determination of bubble diameter and axial velocity for a polyethylene fluidized bed using x-ray...
TRANSCRIPT
www.elsevier.com/locate/powtec
Powder Technology 1
Determination of bubble diameter and axial velocity for a polyethylene
fluidized bed using X-ray fluoroscopy
Ian Hulmea, Apostolos Kantzasa,b,*
aDepartment of Chemical and Petroleum Engineering, University of Calgary, 2500 University Drive, Calgary, Alberta, Canada T2N 1N4bTomographic Imaging and Porous Media Laboratory, Calgary, Alberta, Canada T2N 1N4
Received 10 July 2003; received in revised form 15 July 2004; accepted 10 August 2004
Available online 2 November 2004
Abstract
Gas–solid fluidized bed reactors have many industrial applications and have been studied extensively. Yet despite the amount of research
on the subject there is considerable disagreement to their behaviour. Part of the disagreement is due to the varied techniques used to examine
them. In this study we are examining, using non-intrusive techniques, the hydrodynamics of a partially scaled down reactor used in the
production of polyethylene. In particular, we are using digital fluoroscopy to determine the properties of bubbles that form in the reactor. The
goal of the research is to develop a tool that can be used to validate CFD models. The first part of the paper examines the tools used and
software developed to study the properties of the bubbles. The second part of the paper discusses results from experiments conducted using
glass beads (150 to 250 Am) and polyethylene particles (mean diameter of 850 Am and density of 650 kg/m3 with a wide particle size
distribution), which were used to validate the developed experimental technique and software using correlations in literature.
D 2004 Elsevier B.V. All rights reserved.
Keywords: Gas–solid; Fluidized bed; X-rays; Bubble diameter; Bubble axial velocity
1. Introduction
Due to the contact between the solid and gas, gas–
solid fluidized bed reactors are ideal for processes that
require inter-phase contact. These reactors find many
applications for coating, drying, adsorption, and catalytic
reactions, to mention a few. Despite their widespread use,
and the large amount of literature dealing with these
reactors, they are not fully understood. As a result, many
of the design criteria used for these reactors are empirical.
Also, there are problems when scaling these reactors.
Because most of the research is conducted on laboratory
scale reactors, some people question the usefulness of
these results for industrial reactors. Yet, despite the
criticism of laboratory scale data, this data remains one
of the only means for validating computational fluid
0032-5910/$ - see front matter D 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.powtec.2004.08.008
* Corresponding author. Tel.: +1 403 2205752; fax: +1 403 2825060.
E-mail address: [email protected] (A. Kantzas).
dynamic models, which is considered by many to be the
next design tool.
There are many interesting properties that can be
measured in these reactors. However, the flow of gas,
whether in the form of bubbles or emulsion, is one of the
most important factors determining the behaviour of a
fluidized bed [1]. Many design parameters, in gas–solid
fluidized bed reactors, require knowledge of the bubble
diameter. The mass and heat transfer coefficient of Davidson
and Harrison [2], Kunii and Levenspiel (1969), Sit and
Grace (1981), and Peters et al. (1982) cited in Hatzantonis et
al. [3] all require the bubble diameter and are important
design equations. As a result, many attempts have been
made to understand the behaviour of the gas phase, and in
pa rticular the bubbles, starting from the btwo-phaseQ theoryof Toomey and Johnstone [4], the Davidson theory for
bubble movement (1961), and the Geldart classification of
fluidized particles [5].
Early work on visualizing bubbles centred on two-
dimensional columns. However, due to the narrow width,
47 (2004) 20–33
Fig. 1. Picture of the experimental setup.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 21
wall effects are considerable and the applicability of the
results to a three-dimensional column is questionable. While
the emphasis shifted to developing experimental techniques
to determine the properties of bubbles in cylindrical
columns, most of the data available in open literature is a
result of studying a limited number of bubbles. Often
studies are conducted using probes inserted into the column,
which disturbs the shape of the bubbles. While non-
intrusive techniques have been developed to study the
behaviour of the gas phase, they have not been applied to
study the behaviour of a large number of bubbles in a
bubbling bed.
