Está en la página 1de 17

Pergamon Chemical Engineering I

Science, Vol. 52, No. 13. pp. 21 l-2127. 1997


C 1997 Elsevm Science Ltd. All rights reserved
Printed in &eat Britain
PII: SOOO9-2509(97)00038-9 OOO!-2509/97 $17.00 + 0.00

Investigation of the two-phase flow in


trickle-bed reactors using capacitance
tomography
Nicolas Reinecke and Dieter Mewes*
Institut fur Verfahrenstechnik, Universitat Hannover, Callinstralje 36, D-30167 Hannover,
Germany

(Received 19 March 1996; in revised form 29 July 1996; accepted 27 August 1996)

Abstract-Investigations of fluid dynamic properties of two-phase flows in regular and irregu-


lar packings of trickle-bed reactors require a detailed knowledge of the flow phenomena inside
the packing. A visual observation of the flow from the outside of the column does not yield the
desired information, especially for large-scale pilot-plant reactors. Therefore, new measurement
techniques are required to obtain the necessary information for the physical description of the
flow, especially for scale-up calculations. In this paper, the possibilities of tomographic imaging
of the two-phase flow in trickle-bed reactors using capacitance tomography are discussed. The
current state of the art of this new measurement technique is discussed and the details of the
system used by the authors are described. The principle of the measurement technique is
explained and the possibility of obtaining new information pointed out. Preliminary results of
this new imaging technique are presented. 0 1997 Elsevier Science Ltd. All rights reserved

Keywords: Two-phase flow; trickle-bed reactors; tomographic imaging; capacitance tomo-


graphy.

1. INTRODUCTION Simons et al., 1993), y- and X-ray transmission


Tomographic measurement techniques have been and scattering (CAT) (Ikeda et al., 1983; Toye et al.,
applied to medical imaging for over 20 years and they 1996; Kantzas, 1994), nuclear magnetic resonance
are now a standard diagnostic tool for many pur- (NMR, MRI) (Gladden, 1994, 1995; Fordham et al.,
poses. The use of tomographic imaging techniques in 1993), microwave reflection and diffraction (Bolomey
process engineering applications has not been con- et al., 1990), ultrasonic/acoustic techniques (Seiraffi,
sidered as long, but is becoming increasingly popular. 1993), interferometry and holographic tomography
These measurement techniques allow for two- or (Mewes and Fellholter, 1993) and electrical and elec-
three-dimensional imaging of various processes in tromagnetical field interaction (ECT, EIT, EMT)
chemical and mechanical engineering industries. (Klug and Mayinger, 1993; Wang et al., 1993, 1994;
Using several possible physical principles to measure Abdullah et al., 1993; Xie et al., 1992; Isaksen and
the integral value used for the reconstruction, the Nordtvedt, 1992; Boddem et al., 1994; Yu et a/.,1993).
phase-, the temperature- and the concentration-distri- The different measurement principles can be classified
bution, as well as velocity distribuitions can be cal- according to the interaction with the medium to be
culated. In addition, some techniques allow for the imaged and the type of physical property measured.
property sensitive investigation of labeled compo- The interactions range from transmission, diffraction,
nents or phases. The advantage of all these techniques refraction, reflection, scattering and emission, while
is the possible space-imaging without intrusion into the physical property measured can be nucleonic or
the process. photon rays, high- and low-energy electromagnetic
For the tomographic reconstruction, integral mea- fields or waves and physical waves. Depending on the
surements have to be made. These can be obtained in specific technique used, different advantages and dis-
numerous ways. There are, for example, positron or advantages regarding the accuracy, frequency and
photon emission (PET, SPECT) (Parker et al., 1994; resolution of the reconstructed images and thus ap-
plications can be identified. Furthermore, for indus-
trial applications, the cost of the system as well as the
fault-tolerance are decisive.
*Corresponding author. Tel.: + +49-(0)51 l-762-3826; fax: Most applications of process tomography in the
+ +49-(0)51 l-762-3031; e-mail: nico@c36.uni-hannover.de. chemical industry require the imaging of fast changing,
2111
2112 N. Reinecke and D. Mewes

