Chemical Engineering Science 54 (1999) 2113}2122

Dynamics of #uidized-bed reactors. Development and application of

a new multi-"ber optical probe
Davoud Tayebi *, Hallvard F. Svendsen , Arne Grislinga s
, Thor Mejdell,
Kjetil Johannessen
Department of Chemical Engineering, Norwegian University of Science and Technology, NTNU, N-7034 Trondheim, Norway

STATOIL Research Centre, Trondheim, Norway
SINTEF Applied Chemistry, Trondheim, Norway
SINTEF Electronics & Cybernetics, Trondheim, Norway

Abstract
There seems to be no method available to measure all parameters of interests in multiphase #ow systems with high solids
concentration and with reasonable accuracy. Parameters include, local particle velocity vectors, solids volume fractions and bubble
rise velocity, size, frequency and volume fraction. In this study a novel method based on a "ber optical technique and tracer particles
has been developed for simultaneous measurements of the mentioned local #ow properties in highly concentrated multiphase #ow
systems such as gas}solid #uidized bed reactors. A particle present in the measuring volume in front of the probe is marked with
a #uorescent dye. A light source illuminates the particles and the detecting "bres receive re#ected light from uncoated particles and
#uorescent light from the tracer particle. Using optical "lters, the #uorescent light can be distinguished and together with a small
fraction of background light from uncoated particles can be used for the determination of local #ow properties. The method gives the
possibility for simultaneous measurement of the local movement of a single tracer particle, local bubble properties and the local solids
volume fractions in di!erent positions in the bed.  1999 Elsevier Science Ltd. All rights reserved.
Keywords: Particle velocity; Bubble properties; Solids volume fraction; Tracer particles; Fiber optics; Fluidized beds

1. Introduction
Fluidized bed- and circulating #uidized-bed reactors
are used in many areas in the process industry, especially
in the oil and petrochemical industries. There are still
many uncertainties in the design, operation and scale up
of such reactors, particularly when the movement and
circulation of solids is concerned. In these systems the
local values of solids volume fraction may range from
0 to approximately 60 vol.%. For better understanding
of the #ow properties, behaviour and for fundamental
modelling of these multiphase systems, it is of prime
importance to know the details of the #ow. The local
velocities of both phases and the local distribution of

*Corresponding author. Fax: 47 22 06 7350; e-mail: davoudt@chem.

sintef.no.

solids volume fraction are particularly important. Also

any phenomenological model developed in this area require empirical data. The reliability, accuracy and quality
of these models therefore depend on the accuracy of the
employed measurement techniques.
Pressure measurements have long been used in #uidization studies. The time-averaged values of the pressure
measurements at di!erent locations in the bed give qualitative information of the bed behaviour (Glicksmann
et al., 1993; Marzocchella et al., 1997).
For suspensions with a solid volume fraction greater
than a few percent the optical depth is of the order a few
particle diameters. As a result, techniques such as visualization or Laser}Doppler velocimetry are impractical
under such conditions (Arastoopour and Yang, 1992;
Horio and Kuroki, 1994).
Electroresistivity techniques are sensitive to the changes in the e!ective dielectric properties. The e!ective
properties of arbitrary powders which are a!ected by

0009-2509/99/$ } see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 8 ) 0 0 3 1 9 - 4