Due to the limited amount of information in literature
dealing with the measurement of bubble properties, and the
smaller amount of information on bubble properties for low
density particles, laboratory measurements were needed.
Not only were new measurements needed, but a new
technique needed to be developed that could capture and
analyze a statistically relevant number of bubbles. The result
was an experimental system, X-ray fluoroscopy, that can
capture real time footage of bubble movement, and software
that can analyse this data. The number of bubbles studied is
approximately 10,000 for each run. These experimental
results can be used as a tool to compare with the results
obtained from computational fluid dynamics. While other
parameters can be used, bubble size and velocity provide a
good basis for comparison of process conditions and model
parameters.
2. Experimental setup
2.1. General configuration
The experiments were conducted on a scaled down gas–
solid fluidized bed reactor used for the production of
polyethylene. The reactor is 0.1 m in diameter and 1.0 m in
height with a disengagement zone. Fig. 1 shows the
experimental setup with the column outlined.
The column is made of Plexiglas which allows for
visual confirmation of what is observed used X-ray
fluoroscopy. The vessel was fitted with a porous plate
distributor through which air, at ambient conditions, was
passed at varying superficial gas velocities. Gas ve-
locities ranged from 2.5 to 3.5 times minimum fluidiza-
tion for the polyethylene solid bed (minimum fluidization
velocity is 7.7 cm/s). The lower value of 2.5 times
minimum fluidization was chosen since bubbling
was not observed below this velocity and above 3.5
times minimum fluidization the bed was predominantly
slugging.
2.2. X-ray analysis
Images of bubble movement were obtained using X-ray
fluoroscopy. The generator was an MPX 100 with the
bucky (X-ray tube, table and intensifier) being a GE
RFXII. The tube operates from 60 to 120 kVp and an
image intensifier transforms the X-rays into visible light
for display. The field of view of the image intensifier is a
circle of diameter 0.135 m. As a result of the image size,
three different camera heights were required to capture the
movement of bubbles throughout the solids bed of
approximately 0.4 m in height. The images were captured
at a rate of 30 frames/s using a Matrox Meteor II frame
grabber for 2 min, and a high speed hard drive was used in
order to store the large amount of data resulting from the
images. Fig. 2 is a sample image obtained using X-ray
fluoroscopy.
Once the images were acquired they were processed off
line using software developed in house.
2.3. Bed material
The solid phase used in these experiments was a linear
low-density polyethylene (LLDPE), which has the same
appearance as laundry detergent. The sample used has a
wide particle size distribution ranging from particles less
that 100 Am, to particles larger than 1500 Am with a volume
mean diameter of 850 Am. The bulk density of the
polyethylene particles, measured in the laboratory, was
approximately 650 kg/m3 (Wright, 1999). The polyethylene
particles are porous and irregular in shape, as shown in
Fig. 3.
In addition to the LLDPE, experiments were also
conducted using glass beads to validate the software
Fig. 2. A sample image from X-ray fluoroscopy.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3322
developed for this work. The glass beads had a size
distribution ranging from 150 to 250 Am.
3. Software to analyze experimental results
In order to retrieve information from the images obtained
using fluoroscopy, software had to be developed. Two
different programs were written; one identifying the bubbles
and the other tracked the bubbles from frame to frame.
Identifying bubbles automatically required the use of image
processing tools, while tracking the bubbles required certain
brulesQ be developed to account for bubble actions.