the so-called transient processes. This is the case for


most multiphase flows of interest. It requires the
tomography system to have the capability of very fast
data acquisition and for control purposes, often on-line
image reconstruction. Furthermore, the measurement
volume is often not a simple circular pipe but a rather
complex geometry, such as process equipment con-
taining impellers or catalysts (McKee et al., 1994; Xie
et al., 1995). In addition to that, more than two differ-
ent phases can be present, making the imaging tasks
even more difficult.
The application of tomographic measurement tech- measurement
volume step k
niques for research purposes poses different demands
on the tomographic system. Here, the accuracy and
Ul
the resolution of the imaging system are very impor-
tant as well as the measurement speed. The necessity
for on-line image reconstruction is often relaxed, so
that more sophisticated reconstruction algorithms
can be applied off-line. Furthermore, in laboratory
applications the robustness of the system is not strictly
required.
One tomographic measurement technique that is
well suited to both research and industrial applications
is capacitance tomography. The sensors required for
the imaging are robust and relatively cheap and the measurement
step k+l
actual data generation and acquisition well reported volume
(see, for example, Xie et al., 1995). An on-line
reconstruction of the measured phase distribution is (cl
possible using simple reconstruction algorithms such
as backprojection running on parallel computing
architectures. A more sophisticated iterative recon-
struction like ART yields better results but needs
more computing power (see, for example, Reinecke
and Mewes, 1994). Altogether, capacitance tomogra- U,
phy offers a very useful tool for a number of process
engineering applications. In this paper, the applica-
tion of capacitance tomography for the imaging of
two-phase flow in trickle-bed reactors is reported.
Particular interest will be paid to the highly instationary
flow regimes encountered in these types of reactors. step k+2

2. MEASUREMENT PRINCIPLE
Fig. 1. Schematic representation of the capacitance as
a function of the size and distribution of the permittivity.
In capacitance tomography, the interaction of a
permittivity distribution with an electrical field applied
For a homogeneous permittivity distribution, eq. (1)
to the measurement volume is used for the generation
can be simplified to yield the well known Laplace
of the integral measurements required for the recon-
equation
struction. The capacitance of two electrodes is a func-
tion of the permittivity of the components in the grad[s(x,y)] =0 + Acp(x,y) =O. (2)
measurement volume as well as their distribution.
The integral capacitance C which is measured by the
This is schematically shown in Fig. 1. The relation-
peripheral electrodes 1 and 2 can then be obtained
ship between the permittivity, the spatial distribution
through
of the permittivity and the resulting capacitance can
be derived from Maxwell’s equations (Maxwell, 1892).
Cc-= Q $j&,y)EdA
When the dielectric displacement D and the electrical
‘pz - ‘pl 42 -41
field density E are replaced by the spatial distribution
of the electrical potential cp(x,y) and permittivity = ffa&(x,y)grad(cp(x,y))dA
E(X,y), Poisson’s equation can be derived (3)
(P2 -‘PI

1 where cp2 - cpl is the potential difference between the


A+> Y) + - grad Cdx, ~11grad CG, Y)I =O.
4% Y) driving and measuring electrode and A the directed
(1) normal area of integration on the electrode surface.
Two-phase flow in trickle-bed reactors 2113

From eq. (1) it is evident that Poisson’s equation is exchange of the different components and a better
linear with reference to the electrical potential distri- assessment of the measurement errors associated with
bution but non-linear with reference to the permittiv- the individual part. The sensor generally consists of
ity distribution. Therefore, the integral capacitance peripherally mounted electrodes manufactured from
calculated from eq. (3) is a non-linear function of the conducting foil and a casing giving it mechanical
permittivity distribution. In contrast to any ray trans- stability and electromagnetic stray immunity. The
mission techniques, capacitance tomography is thus sensor electronics are used to do the sequential sam-
a non-linear technique yielding either very computa- pling of the electrodes, while the measurement instru-
tionally intensive iterative forward solutions or ment is used for the measurement of the capacitance
linearization procedures with inherent errors in the between the electrodes. The set of measurements is
reconstruction. In addition, it is also evident that an transfered to a computer, where the reconstruction is
inversion of eqs (1) and (3) with reference to the done.
permittivity distribution is not possible, which is com-
mon to most tomographic techniques mainly because 3. CAPACITANCE TOMOGRAPHY SYSTEM

