficient can
be considered to be constant and thus, the changes in mass
release over time give rise to proportional pressure changes.
4. Results and discussion
4.1. Fluctuations
Fig. 1 presents power spectra of pressures measured in the
windbox, at the bottom of the furnace and in the cyclone in-
let of the 12 MWth boiler under two operation conditions: the
reference case and the case without a bottom bed. The power
spectrum of the differential pressure across the total-air venturi
is also shown. In the reference case with a bottom bed present,
the power spectral density is found mainly in two frequency re-
gions: around 1 Hz and below 0.3 Hz. The peak around 1 Hz in
the frequency spectrum has previously been attributed to bub-
ble passage (e.g. Johnsson et al., 2000), although it is less pro-
nounced compared to the region below 0.3 Hz (but yet present,
1
pci < > pwb
Coherence [-]
0.8
pfb < > pwb
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Frequency [Hz]
Fig. 3. Coherence between the pressure in the windbox and at the cyclone
inlet or in the bottom bed.
(6)
as shown in Fig. 3). Without a bottom bed there are significantly
less fluctuations, even if the low-frequency part of the spectrum
remains to some extent. Fluctuations below 0.3 Hz were found
also for the signals from the mass spectrometer and the zirco-
nia cell. Such low-frequency fluctuations have been assumed
to be a consequence of variations in operating conditions, but
they have not been associated with any relevant process in the
(7) furnace.
Simultaneous pressure measurements in the air-feed system,
in the windbox and in various positions inside the furnace show
that these signals are more or less correlated. This is indicated
in Fig. 2a where the pressure in the windbox is cross-correlated
with the pressures at various heights above the air nozzles in
the reference case. The side-lobes of the cross-correlation curve
are caused by the 1 Hz pressure fluctuations. The highest cross-
correlation between the pressure signal in the windbox and the
signals in the furnace is in the freeboard positions. The corre-
lation with the windbox pressure becomes poor when the pres-
sure signal originates in the dense bottom bed (at 0.0 m). The
poor correlation could be an effect of that a single pressure tap
at the bottom of the dense bed cannot display all movements
in the bed and therefore mainly captures the pressure fluctu-
ations in the vicinity of the wall where the tap is mounted.
This explanation is also in line with the findings of Pallarès
et al. (2006), who showed that in the bottom part of the furnace
neither the time-averaged nor the fluctuating in-bed pressure
measured at the wall of a large fluidized bed can be assumed
to represent the whole cross-section of a fluidized bottom bed.
This result might explain the lack of correlation between fluid
dynamic parameters and gas composition in the investigations
by Stubington and Chan (1990) and Lin et al. (2004). Explic-
itly, they only measured in the bottom bed where it is difficult
to get a representative measure of the pressure, due to cross-
sectional heterogeneity of the pressure field.
Fig. 2b shows the cross-correlations between the pressure in
the tip of the pressure probe pointing upward correlated against
the pressures measured in the downward tip and the wall pres-
sures at 0.0, 1.0 and 10.9 m height. The probe was inserted at
0.57 m height. The only significant correlation is with the pres-
sure in the downward tip (h = 0.55 m). This is expected as the
distance between the tips of the probe is short. It is notewor-
thy that this cross-correlation is only slightly higher than that
between the pressure in the windbox and at the cyclone inlet
in Fig. 2a. The result confirms that pressure taps mounted in
 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560 555