3.1. Image processing goal
For this research, the objective is to develop an
algorithm, using image processing tools, to automate the
detection of bubble boundaries. Previous work to determine
bubble boundaries was based on visual inspection [6],
calibration wedge [1,7] and minimum entropy [8]. The
criterion for detecting bubble boundaries, in this study, is
based on visual inspection. Many different algorithms were
Fig. 3. Micro CT picture of a 650 Am polyethylene particle showing the
non-spherical shape and porous nature.
attempted to determine the most accurate and repeatable
method for bubble identification. Pictorially, the imaging
objective is shown in Fig. 4 where the original image with
poor contrast and noise is processed to produce the final
binarized image from which data can be extracted.
Through a series of imaging operations the bubble
boundaries are identified. While a reduction of the random
noise can be done with smoothing and opening and closing
operations [9] to remove constant noise in the system a
background subtraction was required. Noise, perhaps is not
the best description for the constant noise in the system
because it is due to several factors such as object geometry,
X-ray beam geometry and heeling effect.
3.2. Background subtraction
The object geometry can affect the path length of the X-
ray. Since the object is circular, the beams traveling through
the edge of the column will go through less material than the
centre. As a result, the beam will attenuate less according to
the Beer–Lambert relationship and there will be a gradient
from dark in the middle to bright on the side.
I ¼ I0exp � lqlð Þ ð1Þ
A conical X-ray beam increases this gradient. The last
effect is called the heeling effect [10]. Due to the X-ray
anode angle, there is a brighter spot on the right hand side of
the image, which is present in our system. As are result of
these phenomena, a correction must be made to ensure
uniform intensity across the image. To accomplish this, a
background subtraction was performed.
While background subtraction initially seems simple,
there are some complications. The first obvious complica-
tion is creating a background image. The background image
in this work was constructed by averaging a series of images
from the run. The theory is that, if the bubbles are randomly
occurring, the brightness created by them in each frame
should average out when a sequence of frames is averaged.
Previous attempts were made to use images from a non-
Fig. 4. The image on the left is the original cropped image obtained from
the fluoroscopy unit and the image on the right is the final image with the
bubbles identified after image processing.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 23
fluidized column. However, this was not useful for the
algorithm used in this study to determine the bubbles. The
algorithm used in this study revolves around subtracting the
background from an image and being left with only bright
regions, or bubbles. Since a packed bed has a higher bulk
density than a fluidized bed, because of decreased voidage,
the images are darker. Since a dark image has lower pixel
intensity values, upon subtraction additional pixels remain.
Therefore further modifications are necessary. Operating the
bed at minimum fluidization and to obtain a background
image resulted in similar problems. The background image
was an average of 200 frames chosen mid way through the
experimental run. This minimized any potential intensity
variations caused by the X-ray source or image intensifier.
3.3. Segmentation
Typically, when images are segmented, to separate an
object from the background, a global threshold is used. This
means that the user specifies a greyscale value so that every
value either above or below (or sometimes in between two
values) is the region of interest and is therefore separated
from the image. The reason that global thresholding does
not work in this study is due to the wake of the bubbles.
Since a wake is present on all bubbles, care has to taken
when distinguishing between small bubble and the wake
around large bubbles. It should be noted that the wake of the
small bubbles is proportionally less than large bubbles [11].
If global thresholding is used, it tends to either over-predict
Fig. 5. The steps for determining the bubble b
the size of the large bubbles and identify the smaller
bubbles, or under-predict the size of large bubbles and not
identify small bubbles. Since the amount a beam is
attenuated is due in part to the amount of material along
the length of the X-ray beam, it can shown that the amount
of mass along a beam that passes through the wake of a
large bubble can be greater than or equal to the amount of
mass along a beam that passes through a small bubble.
Local thresholding uses successive erosion and logic to
determine the number of bubbles in an image. Fig. 5 shows
the steps involved when separating the bubbles.
The image chosen for this example is one that after
examining the images before and after five separate bubbles
were identified. The present algorithm works by taking the
original image and eroding the boundaries until the bubbles
are separated. Stepping through the images in Fig. 5, Fig. 5a
is the original image. This image is binarized and the result
is stored in a separate image buffer, Fig. 5f. The original
image is then eroded by one intensity value as shown in Fig.