2114

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

humidity, temperature, and particle shapes may be di$cult to establish (Louge and Opie, 1990).
The main limitations of capacitance tomography techniques are the poor resolution and the blurring of voidage boundaries. In addition comes the uncertainties
prevalent in all tomographic techniques, relating to the
deconvolution of the current measurements. False images may also occur, especially when there are more than
two objects within the same imaging "eld (Halow and
Nicoletti, 1992; Williams and Beck, 1995). The major
disadvantage of X-ray, c-ray and C scanners is the
time-averaged nature of the measurements, which limits
the usefulness of these techniques in process diagnostic
applications (Azzi et al., 1991, Yates and Cheesman, 1992;
Kumar et al., 1997).
The overall #ux of solids circulation in high-velocity
systems may be determined by collecting the total #ow
over a period of time and measuring each phase separately by means of sampling probes. In #uidized beds,
particle motion appear in di!erent directions. It is necessary to collect solids from various positions with the
probe opening in di!erent directions at separate time
periods. Time averaging is therefore necessary (Rhodes
and Laussmann, 1992; Werther et al., 1993).
The particle-tracking techniques such as CARPT
(Computerized Automated Particle Tracking) or PEPT
(Positron Emission Particle Tracking) are limited to
the use of a de"ned particle. In a #uidized bed, particles are distributed over a certain range of sizes, and
generally may have non-spherical form. Therefore, the
tracer particle will at best represent only a narrow
fraction of the bed particles. The implementation of
the above techniques require many detectors and much
calibration works. The experiments may also be of
long duration. (Lin et al., 1985; Sannvs, 1997; Simons
et al., 1993).
Many investigators have attempted to measure local
solids motion and volume fractions in dense suspensions
using optical "bre sensors. Optical "bre probes are
simple, yield high signal-transmission-to-noise ratios,
and create a minimum disturbance to the #ow. The
sensors used are either based on forward light scattering
between emission and detection "bres separated by
a short distance, or on backscattering onto an optical
"bre system. Forward scattering is limited to relatively
low solids volume fractions and also is considered to
disturb the #ow structure more than the backscattering
method. It is therefore impractical for dense systems
(Okhi and Shirai, 1976; Hartge et al., 1989). Because
backscattering sensors are simpler and less intrusive, they
have received wider attention, see Oki et al. (1975)
Lischer and Louge (1992) and Cocco et al. (1995).
In the measurement of particle velocities with "bre
optical probes, almost all detection systems apply two
parallel sensors placed at a certain distance. This design
can determine the magnitude of the velocity vector in one

dimension. As shown in this paper, the values obtained

by such a method can be misleading.
As discussed brie#y above, and according to the detailed review of the measurement techniques by Tayebi
(1998), to the knowledge of the authors there is presently
no method available to determine all the parameters of
interests in multiphase systems with high solids concentration with a reasonable accuracy.
Any successful technique for measuring the abovementioned parameters, particularly solids movement
has to combine the ability to work over a wide range
of solids concentrations and particle sizes. It requires
also minimum disturbance of the #ow structure and
robustness under harsh conditions. A successful design of
optical probes can ful"l the necessary conditions, and
this work is therefore based on the application of this
technique.

2. Measuring principle
The local motion of a single particle in gas}solid #ow
can be determined using two parallel optical "bres separated by a known distance as light detectors. For a su$ciently high light sampling frequency the movement of
a self-lighting particle in front of the probe may be
detected as two bell-shaped curves separated in time.
One component of the particle velocity vector, u can be
NV
obtained by dividing the distance between the sensors,
Dx, by the measured time delay, Dt.
Dx
; " .
NV Dt

(1)

This measuring principle is directly applicable only in

very dilute gas}solid #ows. In #uidized beds, especially in
the dense region, the above technique can only be applied
by using tracer particles. These tracer particles should
have exactly the same physical properties as the other
particles in the bed apart from the radiation properties.
This can be achieved by a #uorescent dye impregnation
of the tracer particles to be. Only one tracer particle at
a time should be present in the measuring volume. It is
then possible to distinguish between the radiation from
the main particles and that from the tracer particle. The
re#ected light from uncoated particles in the bed and the
#uorescent light from the tracer particle may be separated using a proper optical "lter combination. Fig. 1
illustrates this method whereby the obtained signals from
the passing tracer particle are similar to those from
a system with only one single particle.
A two-dimensional picture of the tracer particle movement in front of the probe can be obtained by arranging
a number of detectors in a certain geometrical con"guration. Knowing the relative positions of the detectors,
the local particle velocity vector in two dimensions is

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

obtained by a simultaneous treatment of the time-series

signals from all detectors.

2115

The probe is made as small as possible with a sharp

edge and smooth face to reduce the distortion of the #ow.
In #uidized beds where the particle movement may occur
in all directions, a circular design of the probe tip is an
optimal solution. For velocity measurements usually two
or more detectors/probes are required. Using separate
probes will disturb the #ow and create probe interference. This can be avoided by embedding all detectors in
the same probe.
For particle velocity measurements an important factor is determination of the e+ective distance between two
sensors. Consider a particle passing in front of two parallel detecting "bres separated by a known distance. If the
moving path of the particle is parallel to the end face of
the "bres, the in-line component of its velocity can be
determined from Eq. (1). Maximum light intensity is