the wall in the bottom bed cannot be considered as revealing
of all pressure fluctuations in the bed. This conclusion is fur-
ther supported by Fig. 3, which plots the coherences between
the windbox pressure and the pressure in the bottom of the
furnace (h = 0.0 m) and at the cyclone inlet (h = 10.9 m). As
can be seen, only the 1 Hz frequency region changes signif-
icantly with measurement position and no obvious difference
is seen for the 0.3 Hz frequency region. This indicates that the
low-frequency fluctuations (those below 0.3 Hz) are caused by
rather global pressure fluctuations, which can be detected in
the entire furnace, whereas the 1 Hz fluctuation region is at-
tributable to more local phenomena (such as bubble passage).
In addition, a comparison of the coherence between the fluc-
tuations in the air-feed system and the pressure in the lower
1 wall the bubble passes).
0.8 pwb < > tot. airflow
Coherence [-]
part of the splash zone in the 12 MW boiler (h = 0.99 m) and
235 MWe boiler (h = 0.25 m) is shown in Fig. 4. The coher-
ence in the low-frequency domain covers a broader spectral
range in the 235 MWe boiler than in the 12 MW boiler. Pres-
sure drop measurement indicates that the former boiler is oper-
ated without a dense bottom bed (see Johansson et al., 2006).
This should be part of the explanation for the absence of a re-
gion in the frequency spectrum of the pressure in the 235 MWe
boiler which can be devoted to bubble passage (such as a clear
peak around 1 Hz). However, even if there would have been a
dense bed with clear bubbles in the 235 MW boiler the wide
cross-section would result in many bubbles distributed over the
cross-section. Consequently, wall-mounted pressure measure-
ments would probably not yield a clear peak in a spectrum due
to the interaction of the bubbles as well as that wall measure-
ments are influenced differently depending on how far from the
The essential conclusion from Figs. 1–4 is that two frequency

p0.99 < > tot. air flow regions dominate the fluctuations in the 12 MW boiler: one
0.6
originating from bubbles in the dense bed (1 Hz), and one found
0.4
at low frequencies (< 0.3 Hz). The source of fluctuations in
0.2 the low-frequency region is analysed below. To better visualize
0 these low-frequency fluctuations, the time series of pressure
0 0.5 1 1.5 2 2.5 3 3.5 4 fluctuations are filtered with a moving average of 1 s (40 points).
(a) Frequency [Hz]

1
4.2. Fluctuations from the fuel feeding
0.8 pwb < > tot. airflow The influence of the fuel feeding on the low-frequency fluc-
Coherence [-]

p0.99 < > tot. airflow

0.6 tuations is illustrated in Fig. 5, where a time series of pressure
0.4 fluctuations in the case without a bottom bed is shown. High
amplitude, low frequency fluctuations are present both in the
0.2
bottom bed and close to the cyclone inlet. The character of
0 these fluctuations, especially at the cyclone inlet, changes when
0 0.5 1 1.5 2 2.5 3 3.5 4
the fuel feed is stopped. Thus, it can be concluded that the fuel
(b) Frequency [Hz]
feeding is accompanied by low-frequent fluctuations. Mea-
Fig. 4. Coherence between pressure in the air feed system and in the splash surements in the reference case showed that low-frequent fluc-
zone. (a) The 12 MW boiler. (b) The 235 MWe boiler. tuations may be produced by other sources as well (cf. Fig. 5).

Fig. 5. The pressure at the bottom of the furnace and in the cyclone inlet, with and without fuel feeding in the case of low bed height. The signals are filtered
with a moving average of 1 s.
556 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560

Fuel feed
3 Time-averaged fuel feed
Mass flow [kg/s]

0
0 20 40 60
Time [s]

Fig. 6. Variations in the fuel feed.

Table 3
Number of high inputs of fuel and the corresponding frequencies
Fig. 7. Calculated release of volatiles and moisture at a fuel input from one
Fuel inputs  2 × F̄F  4 × F̄F  6 × F̄F  8 × F̄F scrape of a typical size distribution as obtained from Eq. (3). The (overall)
No. (dimensionless) 158 61 25 7 average fuel input weighs around 0.3 kg and the content of volatiles and
Frequency (Hz) 0.44 0.17 0.07 0.02 moisture is around 40%. Mass release due to char combustion is excluded
in this plot.
The inputs are related to the average fuel feed, F̄F .