5b. This image is binarized and also stored in a separate
buffer as shown in Fig. 5g. This image is then compared to
the previous binarized image to see if any new bubbles have
appeared. If any new bubbles are found, they are stored in
the binarized buffer as only the bubble that is separated is
changed while the other bubbles maintain their maximum
size. This continues for five iterations until the final image
has separated the bubbles as seen in Fig. 5j. Five iterations
were chosen after examining many frames and determining
when bubble separation had occurred and when bbubblesQ
oundaries based on local thresholding.
Fig. 7. Picture showing the difference between the real bubble’s centre of
gravity and the calculated centre of gravity based on the image frame.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3324
were being artificially created by intensity variations in the
real bbubblesQ.The last step in determining bubble boundaries is to
superimpose the final binarized image as onto the back-
ground subtracted image and to threshold the image at
30% of the maximum intensity of each individual bubble.
A sensitivity test was done on this and was found to be the
best value based on visual inspection and matching
correlation.
3.4. Tracking bubbles from frame to frame
The output file from the bubble identification program
is a text file that contains the frame number and individual
bubble properties for all the captured frames. Additionally,
each bubble is given a label, one to the number of bubbles
in the frame. The bubbles are labeled sequentially as they
are met, scanning left to right and top to bottom. Problems
arise if bubbles coalesce, break-up, go out of the frame,
come into the frame, appear or disappear. This is because
these actions change the relative labeling in successive
frames. A simple example shows the difficulties with
tracking bubbles. If there are four bubbles in one frame,
with bubble one and two joining, and three bubbles in the
next frame (Fig. 6), without any logic statements, which
account for joining bubbles, the program (details are
available in Hulme, [12]) would think the bubble labels
match (i.e. 1 with 1 and 2 with 2 and so forth). However,
since two bubbles have joined, bubble 1 in the next frame
is actually two bubbles (1 and 2) in the previous frame. It
can similarly be shown that other actions cause problems
with matching bubbles, which are resolved within the
program so that bubbles are match correctly giving them
the proper velocities.
To match the bubbles, a series of logic is used. More
detail is available in Hulme [12].
There is also one additional problem that arises because
of the size of the images. Since the diameter of the image
Fig. 6. Image showing the problem with matching bubbles when bubbles
join.
frame is only 13.5 cm, various camera heights have to be
used in order to obtain a complete picture of the bubble
properties for the entire height of the reactor. Because of the
image edges (both top and bottom), sometimes only part of
a bubble is visible in the frame. As a result, when the centre
of geometry is calculated, it is, in actual fact, not of the
brealQ bubble but only a portion, as shown in Fig. 7.
Therefore, bubble velocities will be incorrectly calculated.
Outgoing bubbles are accounted for by checking the
frame after the current frame being processed. If there is a
bubble that has a minimum point of 0 in the y direction (top
of the frame) then it is in the process of going out of the
frame. If this is the case, then the properties of the bubble on
the current frame are passed to the next frame. While the
area of the bubble remains the same, the centre of gravity is
now shifted by the amount the bottom of the bubble moves.
This is continues until the bubble completely disappears
from the frame.
Bubbles entering the frame also require modification.
When the data is read into the program, seven frames ahead
of the one being processed are also read. If a bubble has
bappearedQ in the frame from the bottom, then a bubble that
matches the space it occupies is found in subsequent frames.
This is done by looking at the following frame and
determining if the current partial bubble is within the
bounds of a bubble in the next frame. This logic continues
until a bubble is found in a subsequent frame that is
completely within the image frame.
4. Experimental results
To validate the software developed for this work, data
sets were created to simulate bubble movement. Bubbles
were drawn on graph paper, representing the resolution of
the image. The bubbles were then bmovedQ to simulate the
actions of real bubbles in the experiments. Once it was
determined that the program could pick up these actions,
the program was tested on a trial set of data using a
narrow particle size distribution of glass beads (150 to 250
Am). Several parameters had to be adjusted through trial
and error in order for the program to track the bubbles
Fig. 8. Graph showing the average bubble diameters as a function of bed height for glass beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 25
correctly and predict values similar to those found by other
researchers.