obtained when the particle has the shortest distance to

the "ber end centre. The distance between two light
maxima is the e+ective distance, l which in some cases
C
can be di!erent from the geometric distance, l , (Patrose
E
and Caram, 1982). The e!ective distance, l , is seen to be
C
a function of the viewing angle of the receiving "bres and
the nature of illumination at the measuring site, see
Fig. 2.
To overcome the above problems, uniform illumination should be applied with both central and surrounding
light emitting "bres as shown in Figs. 2D and 3. A thin
glass window in front of the probe covers the blind spots
and together with the right "ber con"guration ensures
optimum illumination/detection uniformity.
Fig. 3 shows the "ber arrangement and details of the
probe tip. Thirty seven equal diameter plastic "bres are
"xed by epoxy glue in a specially made steel pipe with
a hexagonal shaped inner wall. Seven of the "bres are
selected as detectors while the other 30 are used as
emitters. The geometrical con"guration of "bers is such
that six of the detecting "bres are positioned at the
corners of an inner hexagon and one in the centre of the

Fig. 1. Schematic illustration for velocity measurement of a passing

single #uorescent dye impregnated tracer particle in the presence of
other uncoated particles using two parallel optical "bres as detectors.
The re#ected light from uncoated particles is blocked and only the
#uorescent light from the tracer particle is recorded.

Fig. 3. Optical "bres con"guration and details of the probe tip design.
The external steel cover keeps the sapphire window "xed in front of the
probe.

3. Probe design

Fig. 2. Schematic illustration of the in#uence of illumination and "bre viewing angle on the e!ective distance, l , relative to the geometric distance l .


(A) the end surface plane is not perpendicular to the "bre axis (B) The illumination light placed between receiving "bres (C) external illumination; (D)
uniform illumination. The distribution of illuminating light, I is schematicaly illustrated in the right-hand side of each "gure.


2116

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

probe. Each detector is surrounded by six emitters ensuring symmetric and &&uniform'' light illumination of the
measuring site. The end face of the "bres are polished by
a multi procedure operation using ultra "ne sand papers.
The tip of the probe is smoothed and protected by a thin
sapphire window with a thickness of 0.5 mm and diameter of about 2.5 mm. The distances and angles between
all detecting "bres were measured by microscopy and
image processing.

4. Tracer particles
When tracer particles are used to represent the #ow of
particles, their physical properties must be close to the
other particles in the bed. As tracer particles, one may use
a size fraction, or only one size of particles with the same
physical properties as the bed particles. Movement of
particles with other physical properties than the particles
in the bed can be measured using a small fraction of these
particles as tracer.
When tracer particles are made by #uorescent dye
impregnation as in this work, the weight of the dye is
negligible. The physical properties of the tracer particles
are then almost unchanged. In a #uidized bed a tracer
particle is always surrounded by and collides with other
particles. The #uorescent dye should not be lost or transferred to other uncoated neighbouring particles, but "xed
to the tracer particles both physically and chemically.
In this study "ve batches of spent #uidized cracking
catalyst (FCC) particles with the size distribution, shape
and density as the bed particles were coated with di!erent #uorescent dyes. Two batches were impregnated by
Molecular Probes Inc., USA and three batches by SINTEF Applied Chemistry, Norway. The #uorescence e$ciency and light spectra of all "ve tracer groups were
analysed by exposing them to the same laser source with
constant output power and wavelength (488}514 nm).

The analysis were carried out using a light spectrometer

(MCS UV-NIR 190}1015 nm, ZEISS), Fig. 4A. The spectrum of the chosen dye is shown in Fig. 4B. The dyeimpregnated tracer particles can easily be deactivated by
exposure to low concentrated ozone.

5. Particle velocity vector

When a group of particles passes the probe, a tracer
particle in the measuring volume will be detected by up
to seven detectors. The number of detecting sensors depends on the local particle concentration in front of the
probe and the tracer particle path. The local particle
velocity vector in two dimensions is then obtained by
simultaneous data treatment of the time series signals
from all detectors.
Fig. 5 shows a small section of a time series measurement conducted in a 7 cm ID bed of SiC particles. In this
example the sampling frequency was set to 20 kHz per
detector. As shown in this "gure, the signals from the
passing tracer particles and the background signals from
uncoated particles are clearly distinguishable. A threshold can be set as indicated by the dashed line.
The velocity vector determination for a passing tracer
particle is based on the maximum points of the signals
from the individual detectors. The projected distances
between the detecting "bres for each individual case are
calculated based on the measured maximum intensities,
the order of detection, and the geometrical positions of
the sensors in the probe.
Fig. 6 illustrates a particle passing two sensors with
a "xed distance, X. In both cases (A) and (B), the magnitude of the particle velocity vector, u and its angle of
N
direction, b are determined by
Xcos b Dx
u"
" .
N
Dt
Dt