These could presumably originate from an irregular solids flow 0.25

from the loop seal, varying lateral displacement of the bed, Simulated pressuresignal
Pressure [kPa]

poorly fluidized zones, etc. 0

In Fig. 6, a time series of the fuel-feed rate from the conveyor
is shown together with its average value, as obtained from the -0.25
fuel-feed rate experiments explained above (when fuel was then
by-passed the furnace). As can be seen the discharge from the -0.5
fuel conveyer shows large variations in fuel batches, including
-0.75
a period of 8 s without any fuel input, and a period with a 0 20 40 60
peak flow rate of more than 10 times the average value. The Time [s]
periods of high flow rates occur at low frequencies, and as given
by Table 3, where the fuel peaks are counted, the higher the 0.25
fuel input, the lower the frequency of the peaks. Thus, these Measured pressure signal
Pressure [kPa]

variations in fuel flow should be the main reason for the low 0
frequency region in the power spectra of Fig. 1. The variation
-0.25
in fuel feed is due to the escape of small fuel particles from the
scrapes through the gap between the scrapes and the wall of -0.5
the conveyer. These particles are then caught by other scrapes,
where this gap is smaller (variations were observed) or sealed -0.75
by larger particles. The irregularities in the fuel-feed result in 0 20 40 60
variations in the gas production in the furnace, since drying and Time [s]
devolatilization occur rapidly. This affects the furnace pressure,
Fig. 8. Time series of simulated (Eq. (7)) and measured pressure fluctuations
which varies with gas production.
at 1 m height.
Table 2 shows typical variations between weighed and sieved
batches from three scrapes of the feeder. The sieving shows
that large differences are also present in the size distribution of
each fuel batch: the larger the sample, the larger the fraction of of fines, the faster the release of volatiles. The largest fuel input
fines. An attempt to illustrate the impact of the variations in the gives a release of 0.6 kg of volatiles during the first second. This
fuel feed is given in Fig. 7, which shows the calculated mass corresponds to more than 20% of the total mass flow through
release of volatiles (Eq. (3)) for the samples in Table 3 plotted the furnace and is high enough to create a temporary overall
versus time. Also shown in Fig. 7 is the release of volatiles sub-stochiometric condition in the furnace.
from the highest fuel input in Fig. 6, assuming the same size In Fig. 8, time series of the simulated pressure fluctuations
distribution as in Sample 3. As expected, the larger the fraction (Eq. (7)) at 1 m height are plotted together with the measured
 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560 557

Fig. 9. Simultaneous time series of two zirconia cell probes during the reference case. The horizontal distance between the probe tips were 1 m (h = 3.7 m).