of the severely underdetermined system of equations. 3.1. Sensor


Therefore, it is only possible to calculate the In Fig. 3 the primary capacitance sensor used for
capacitance from a known permittivity distribution the present study is shown schematically. It consists of
and not vice versa. three planes of identically segmented electrodes.
In order to obtain a two-dimensional tomographic A grounded shield is placed around them to ensure
image of a phase distribution inside the measurement electromagnetic stray immunity. Two sets of driven
volume, several linearly independent measurements of shield electrodes above and below the measurement
the capacitance between electrodes have to be made. electrodes are used to focus the electrical field to
The electrodes used for the measurement are placed a known volume. This was implemented to allow for
peripherally around the measurement volume. Rather a two-dimensional representation of the electric field
than having one sender and one receiver electrode as within the measurement plane, as well as a defined
would be used for linear transmission techniques like measurement volume. In Fig. 4 the resulting sche-
CAT, both electrodes are involved in generating an matic representation of the axial distribution of the
electrical field which is distorted by the permittivity electric field is shown. The electric field across the
distribution within the measurement volume. Due to measurement volume is very homogeneous. There-
the electrical field interactions between fields gene- fore, change of permittivity approaching the measure-
rated by different electrodes, the measurement of the ment volume only results in a change of the measured
different linearly independent capacitances is conduc- capacitance when the tip of the measurement elec-
ted by sequentially sampling different pairs of elec- trodes is reached. Due to the increase in homogeneity
trodes This can be explained using Fig. 1. In the first of the electric field and the lack of effects it is possible
cycle the capacitance is measured between electrodes to reduce the dimension of the measurement plane to
1 and 2, in the second cycle between 1 and 3, in the 30 mm. This reduction yields a more exact two-dimen-
third cycle between 1 and 4 and so on. Thus, a set of sional representation of the measurement and elimin-
linearly independent measurements is obtained. ates the axial integration effects.
In Fig. 2, a schematic representation of the When operating the 16-electrode sensor, a versatile
measurement chain for capacitance tomography is programming was developed. This allows for several
shown. It consists of the primary sensor system, of the electrode segments adjacent to one another
the sensor electronics and capacitance measurement to be used as one electrode. This procedure does
instrument and the computer for the reconstruction. not immediately increase the number of linearly
This type of generalization allows for a better independent measurements since they are determined

sensor

permittivity distribution
measurement volume
Fig. 2. Schematic representation of the measurement chain.
2114 N. Reinecke and D. Mewes

(a)

shield
electrode
(b)

Fig. 3. Schematic representation of the primary sensor.

by the smallest segmentation of the electrode plane, linearity, the spatial amplification and the resolution
i.e. the number of single electrode segments. With an in the center of the sensor (Reinecke and Mewes, 1994,
increase in the number of electrode segments per 1995a).
electrode, though, the signal level and thus the signal-
to-noise ratio (SNR) is increased, while the spatial 3.2. Sensor electronics and capacitance measurement
resolution is decreased. Considering two sensors with In capacitance tomography, the encountered
equal electrode area (and thus signal level) but differ- capacitances are usually high (several pF) with the
ent numbers of segments per electrode, the sensor variations to be detected being very small (typically
with the higher number of segments per electrode will a few fF) at very high sampling rates (typically
yield an increase in the total number of linearly inde- 5-15 kHz). The electrical current then flowing onto the
pendent measurements, since a smaller rotational electrode does so in only a finite time, setting a physical
step-size is possible. The total number of measure- limit to the sampling frequencies. In addition, the stray
ments per image determines the imaging speed and immunity of the circuit is very important to avoid
the image resolution. The electrodes not involved in extensive noise on the very small signal level.
a measurement are left floating, which results in vir- The measurement technique used in the present
tual source electrodes around the measurement vol- study consists of a commercial capacitance meter
ume. This increases the signal level, the SNR, the based on an AC-bridge capacitance measurement
Two-phase flow in trickle-bed reactors 2115

U, I 1-I u,=o

Fig. 4. Plot of the electrical field for driven guard electrodes.

low
pass

*
low
pass

-1
_---------- I I
I
I
I
II 18bit
DAC I

16 bit I
I
PC
8OC166 OAC I
18 bit I
OAC
I
reference impedance I
I
I
I feedback loop. balancing clrcult I
-----------------------~~----- --------_

Fig. 5. Capacitance measurement circuit.

circuit. This is shown schematically in Fig. 5. A signal tional to the capacitance is available. Using this tech-
generator is used to drive one electrode of unknown nique, accurate and stable capacitance measurements
capacitance. The resulting current on the other elec- are possible due to a low baseline drift and high SNR.
trode is converted to a voltage and then amplified. The same technique is also employed by other investi-
By using a reference voltage 90” phase shifted, a de- gators including Yang et al. (1992, 1993) and Klug
modulation is possible and an analog signal propor- and Mayinger (1993). Commercial instruments can
2116 N. Reinecke and D. Mewes
typically be operated at measurement frequencies of
14 kHz with accuracies of 0.5-0.1 fF. The accuracy of
the measurement increases with decreasing measure-
ment frequency. A further improvmment of the
measurement accuracy can be obtained when the AC-
bridge is balanced using a micro-controller in a feed-
back loop. This is indicated in Fig. 5 by the boxed
components. This technique is employed by Yang linearizationdata
(b) 1 [i.e. sensitivity]
et al. (1994). Due to the iteration measurement,
though, the sampling frequencies are somewhat lower.