(2)

Fig. 4. (A) Schematic diagram of the light spectral assembly. (B) The light spectra from spent FCC particles coated with #uorescent dye and used as
tracer particles in this study, no optical "lter present.

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

2117

Fig. 5. A section of a time-series measurement conducted in a laboratory #uidized-bed reactor of uncoated SiC particles and #uorescent dye
impregnated spent FCC tracer particles. The measurements are carried out at the centre of a 7 cm ID bed and axial height of h"160 mm above the
gas distributor with a super"cial gas velocity of 0.065 m/s.

Fig. 6. Schematic illustration of the determination of the particle velocity vector based on the e!ective distance, *x: (A) the particle passes in
between two sensors, (B) the particle passes the detecting "bres without
passing in between them.

Here, Dx is the projected distance between the sensors

and Dt is the measured time delay between detected
signals. Eq. (2) has two unknown variables and one
additional equation is required for calculation of u and
N
b. Therefore, successful signals from at least three detectors are needed for determination of the velocity vector. It should be emphasized that the velocity vector
determined in this way represents an averaged velocity
across the probe head. Since the probe diameter is
2.5 mm, this is considered to be a local velocity and
determined as a straight vector.
Fig. 7 shows an example of the signals received from
a tracer particle passing the probe. In this case the particle has an average speed of 0.41 m/s and a moving
direction angle of about 3503 relative to the main direction of the probe. It has been observed "rst by detector
number 5, with a maximum signal intensity at point (a),
then detector number 6, with a maximum signal intensity

Fig. 7. Signals from various detectors during the passage of a tracer

particle in front of the probe The particle has a straight moving path,
u "0.41 m/s and b"3503.


at point (b). Detector number 1 has also seen the tracer

particle, but since it is far from it, the signal level is much
lower than the others. Finally, the tracer particle has
been observed by detector number 7, with a maximum
signal intensity at point (d). The maximum intensities
may indicate in-plane distances to the detectors, but will
be strongly in#uenced by the distance from the probe
surface. Therefore, only the time sequences of maxima,
denoted by a}d in Fig. 7 are used for velocity determination. These correspond to the shortest distances to the
detectors in the probe tip plane as depicted in Fig. 6.

6. Local particle volume fraction

The probe was calibrated experimentally in two extreme cases, "rst in a dark and particle-free space and

2118

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

Fig. 8. The probe calibration curves viewing a single #uorescent tracer

particle: (A) In a dark and particle-free space; (B) In an expanded bed of
SiC particles.

then in a &&uniformly'' expanded #uidized bed of uncoated

SiC particles using a speci"c tracer particle. A single
average-size tracer particle was "xed at the end of an
adjustable steel wire placed in a dark and particle-free
space. The optical probe was "xed on the opposite side of
the wire in the same horizontal and vertical plane. The
particle was then moved away gradually and the received
signal intensity was recorded, Fig. 8A.
The probe was also calibrated in a &&uniformly'' expanded bed of uncoated SiC particles by the same procedure as described above, see Fig. 8B.
Based on the above calibration results it is possible to
de"ne the limits of the measuring volume. As shown in
Fig. 8A, at a distance of 1.5 mm the signal intensity
reaches 0.1 V, corresponding to the background signal
level, see Fig. 5. This may be interpreted as the outer limit
of the measuring volume. Any tracer particle further
away will be practically undetectable regardless of local
particle concentration. The maximum depth of the
measuring volume at minimum #uidization velocity is
about 0.5 mm, as shown in Fig. 8B. Since the #uorescent
tracer particles applied in this study have a wide size
distribution, a lower signal intensity is obtained from
particles smaller than the average-size tracer particle
used in the calibrations. The measuring depth will therefore, be limited to between about 0.3 and 1.5 mm, depending on the local particle concentration and the size
of the passing tracer particle.
As shown in Fig. 5 the re#ected light from the uncoated particles passing through the long-pass "lter gives
a visible background signal. Using a proper calibration
function, this background signal can be used to determine the local particle concentration.
The design of the probe is such that each sensor can be
considered as a separate coaxial sensor giving a maximum uniform illumination intensity, see Fig. 2D and

Fig. 9. Characteristic curve of coaxial sensors when a glass window is

used to cover the dark region. The re#ected light intensity decreases
from its maximum with increasing distance from the outer surface of the
glass window.