pressure at the same height. It shows that the simulations give

pressure fluctuations which qualitatively resemble the measured
signal with respect to frequency and amplitude of fluctuations.
It should, however, be pointed out that the comparison is not
completely straight forward, because it is not possible to carry
out simultaneous measurement of fuel-feed rate and pressure
fluctuations for the boiler used (the variations in fuel feed, used
as input in the simulations, were obtained when the boiler was
out of operation, as described above).
The link between pressure and a high release of moisture
and volatiles are shown in Fig. 9 in the form of time series of
simultaneous measurements with the two zirconia-cell probes
at 3.7 m, with their tips separated by a distance of 1 m, and
with the pressure measurements above the primary air distrib-
utor (h = 0.03 m). The zirconia-cell signals are made binary by
Eq. (1) and then added (the vertical axis indicates the number
of zirconia cells recording reducing conditions (0, 1 or 2)). In
this comparison reducing conditions are interpreted as a mea- Fig. 10. Time series of the zirconia cell signal at 7.9 m and the discretized
sure of high release of volatiles. Such conditions were often pressure at 0.0 m in the furnace.
registered at the same time by both cells, which implies that the
presence of regions with high concentrations of volatiles were
of considerable lateral extension and rather global, as the width signal, even though the zirconia-cell probe only measures in
of the furnace is 1.4 m. Reducing conditions of this kind could one position and thereby gives an incomplete picture of the gas
not be due to a maldistribution of fuel over the cross-section composition over the cross-section. The reason for the smaller
(as a maldistribution that covers most of the cross-section is no number of occasions of reducing conditions than the number
maldistribution). Instead, they should be due to irregularities in of pressure peaks is probably that some of the reducing (fuel
the fuel feed as discussed above. Fig. 9 shows that the pressure rich) regions are burnt out while passing through the furnace.
at the bottom of the furnace is high when globally reducing Considerably more reducing periods are detected in the lower
conditions are detected (highlighted by the horizontal brackets position (3.7 m) as can be seen in Fig. 11a. Fig. 11 also shows
in Fig. 9), whereas it is low when the environment is oxidizing a correlation between the zirconia cell and the air flow enter-
or when only locally reducing conditions are present. This con- ing the furnace (Fig. 11b). As in the previous figures, the signal
firms that there is a significant correlation between the signal of the air flow is filtered and made binary (here, a peak in p̂
of the zirconia cell and the furnace pressure. corresponds to a dip in the air feed). The correlation is poor
The connection between the reducing condition and pres- in the lower position (Fig. 11a) but much better in the upper
sure can also be seen in Fig. 10, which shows time sequences position (Fig. 11b), where almost all reducing conditions oc-
of pressure at 0.0 m and the signal of a zirconia cell at 7.9 m. cur simultaneously with low air-feed flow. The lack of corre-
The pressure signal was filtered and then made into a binary lation in Fig. 11a should follow from reducing conditions in
signal, using a moving average of 3 s and applying Eq. (1). the lower position that mostly originate from maldistribution of
The increased pressure at the bottom of the furnace often co- fuel or air over the cross-section of the bed. Most likely such
incides with reducing conditions, i.e., with a high zirconia-cell reducing regions have mixed to a large extent with air before
558 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560
Fig. 11. Time series of the zirconia cell signal and total air flow in the reference case at two levels in furnace. (a) 3.7 m, time fraction of reducing conditions
is 0.099. (b) 7.9 m, time fraction of reducing conditions is 0.012.
zirc. cell, pfb
Normalized coherence [-]
0.8 zirc. cell, pwb
zirc. cell, pci
zirc. cell, tot. air flow
0.6
0.4
0.2
0
0 0.5 1 1.5
Frequency [Hz]
Fig. 12. Coherence between the zirconia cell and various pressures and the
total air flow in the reference case. The zirconia cell probe was applied at
h = 3.7 m.
reaching the 7.9 m level, leaving only reducing regions from
occasionally high fuel input. Fig. 12 shows a high resemblance
between the signals of the zirconia cell, the pressure, and the
air flow at low frequencies (cf. Table 3), which confirms the
connection between reducing conditions, pressure peaks and
a low air feed. Similar relationships were found at all loads
investgated.
The course of events caused by the irregular fuel feed is
schematically summarized in Fig. 13, illustrating the correlation
between the pressures in the furnace and in the windbox, as well
as between the air flow entering the furnace and the oxygen
concentration. The pressure in the furnace, and consequently
the pressure in the windbox, follows the release of volatiles and
moisture. The increased pressure reduces the air flow into the
furnace. The decrease in airflow coincides with a high release
of volatiles and, as a result, the conditions become reducing
and the oxygen concentration in the back pass decreases. The
irregularities in the fuel feed are, of course, an effect of the
design of the conveyer and unique for the specific boiler applied
in this work. In the particular case studied here, the scrapes were
more or less worn out, but this condition could be used to study
the link between variations in fuel feed and pressure drop in the
2
Fig. 13. A qualitative sketch of time series showing the effects of the low
frequency fluctuations.
furnace. Normally, for the fuel used (coal) the fluctuations in
fuel feed should be less than observed in this work. Yet, several
more inhomogeneous fuels such as waste fuels and biomass
are known to exhibit considerable fluctuations with respect to
the heating value vs time. Also such variations are known to
cause fluctuations in furnace pressure, although the reason not
necessarily being fluctuations in fuel-feed rate.
Similar low-frequency fluctuations and similar links
between pressure and air feed as reported above are seen in
the 235 MWe boiler, as shown in Fig. 14. The reason for these
fluctuations is not known, but it is believed that these will have
a similar impact on the combustion process as in the 12 MWth
boiler.
Simultaneous measurements of various parameters in the
12 MWth boiler show that reducing conditions are accom-
panied by high amounts of gas compounds, typically found
during pyrolysis, such as carbon monoxide and hydrocarbons.
The coexistence of such compounds and reducing conditions
were observed simultaneously by the mass spectrometer and
the zirconia cell, as exemplified in Fig. 15, where the zirconia-
cell signal is filtered with a moving average to make the char-
acter of the signal more analogous to the signal from the mass
 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560 559