4.1. Glass beads
The average bubble diameter is comparable with a
correlation by Werther et al. [13] and the correlation by
Mori and Wen [14]. However, it appears to taper off at
higher bed heights, as seen in Fig. 8, which could be due to
the scale of the reactor.
The average axial velocity, as seen in Fig. 9, also follows
trends found in literature.
Fig. 9. Figure showing the average axial velocity as a function of bed height fo
correlation by Kunii and Levenspiel [15] based on data by Werther et al. [13] using
single bubble velocity proposed by Davidson and Harrison [2].
While the absolute value for the bubble velocity is
different in Fig. 9, the shape of the curve is comparable to
the correlation of Kunii and Levenspiel [15] using the single
bubble velocity of Wallis (1969), cited in Kunii and
Levenspiel [15] which accounts for wall effects.
4.2. Polyethylene using a settled bed of 0.4 m
The system of interest for our research is a polyethylene
air system. To compare with literature a settled particle bed
height of 0.4 m was used. The raw data will be presented
using a superficial gas velocity of three times minimum
r glass beads. The difference in the correlations is that KL Wallis uses a
the velocity of a single bubble by Wallis (1969) [18]. KL Davidson uses the
Fig. 10. The frequency distribution of bubbles (for polyethylene beads) throughout the column, showing that the number of bubbles become less farther away
from the distributor.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3326
velocity to demonstrate the distribution of the data and then
averaged to determine trends.
Before analyzing the diameters and velocities of the
bubbles, the gross distribution of bubbles in the bed was
examined. According to literature, the number of bubbles
should decrease with increased height above the distributor,
as shown in Fig. 10.
Fig. 10 represents the total number of bubbles at each
respective binned height divided by the total number of
bubble throughout the column for the duration of the
experiment. Fig. 10 agrees with what is observed visually,
Fig. 11. The raw data of the effective bubble diameters showin
in that there is a larger number of bubbles near the
distributor with fewer at higher bed heights, due to bubbles
coalescing.
The first property examined was the bubble diameter,
which can be difficult to define due to the irregular shape
of the bubbles in a bubbling bed. While the program
created can determine the feret diameter of the bubbles,
authors in literature use an effective bubble diameter. The
effective diameter is calculated assuming that the bubbles
are spheres. As such, since the present images are two-
dimensional, the diameter of the bubbles is calculated
g the wide distribution of sizes for polyethylene beads.
Fig. 12. The frequency of bubbles at certain diameters for polyethylene beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 27
assuming the area of the bubble, calculated by the
program, as that of a circle. Therefore the diameter of
the bubble can be calculates as:
Db ¼�4� Total area of bubble
p
�0:5
ð2Þ
The feret diameter, calculated in the program, is a more
accurate measurement of the bubble diameter since it does
not assume the bubble is a circle. Instead, it calculates the
Fig. 13. 3D graph showing the frequency distribution of the bubbles diameters (fo
heights small bubbles are present.
length of a chord that joins two points on the surface of the
bubble. Determining this length at eight different angles, the
program then outputs the minimum, maximum, and mean
feret diameter. These values give a better idea of the shape
and size of the bubbles. However, a considerable portion of
the work is a comparison with literature and, as a result, the
effective diameter is used as a basis for comparison.
The effective bubble diameters can be plotted and
compared to the bed height as seen in Fig. 11, to get a
feel for the size and distribution of the bubbles in the reactor.
r polyethylene beads) at different binned heights. This shows that at all bed
Fig. 14. Average bubble diameter as a function of bed height for polyethylene beads. This graph also shows the difference between the average ferret diameter
and the effective diameter.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3328
Fig. 11 demonstrates that, at every bed height, there are
small bubbles, which is confirmed by visual observation.