Reh and Li (1991). One of the advantages of this design is

that the characteristic curve of the sensor will be similar
to that of a single emitting/detecting "bre as shown in
Fig. 9. As shown in this "gure, the signal contribution
from the blind spot region of a coaxial "bre arrangement
is blocked by a glass window. This design can be used for
local solids volume fraction measurements with the advantages of having a more uniformly distributed emitting
light over the entire measuring cross-section and having
several detectors in the probe.
For the measurement of local solids volume fractions,
one of the major di$culties with optical probes in general, and the re#ection type in particular, is that they are
very di$cult to calibrate experimentally, especially in
gaseous suspensions (Mastuno et al., 1988; Tung et al.,
1988; Hartge et al., 1989; Lischer and Louge, 1992; Amos
et al., 1996). Various methods have been used for calibration of single "bre emitting/detecting probes. Despite
all attempts, no exact and accurate method or function is
available for the entire range of concentrations. However,
reasonable results have been reported using the empirical
correlation given by Eq. (3) in gas}solid suspensions
(Hartge et al., 1989; Werther et al., 1993; Rensner and
Werther, 1993; Werther and Hage, 1996).
;!; "aCI .

T

(3)

Here, ; is the measured voltage and ; is the voltage at


C "0. To use this correlation for SiC particles, the
T
coe$cients a and k need to be determined. According to
the theoretical model (Rensner and Werther, 1993), the
exponent k depends mainly on the size distribution of
particles whereas the factor a depends mainly on the
refractive index of the #uid and on the optical properties
of the particles. Thus, in a gas}solid system the numerical
value for k will be the same as in the liquid}solid system.
k"0.57 was chosen based on the calibration data given
by (Werther et al., 1993) and a"0.094 for the gas}solid

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

2119

Fig. 10. Schematic illustration of detected signals by two sensors in the

probe from the passage of a bubble.
Fig. 11. The calculated particle velocity based on u ) r versus the
measured particle velocity by the probe.

system was obtained from Eq. (3) by measurements of the

voltage ; in a "xed bed and at C "0, respectively.

7. Bubbles
As discussed earlier, when the signal level in a measured time series is lower than a certain value, a bubble is
identi"ed. By simultaneous treatment of all time series
from various detectors the bubble rise velocity can be
determined. This is done based on the time delays between detected signals from various sensors during the
passage of a bubble as illustrated in Fig. 10. The rise
velocity at the front and tail may be measured separately
based on the measured time delays D and D .
@D
@R
In a bubbling #uidized bed, bubbles are mainly rising
upward along the bed as vertical axis. Therefore, the time
delay between signals from sensors placed in the same
horizontal level are close to zero. The e!ective distances
between these detectors are also zero and the resulting
velocity will be in"nite. Therefore, only signals from
detectors with di!erent axial positions are compared
with each other assuming the bubble motion to be vertical. Bubble chord length is determined from the average
value of the rising velocities and their average passage
time by
l "uN )  .
(4)
@A
@ @
The size of bubbles are determined from the measured
chord length (Chan et al., 1987)
dN "1.43 u ) .
(5)
@
@ @
The bubble frequency is obtained from the number of
identi"ed bubbles per time unit.

8. Tests and veri5cations

In order to verify the accuracy of the velocity measurements, one average-size tracer particle was "xed on
a rotating disk. The particle was positioned at a certain

radius by means of a microscope. The rotation frequency

of the disk, u, could be adjusted with high accuracy. The
particle velocity was then determined based on its position on the disk and the rotation frequency. The data
presented in Fig. 11 shows a comparison between the
measured particle velocities and the corresponding
values determined by means of ur.
The probe tip was positioned at various distances and
di!erent angles relative to the particle. The velocity determination accuracy was always better than $0.5%.