Fig. 14. Pressure obtained in the splash zone and in the air feed system of the 235 MWe boiler.

Fig. 15. Time series of concentration of benzene and reducing conditions in the high load case (h = 3.7 m).

spectrometer. The agreement between reducing conditions be kept as low as possible in order to maintain high boiler
measured by the zirconia cell and benzene measured by the efficiency.
mass spectrometer supports the validity of both measurement
methods. Notation

5. Conclusions a, b, c, d, e fuel dependent empirical coefficients, di-

mensionless
There is an interaction between the fluctuations in the fluid C air velocity, m/s
dynamic parameters and in the gas composition in a CFB boiler. p(t) pressure, Pa
The signals studied show two dominant frequency regions of p threshold level for binary signal (Eq. (1)),
fluctuations, around 1 Hz and below 0.3 Hz. The former is re- Pa
lated to bubbles, as shown in previous work, and the latter p̂ binary pressure signal, Pa
frequency region should be due to variations in the fuel feed, Rp radius of coal particle, m
affecting the gas production in the furnace and thereby also tdev devolatilization time, s
the pressure. Modelling the relation between volatile release x relative volatile content, dimensionless
and in-furnace pressure gives the right order of pressure fluc- X fuel distribution, dimensionless
tuations with respect to amplitude and characteristic time scale
Greek letters
of the pressure fluctuations, i.e., fluctuations with a frequency
from 0.3 Hz and below. Thus, variations in fuel-feed rate (or  friction coefficient, dimensionless
fuel composition) should be an origin for pressure fluctuations f density, kg/m3
in the furnace.  time for which a particle is subjected to de-
The fluctuations in pressure lead to variations in the air feed: volatilization, s
low air feed rate occurs simultaneously with high release of
Index
volatiles. This gives rise to reducing conditions of considerable
magnitude and duration, which requires enhanced excess air to ci cyclone inlet
prevent unburned fuel from escaping the boiler. Thus, variations fb bottom of the furnace
in fuel feed (mass and/or heating value per time unit) should wb windbox
560 A. Johansson et al. / Chemical Engineering Science 62 (2007) 550 – 560
Acknowledgements
This work is financed in part by the Swedish Energy Agency
and the research programme: Advanced CFB for clean and
efficient coal power (CLEFCO) of the research fund for Coal
and Steel. The assistance of the people operating the 12 MW
CFB boiler at Chalmers University of Technology is also
acknowledged.
References
Dearth, M.A., 1999. Evaluation of a commercial mass spectrometer for its
potential to measure auto exhaust constituents in real time. Industrial
Engineering Chemical Research 38, 2203–2209.
Goidich, S.J., Hyppänen, T., 2001. Foster wheeler compact CFB boilers for
utility scale. In: Geiling, D.W. (Ed.), 16th International Conference on
Fluidized Bed Combustion, ASME, New York, Paper FBC01-0099.
Jang, H.T., Kim, S.B., Cha, W.S., Hong, S.C., Doh, D.S., 2003. Pressure
fluctuation properties in a combustion of mixture of anthracite and
bituminous coal in a fluidized bed. Korean Journal of Chemical Engineering