The scatter in the data also agrees with observations by
other researchers [16–19].
The bubble diameter distribution was also examined,
seen in Fig. 12.
As expected, Fig. 12 shows what is visually observed,
that there are a large number of small bubbles and fewer
large bubbles.
The distribution of bubble diameters at binned bed
heights further elucidates the distribution of bubble diam-
eters. The best way to visualize the distribution is on a three-
dimensional graph seen in Fig. 13, where the z-axis is the
Fig. 15. Ratio of the maximum feret diameter to the minimum feret diameter (for p
for a circular bubble would be 1.0.
binned bed heights. The bubble properties are binned in 4
cm intervals to produce the graph.
It is interesting to note that, at higher bed heights, the
fraction of small bubbles present is significant. This is
confirmed by examining the images. In many instances,
small bubbles appear and disappear brandomlyQ at all bedheights. Sometimes these small bubbles do not coalesce but
travel axially without interaction. It should also be noted
that the distribution of bubble diameters is not uni-modal.
The next step was to average the bubble diameters over
the binned bed heights. While statistically taking the mean
of a sample that is not uni-modal (see Fig. 13) is incorrect, it
does provide a means of comparison with other correlations
olyethylene beads) showing that the bubbles are nor circular, since the ratio
Fig. 16. Axial bubble velocity as a function of bed height for polyethylene beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 29
and gives a feel for the predominant bubble size. The
averaged results can be seen in Fig. 14.
Fig. 14 shows a comparison of the effective and feret
diameter with correlations by Werther et al. [13] and Mori
and Wen [14]. It should be noted that there is a difference
between the feret diameter and the effective diameter.
Clearly the shape of the bubble are not circular in these
images and examining the ratio of the maximum to
minimum feret diameter, Fig. 15, the ratio deviates from
one significantly.
The bubble axial velocities were calculated based on the
movement, from frame to frame, of the geometric centre of
Fig. 17. The frequency distribution of bubb
the bubbles. Fig. 16 shows the bubble axial velocities
plotted as a function of the bed height.
Similar to the plot for bubble diameters, seen in Fig. 11,
there is considerable scatter in the velocities, which is
consistent with literature.
As with the bubble diameters, the distribution of bubble
velocities were examined to ensure a proper distribution and
to ensure there was not a large number of negative bubble
velocities, since bubbles were not observed traveling down
the column (although it has been reported by Clift and Grace
[20] that such velocities may well occur). Negative bubble
velocities were attributed to bubbles being matched incor-
le velocities for polyethylene beads.
Fig. 18. 3D graph showing the distribution of the bubbles velocities at different binned heights for polyethylene beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3330
rectly and discarded in the averaging process. The frequency
distribution, seen in Fig. 17, shows that the distribution is,
as expected, Gaussian, with a few bubbles traveling quickly
(as they coalesce).
The distribution of the bubble velocities follows an
almost uni-modal distribution, as shown in Fig. 18, a three-
dimensional graph. This is similar to the graph presented for
Fig. 19. Average axial velocities obtained experimentally with correlation by K
experiment for polyethylene beads.
the bubble diameters (where the bubble properties are
binned in 4 cm intervals to produce the graph).
Due to the near Gaussian distribution of the axial bubble
velocities, averaging the bubble velocities is statistically
valid (as opposed to averaging the bubble diameters). By
averaging the bubble velocities in the same manner as the
bubble diameters, Fig. 19 is obtained.
unii and Levenspiel [15] using diameters from Mori and Wen [14] and
Fig. 20. The average axial bubble velocity for binned diameters compared with a correlation by Kunii and Levenspiel [15] for polyethylene beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 31
Fig. 19 shows a strong agreement between the correlation
of Kunii and Levenspiel [15] while taking into account wall
effects (Wallis cited in [15]), and the experimental results.