3.3. Reconstruction
A simple and very fast reconstruction algorithm is
the backprojection algorithm. It is schematically
shown in Fig. 6(a). The algorithm corresponds to the
filtered backprojection algorithm popular for linear
tomography. The integral measurement value of every
Fig. 6. Schematic representation of the reconstruction
mesurement is distributed across the measurement
algorithms.
plane according to a weighting or sensitivity matrix
Si(x, y) which is unique for an electrode combination
i and obtained a priori by
With the exponent n = 1 the constant h* can be cal-
ACi culated by
Si(X2.V) = z
A-O
It represents a spatial amplification of the sensor
system (AC) at volume fractions of zero (&A-+O). Due
to the non-linearity of this imaging technique, this (8)
linear approach results in relatively large reconstruc-
tion errors. The iteration is continued until the error AC: is
For the present investigation, an iterative recons- below a given value. The reconstruction is done in a
truction technique was used. This is schematically Cartesian coordinate system with 32 x 32 pixels for
shown in Fig. 6(b). In the iteration, the forward calcu- the entire cross section of the imaging plane (Reinecke
et al., 1996).
lation is done using a finite difference calculation of
the electrical field and thus the capacitance. Due to
the computationally intensive FD calculation, the 3.4. Additional data-processing
reconstruction time is high (several seconds per 3.4.1. Time-correction. In most applications of
image) and an on-line reconstruction is not possible. capacitance tomography, the sequentially obtained
The optimization of the assumed permittivity and measurement values for the different combinations of
phase distribution is done using a set of sensitivities the electrodes are used for the reconstruction of one
defined in eq. (4). This calculation procedure is similar image. This is no problem if the imaging time is much
to the algebraic reconstruction technique (ART) used lower than the ratio of the length of the electrodes to
for linear tomography (Sweeney, 1972). For the elec- the structural velocity of the flow. If the structural
trode combination i and the iteration step k the velocity of the flow is high, the time lag between the
calculated dimensionless capacitance CR and the sequential measurements used for one frame is too
measured dimensionless capacitance C& yield the large. This can be the case for instationary flows in
error AC? trickle-bed reactors. Therefore, the velocity of the
measured flow field is limited not only by the frequency
ACTk = CR~i- Cg,i and C~lii =~[cx(x, y)] of the individual measurements but also by the frequency
of the reconstructed frames.
c-co Using a calculation method described by Reinecke
and C*=-.
CL -Co et al. (1995), a time correction between the sequential
measurements can be made. Therefore, the integral
The correction of the assumed phase distribution
measurements for one frame are all at the same virtual
a(x, y) is done using
time. Using an interpolation technique, it is possible
to increase the allowable transients of the measured
ah+’6, Y) = a’(~,Y) + N(x, Y),i, 4, (6) phases and still obtain images which are less blurred.
where h is an additive function of the sort The general shape of the interpolation function to be
fitted into the data can be determined beforehand by
continuously measuring every individual electrode
configuration.
Two-phase flow in trickle-bed reactors 2117

3.4.2. Velocity calculation. When the tomographic From eq. (11) it is evident that the derivative of the
sensor used for the measurement consists of only one measured integral void fraction across the sensor in
measurement plane without any axial extension, the the axial direction is equal to the difference between
correlation of signals from two sensors is the only the local void fraction entering and leaving the sensor.
possible way to achieve a measurement of the velocity. A structure entering the sensor at a time tl and pass-
In the case of capacitance tomography, though, the ing through the sensor without significantly changing
measurement plane of the sensor is extended in the will exit from the measurement volume at a time step
axial direction. For the measurement of the velocity r* later. If a pseudo-autocorrelation function of the
distribution, this axial extension of the measurement form
plane can be used for the calculation of the velocity of
the interphase within the measurement volume from
= @AU
the measured capacitances. ,+r
Due to the driven axial guard electrodes, the axial (12)
sensitivity of the sensor system is a homogeneous
distribution as shown in Fig. 4. The relative motion of is defined, it is possible to determine 7* from
the flow inside the measurement volume can be repre-
sented as the sensor moving over a given void fraction d@m
@ AKF + local maximum, or - =o. (13)
[Fig. 7(a)]. The resulting integral measurement is dr i*
given in Fig. 7(b) and can be calculated by
The absolute value of the maximum is only local since
.X+0 the autocorrelation function has its absolute max-
GI‘+r(x)
=; a(~‘) dx’. (9) imum for 7 =O. It is a global maximum in the sense,
sX
that QAKF(0)~0. It is therefore possible to calculate
Assuming a constant velocity u0 it becomes possible the average velocity of the two-phase flow from the
to replace x. Therefore, one can write integral capacitance measurements (Reinecke and
Mewes, 1995b).
EMCt& + dt)l - 4vo t)