9. Particle velocity vectors

The probe was applied in a transparent laboratory
#uidized bed reactor with 7 cm ID with air at ambient
temperature as the gas phase, SiC particles as the solid
phase and dye impregnated FCC particles as tracers. An
example of the time-series measurements is presented in
Fig. 5.
Tracer particle velocity vectors in two dimensions were
measured at various radial and axial positions in the bed.
The measurements were carried out at three di!erent
super"cial gas velocities of 0.140, 0.065 and 0.035 m/s
respectively. Fig. 12 shows the variations in particle
velocity vectors in the bed at u "0.065 m/s. All the
QE
measurements were made under the same conditions,
including optical properties, ampli"cation of the signals
and sampling rate of 20 000}25 000 scans/s/detector. The
sampling time was set to 60 s in all cases.
Each "gure shows the measured velocity vectors in two
dimensions at di!erent radial positions but at the same
axial level. The vertical axis of the probe is indicated in
the "gures with an arrow in the x-direction and is parallel
to the column vertical axis. The arrows in the "gures
represent the velocity vectors of each successfully measured individual passing tracer particle. The magnitude
and direction of the resultant velocity vector (u and b)
N
for each time-series measurement is given beside the
corresponding "gures. As shown in the "gures, due to the

2120

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

Fig. 12. Tracer particle velocity vectors at h"160 mm and u "0.065 m/s.
QE

rapid variations in moving direction, the value of the

resultant velocity vector is often smaller than the magnitude of the instantaneous velocity vectors. In some cases
this averaged value is almost zero.
In order to study the repeatability of the results,
a number of measurements were repeated and good
agreement was obtained, $5% in resultant velocity and
$3% in moving directions.
Based on the resultant velocity vectors of tracer particles at various radial positions and axial heights at
constant super"cial gas velocity, the radial particle velocity pro"les were determined. An example of these pro"les is illustrated in Fig. 13. As shown in this "gure the
net particle #ow is upward in the core region and downward close to the wall. The particle motion is also seen to
be non-symmetrical, but downward motion is seen close
to the wall on both sides.
Detailed and valuable information can be obtained by
studying the distribution of the particle velocity vectors.
Fig. 14 shows the detailed study of the particle measurements made in the centre of the bed at axial height
h"160 mm and super"cial gas velocity of 0.065 m/s. In
Fig. 14A the tracer particle vectors in two dimensions are
shown. Fig. 14B shows the particle velocity magnitude
frequency spectrum and Fig. 14C shows the frequency of
particles moving directions. The values represented in the

Fig. 13 Resultant particle velocity vectors measured in a bed of SiC

particles at the axial height of h"160 mm and various radial positions
with u "0.065 m/s.
QE

frequency distributions are normalized with respect to

the values obtained from the total number of measured
tracer particles and shows the percentage of the total
measured tracer particles having a certain speed (the
magnitude of velocity vectors) and moving direction. As
shown in Fig. 14 the passing tracer particles move mainly
upward. Almost all particles have a speed greater than
0.06 m/s. Their moving directions appear within
!90}903 with respect to the vertical axis. The net velocity vector was 0.115 m/s.

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

2121

Fig. 14. Tracer particle velocity vectors measured at h"160 mm and u "0.14 m/s. (A) presentation in polar coordinates illustrating the magnitude
QE
and direction of the velocity vectors; (B) the frequency distribution of the velocity magnitude; (C) the frequency distribution of the velocity vectors
directions. From !90 to #90 denotes upward #ow. Zero is parallel to the main bed axis.

10. Conclusions
A method for simultaneous measurements of local
particle velocity vector, bubble properties and solids volume fraction based on the application of a "bre optical
technique and #uorescent tracer particles has been successfully developed. The optical "bre probe has an outer
diameter of 2.5 mm. Its measuring depth is of order
0.3}1.5 mm depending on the local solids concentration.
The probe was successfully calibrated in two extreme
cases, including, a particle-free dark space and a &&uniform'' expanded bed of SiC particles. The calibration
curves were used to estimate the outer limits of the probe
measuring volume. These data were also used to "t the
calibration function of (Hartge et al., 1989) for determination of local solids volume fractions.
The accuracy of the measured velocity vectors by the
probe were tested by the application of a single tracer
particle "xed at a certain radius on a rotating disk at
known rotation frequency. The results obtained by the
probe were in very good agreement with the calculated
ones, ($0.5%).
The method is independent of the physical properties
of the tracer particles such as size distribution, density,
and shape. The technique is also independent of the local
solids volume fractions in the range of 0}60 vol.%, but is
mainly designed for highly concentrated #ow systems.
A software program for simultaneous treatment of the
signals from all detectors has been developed. The program calculates the tracer particle velocity vectors in two
dimensions based on successfully detected signals from at
least three sensors and the order of detection. It determines the bubble rise velocity based on measured time
delays at the front of the bubble. The bubble sizes are
determined based on the measured duration times and