Binning the experimental bubble diameters, in 1 cm
increments, the axial velocities of the bubbles can be
calculated and compared with the correlation by Kunii and
Levenspiel [15] as shown in Fig. 20.
Once again the axial velocities for the respective bubble
diameters adequately match the correlation by Kunii and
Levenspiel [15], with the maximum relative difference
being 10%.
Fig. 21. The effect on bubble diameter at different superficial gas velocities for po
[13].
4.3. The effect of superficial gas velocity on bubble
properties
The last result studied was the effect of superficial gas
velocity on the properties of the bubbles. The average
bubble diameters, at 4 cm binned bubble heights, for the
different gas velocities were calculated and plotted to
determine trends shown in Fig. 21.
Fig. 21 shows that, as the superficial gas velocity
increases, the average bubble diameter increases. This
agrees with two-phase theory, which states that any air in
lyethylene beads. The bubble diameter correleation is that of Werther et al.
Fig. 22. Bubble velocities at different superficial gas velocities for polyethylene beads.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–3332
excess of that required for minimum fluidization will form
bubbles. This also agrees with the correlations in literature
[11,13,14,21–23], where every correlation predicts there
will be larger bubbles at higher superficial gas velocities.
This being said, the correlation of Werther et al. [13] does
not predict as great an increase as the experimental values.
The next property examined was the effect of superficial
gas velocity on the axial velocity of the bubbles shown in
Fig. 22.
Fig. 22 does not show a definite trend, although there
does appear to be a slight increase in the bubble velocity as
the superficial gas velocity is increased. By binning the
bubble diameters, in 1 cm increments, and averaging the
velocities, the effect of different process conditions on
similar size bubbles can be determined as shown in Fig. 23.
Fig. 23. Graph showing that the fluidization velocity, for polyethy
Fig. 23 shows that bubble velocities do not appear to
significantly change with a change in superficial gas
velocity. The velocity correlation by Kunii and Levenspiel
[15] suggests an effect of the gas velocity on the bubble
velocities. However, for the present system, the correlation
by Kunii and Levenspiel [15] does not predict a noticeable
increase in the bubble velocities.
5. Conclusion and recommendations
X-ray fluoroscopy proved to be a powerful tool at
visualizing the movement and determining the properties of
bubbles in a laboratory scale gas–solid fluidized bed reactor.
Using this technique and software developed in house for
lene beads, does not significantly affect the bubble velocity.
I. Hulme, A. Kantzas / Powder Technology 147 (2004) 20–33 33
analyzing the data, the properties of bubbles using glass
beads and a LLDPE for the solid phase, match the bubble
diameter and axial velocity correlations by Werther et al.
[13] and Kunii and Levenspiel [15], respectively. Experi-
ments were also conducted to determine the effect of
superficial gas velocity on the properties of the bubbles. As
expected, an increase in the superficial gas velocity caused
an increase in the bubble sizes.
While the experimental technique does provide good
resolution (0.1�0.1 mm), the contrast of the images needs
to be improved. In addition to further validation of the
experimental technique and software, the next focus will be
on calibrating the X-ray unit. Specifically, using the
intensity values to determine the mass or number of particles
along the length of the X-ray beam. Using this information,
the three-dimensional structure of the bubbles might be
obtainable. Many more experiments are to be conducted to
determine if the behaviour of polyethylene fluidized bed
behaves much differently than accounted for by correlations
presently in literature.
Nomenclature
Db effective diameter of the bubble
Acknowledgements
The authors wish to acknowledge financial assistance
from NOVA Chemicals and the Natural Sciences and
Engineering Research Council of Canada. The contribution
of Sandra Heiber in acquiring the micro CT image of the
polyethylene pellet is appreciated.
References
[1] J.G. Yates, D.J. Cheesman, Voidage variations in the regions
surrounding a rising bubble in a fluidized bed, AIChE Symp. Ser.
88 (1992) 34–39.