20, 138–144.
Johansson, A., Johnsson, F., Leckner, B., 2006. Solids flow pattern in a
235 MWe circulating fluidized bed boiler. Chemical Engineering Science,
doi:10.1016/j.ces.2006.09.021.
Johnsson, F., Zijerveld, R.C., Schouten, J.C., van den Bleek, C.M., Leckner,
B., 2000. Characterization of fluidization regimes by time-series analysis
of pressure fluctuations. International Journal of Multiphase Flow 26,
663–715.
Johnsson, F., Larsson, G., Leckner, B., 2002. Pressure and flow fluctuations
in a fluidized bed—interaction with the air-feed system. Chemical
Engineering Science 57, 1379–1392.
Leckner, B., Golriz, M., Zhang, W., Andersson, B.-AA., Johnsson, F., 1991.
Boundary layer-first measurements in the 12 MW CFB research plant at
Chalmers University. In: Anthony, E.J. (Ed.), 11th International Conference
on Fluidized Bed Combustion. ASME, New York, p. 771.
Lin, C.L., Wey, M.Y., Cheng, H.T., 2004. Relationship between pressure
fluctuations and generation of organic pollutants with different particle size
distributions in a fluidized bed incinerator. Chemosphere 56, 911–922.
Lyngfelt, A., Bergqvist, K., Johnsson, F., AAmand, L.-E., Leckner, B., 1993.
Dependence of sulfur capture performance on air staging in a 12 MW
circulating fluidized bed boiler. In: Clift, R., Seville, J.P.K. (Eds.), Gas
Cleaning at High Temperatures. Blackie, Glasgow, p. 470.
Niklasson, F., Johnsson, F., Leckner, B., 2003. Local air ratio measured by
zirconia cell in a circulating fluidised bed furnace. Chemical Engineering
Journal 96, 145–155.
Pallarès, D., Johansson, A., Johnsson, F., 2006. Interpretation of dynamics of
pressure measurements in CFB boilers 19th International Conference on
Fluidized Bed Combustion, ASME, New York.
Ross, D.P., Heidenreich, C.A., Zhang, D.K., 2000. Devolatilisation times of
coal particles in a fluidized-bed. Fuel 79, 873–883.
Sasic, S., Johnsson, F., Leckner, B., 2004. Interaction between a fluidized bed
and its air-supply system: some observations. Industrial and Chemistry
Research 43, 5730–5737.
Stubington, J.F., Chan, S.W., 1990. The interpretation of oxygen-probe
measurements in fluidised-bed combustors. Journal of the Institute of
Energy 136–142.
Stubington, J.F., Moss, B., 1995. On the timing of primary fragmentation
during bituminous coal particle devolatilisation in a fluidized bed
combustor. The Canadian Journal of Chemical Engineering 73, 505–509.
Svensson, A., Johnsson, F., Leckner, B., 1996. Fluidization regimes in non-
slugging fluidized beds: the influence of pressure drop across the air
distributor. Powder Technology 86, 299–312.
Thunman, H., Leckner, B., 2001. Principals and models of solid fuel
combustion: fuel loading of a fluidized bed combustor. Report Department
of Energy Conversion, Chalmers University of Technology, Göteborg.
Winter, F., 1995. Single fuel particle and NOx /N2 O-emission characteristics
under (circulating) fluidized bed combustor conditions. Academic
Dissertation University of Technology, Vienna.
Zhang, W., Johnsson, F., Leckner, B., 1997. Momentum probe and sampling
probe for measurement of particle flow properties in CFB boilers. Chemical
Engineering Science 52, 497–509.