the rise velocity. Bubble frequencies are determined from

the number of successfully measured bubbles.
The local solids volume fractions were determined
based on the background signals from uncoated particles
in the measuring volume.
The particle moving directions in the bed found to be
rapidly changing and are by no means parallel to the
column vertical axis. Thus, measurements of particle
velocity based on only two parallel detectors separated
with a certain distance as used by many investigators,
may often lead to misleading results.

Notation
a
C
T
d
@
h
k
I

l
@A
l
C
l
E
r
r
"

@
uN
@
u
N
u
N
u
NV
u
QE
;
;

X

coe$cient in the calibration function (Eq. (3))

solids volume fraction, Vol.%
averaged bubble diameter, m
axial height from gas distributor in the bed, m
exponent in the calibration function (Eq. (3))
incident light intensity, mW
bubble chord length, m
e!ective distance between detectors, m
geometric distance between detectors, m
radius from the centre of the rotating disk, m
radial position in the bed, m
averaged bubble passing time, s
averaged bubble rise velocity, m/s
particle velocity, m/s
averaged (resultant) particle velocity, m/s
in-line component of the particle velocity, m/s
super"cial gas velocity, m/s
measured signal intensity, V
measured signal intensity at C "0, V
T
distance between detectors in the probe, m

2122

Dx
Dt
D
@D
D
@R

D. Tayebi et al./Chemical Engineering Science 54 (1999) 2113}2122

projected distance between detectors, m

measured time delay, s
measured time delay at front of the bubble, s
measured time delay at tail of the bubble, s

Greek letters
b
u

direction angle of the velocity vector

rotation frequency, s\

References
Amos, G., Rhodes, M.J., & Benkreira, H. (1996). Calculation of optic
"bre calibration curves for the measurement of solids volume fractions in multiphase #ows. Powder echnol., 88, 107}121.
Arastoopour, H., & Yang, Y. (1992). Experimental studies on dilute gas
and cohesive particle #ow behavior using Laser Doppler Anemometer. In O.E. Potter, & D.J. Nicklin (Eds.), Fluidization vol. <II
(pp. 723}730). New York: Engineering Foundation.
Azzi, M., Turlier, P., Bernard, J.R., & Garnero, L. (1991). Mapping solid
concentration in a circulating #uid bed using gammametry. Powder
echnol., 67, 27}36.
Chan I.E., Sisahtla, C., & Knowlton, T.M. (1987). E!ect of pressure on
bubble parameters in gas-#uidized Beds. Powder echnol., 53,
217}245.
Cocco, R., Cleveland, J., Harner, R., & Chrisman, R. (1995). Simultaneous in-situ determination of particle loadings and velocities in
a gaseous medium. A.I.Ch.E. Symp. Ser., 91 (308), 147}153.
Glicksmann, L.R., Hyre, M., & Woloshun, K. (1993). Simpli"ed scaling
relationships for #uidized beds. Powder echnol., 77, 177}199.
Halow, J.S., & Nicoletti, P. (1992). Observations of #uidized bed coalescence using capacitance imaging. Powder echnol., 69, 255}277.
Hartge, E.-U., Rensner, D., & Werther, J. (1989). Faseroptisches MeBsystem zur Erfassung lokaler Feststo!-Konzentrationen und -Geschwindigkeiten in h+her konzentrierten Gas/Feststo!-Str+mungen.
Chemie-Ingenieur-echnik., 61 (9), 744}745.
Horio, M., & Kuroki, H. (1994). Three-Dimensional #ow visualization
of dilutely dispersed solids in bubbling and circulating #uidized
beds. Chem. Engng Sci., 49 (15), 2413}2421.
Kumar, S.B., Moslemian, D., & Dudukovic, M.P. (1997). Gas-holdup
measurements in bubble columns using Computed Tomography.
A.I.Ch.E. J., 43 (6), 1414}1425.
Lin, S.J., Chen, M.M., & Chao, B.T. (1985). A novel radioactive particle
tracking facility for measurement of solids motion in gas #uidized
beds. A.I.Ch.E. J., 31 (3), 465}472.
Lischer, D.J., & Louge, M.Y. (1992). Optical "bre measurements of
particle concentration in dense suspensions: Calibration and simulation. Appl. opt., 31 (24), 5106}5113.
Louge, M., & Opie, M. (1990). Measurements of the e!ective dielectric
permittivity of suspensions. Powder echnol., 62, 85}94.