[2] J.F. Davidson, D. Harrison, Fluidized Particles, Cambridge University
Press, New York, 1963.
[3] H. Hatzantonis, H. Yiannoulakis, A. Yiagopoulos, C. Kiparissides,
Recent developments in modeling gas-phase catalyzed olefin poly-
merization fluidized-bed reactors: the effect of bubble size variation
on the reactor’s performance, Chem. Eng. Sci. 55 (2000) 3237–3259.
[4] R.D. Toomey, H.F. Johnstone, Gaseous fluidization of solid particles,
Chem. Eng. Prog. 48 (1952) 220–236.
[5] D. Geldart, Types of gas fluidization, Powder Technol. 7 (1973)
285–292.
[6] D.L. Pyle, D. Harrison, The rising velocity of bubbles in two
dimensional fluidized beds, Chem. Eng. Sci. 22 (1967) 531–535.
[7] J.B. Romero, D.W. Smith, Flash X-ray analysis of fluidized beds,
AIChE J. 11 (1965) 595–600.
[8] R.F. Mudde, H.B.M. Schulte, H.E.A. van den Akker, Analysis of a
bubbling 2-D gas-fluidzed bed using image processing, Powder
Technol. 81 (1994) 149–159.
[9] J.C. Russ, The Image Processing Handbook, 3rd edition, CRC Press,
USA, 1999.
[10] T.S. Curry, J.E. Dowdey, R.C. Murray Jr., Christensen’s Physics of
Diagnostic Radiology, Williams & Wilkins, Baltimore, 1990.
[11] P.N. Rowe, C. Yocono, The bubbling behaviors of fine powders when
fluidized, Chem. Eng. Sci. 31 (1976) 1179–1192.
[12] I. Hulme, Verification of the Hydrodynamics of a polyethylene
Fluidized Bed Reactor using CFD and Imaging Experiments, MSc.
Thesis, University of Calgary, 2003.
[13] J. Werther, J.F. Davidson, D.L. Keairns, Influence of the distributor
design on bubble characteristics in large diameter gas fluidized beds,
Fluidization, Cambridge University Press, London, 1978, pp. 7–12.
[14] S. Mori, C.Y. Wen, Estimation of bubble diameter in gaseous fluidized
beds, AIChE J. 21 (1975) 109–115.
[15] D. Kunii, O. Levenspiel, Fluidization Engineering, 2nd edition,
Butterworth-Heinemann, Boston, 1991.
[16] K. Goddard, J.F. Richardson, Bubble velocities and bed expansion in
freely bubbling fluidised beds, Chem. Eng. Sci. 24 (1969) 663–670.
[17] Miwa, K.S. Mori, T. Kato, I. Muchi, Behavior of bubbles in a gaseous
fludized bed, Int. Chem. Eng. 12 (1) (1972) 187–194.
[18] J. Werther, O. Molerus, The local structure of gas fluidized beds: 1. A
statistically based measuring system, Int. J. Multiph. Flow 1 (1973a)
103–122.
[19] J. Werther, O. Molerus, The local structure of gas fluidized beds: 2.
The spatial distribution of bubbles, Int. J. Multiph. Flow 1 (1973b)
123–138.
[20] R. Clift, J.R. Grace, Coalescence of bubbles in fluidized beds, AIChE
Symp. Ser. 67 (23) (1971) 23–33.
[21] R.C. Darton, R.D. La Naueza, J.F. Davidson, D. Harrison, Bubble
growth due to coalescence in fluidized beds, Trans. Inst. Chem. Eng.
55 (1977) 274–280.
[22] M. Horio, A. Nonaka, A generalized bubble diameter correlation for
gas–solid fluidized beds, AIChE J. 33 (1987) 1865–1872.
[23] J.H. Choi, J.E. Son, S.D. Kim, Generalized model for bubble size and
frequency in gas fluidized beds, Ind. Eng. Chem. Res. 37 (1998)
2559–2564.