Marzocchella, A., Zijerveld, R.C., Schouten, J.C., & van den Bleek,
C.M. (1997) Chaotic behavior of gas}solids #ow in the riser of
a laboratory-scale circulating #uidized bed. A.I.Ch.E. J., 43 (6),
1458}1468.
Mastuno, Y., Yamagughi, H., Oka, T., Kage, H., & Higashitani, K.
(1988). The use of optical "bre probes for the measurement of dilute
particle concentrations: Calibration and application of gas #uidized
bed carryover. Powder echnol., 36, 215}221.
Okhi, K., & Shirai, T. (1976). Particle velocity in #uidized beds. In
Fluidization technology, (Vol. 1, pp. 95}110). D.L. Keairns, New
York: Hemisphere.
Oki, K., Akehata, T., & Shirai, T. (1975). A new method for evaluating
the size of moving particles with a "bre optic probe. Powder echnol., 11, 51}57.
Patrose, B., & Caram, H.S. (1982). Optical "bre probe transit anemometer for particle velocity measurements in #uidized beds. A.I.Ch.E.
J., 28(4), 604}609.
Reh, L., & Li, J. (1991). Measurement of voidage in #uidized beds by
optical probes. In P. Basu, M. Horio, & M. Hasatani (Eds.), Circulating -uidized bed technology III, Oxford: Pergamon Press.
Rensner, D., & Werther, J. (1993). Estimation of e!ective measuring
volume of single-"bre re#ection probes for solid volume concentration measurements. Part. Part. System Charact., 10, 48}55.
Rhodes, M.J., & Laussmann, P. (1992). A simple non-isokinetic sampling probe for dense suspensions. Powder echnol., 70, 141}151.
Sannvs, B.H. (1997). Solids movement and concentration pro,les in
column slurry reactors. Dr. Ing. Ph.D. thesis, Norwegian University
of Science and Technology, Norway.
Simons, S.J.R., Seville, J.P.K., Clift, R., Gilboy, W.B., & HosseiniAshra", M.E. (1993). Application of Gamma-ray to gas #uidised
and spouted beds. In M.S. Beck, E. Campogrande, M. Morris, R.A.
Williams, & R.C. Waterfall (Eds.), omographic techniques for process design and operation (pp. 227}238). UK: Computational Mechanics Publ., Southampton.
Tayebi, D. (1998) Simultaneous measurement of local particle movement,
solids concentration and bubble properties in -uidized bed reactors
using a novel ,bre optical technique. Dr. Ing. (Ph.D.) thesis, Norwegian University of Science & Technology, Norway.
Tung, Y., Li, J., & Kwauk, M. (1988). Radial voidage pro"les in a
fast #uidized bed. In M. Kwauk, & D. Kunii (Eds.), Fluidization
&&88 Science and echnology (pp. 139}145). Beijing China: Science
Press.
Werther, J., & Hage, B. (1996). A "bre-optical sensor for high-temperature application in #uidized bed combustion. In Large J.-F. and
Laguerie, C. (Eds.), Fluidization <III pp. (577}584). Eng. Foundation, New York.
Werther, J., Hartge, E.-U., & Rensner, D. (1993) Measurement techniques for gas}solid #uidized-bed reactors. Int. Chem. Engng 33(1),
18}27.
Williams, R.A., & Beck, M.S. (1995). Process tomography: Principles,
techniques and applications. Oxford: Butterworth-Heinemann.
Yates, J.G., & Cheesman, D.J. (1992). Voidage variations in the regions
surrounding a rising bubble in a #uidized bed. A.I.Ch.E. Symp. Ser.,
88 (289), 34}39.