=-[S
00
b
o,t+tr,dt+b

v,t+v,dr
cc(uot’)dt’ -
s
u,t+b

%I 1
cc(uot’)dt’

(10)
.
4. EXPERIMENTAL APPARATUS
In order to assess the application of capacitance
tomography for the imaging of two-phase flows in
trickle-bed reactors, an experimental facility was built.
Differentiating both sides of eq. (10) by dt and divid- In Fig. 8, a schematic representation of the experi-
ing by u. yields mental setup is shown. The liquid and the gaseous
phase (water and air) are fed into the column at the
h, 00 top and flow through the packing cocurrently. They
- =% [u(uot +b) -cr(uot)]. (11)
dt are separated at the bottom in a gas/liquid separator
and the liquid phase is recirculated. The peripheral
measurement techniques for the liquid and the gas
feed include temperature, flow rate and pressure as
(4
well as the pressure drop across the column. The
sensor temperature of both phases can be set as well as the
t / condition and concentration of other media.
In Fig. 9 a trickle-bed reactor is shown schemati-
cally. The feed of both phases into the column is
designed so that both gas and liquid enter the head
of the column through a large number of cocurrent
injection nozzles. Therefore, maldistributions at the
axial coordinate x
head of the column are avoided. The inner diameter of
the column is 120 mm with a total length of 2000 mm
of packing. The packings used are standard spheres of
@I , 10, 5 and 3 mm hydraulic diameter made of Celcore,
a ceramic material. The ceramic itself is porous, so
that the liquid phase wets the packing thoroughly.
In order to avoid errors from a high conducting
liquid-phase, deionized water was used. The conduc-
tivity of the water was thus kept below 1 mS/m. The
measurement plane was located at the bottom of the
column, well upstream of the support tray in order to
axial coordinate x avoid interference. The calibration of the capacitance
tomography was done with a fully wetted and a
Fig. 7. Schematic representation of the measurement. drained column. The full calibration was conducted
2118 N. Reinecke and D. Mewes

Fig. 8. Experimental setup.

with the packing being wetted for 30min in order to 5.1. Tomographic images
assure fully soaked ceramic particles. In Fig. 11, examplary results from the tomographic
imaging of the pulse flow in trickle-bed reactors
are shown. Plotted are individually imaged cross-
5. EXPERIMENTAL RESULTS sections of the pipe during the flow. The correspond-
In general, four different flow regimes encountered ing time goes from left to right and from top to
in trickle-bed reactors are characterized. These are bottom. The time scale between the images is 15ms.
plotted schematically in Fig. 10. For small flow rates The blue color represents the liquid phase while the
of both phases, the trickle flow regime is encountered whiter colors represent an increase of the gas phase.
[Fig. 10(a)] where the catalysts are completely wetted A void fraction of 1 is marked black (transparent).
and the gaseous as well as the liquid phase is conti- From these images it can be seen that the pulse flow
nuous. An operation below this point will yield a itself is established in a strongly three-dimensional
break-up of the liquid film into rivulets. The resulting manner. In the radial direction, even during the pas-
incomplete wetting of the packing makes the opera- sage of a pulse, the distribution of the void fraction
tion of most reactors in this flow regime not sensible. changes. Especially for the liquid-rich zones (pulse)
When the flow rates of both phases are increased, the the void fraction is never 0, so that a bubble flow
pulse flow regime is encountered [Fig. 10(b)]. Here, within these zones can be assumed. This is in accord-
the liquid is bridging the gaps between the catalysts ance with other authors, who have studied pulse flow
and accumulates to form a plug which is then accele- in two-dimensional flat-bed reactors (Melli et al.,
rated through the column by the expansion of the 1990, 1991).
initially blocked gaseous phase. When the flow rate of In order to understand better the axial/time devel-
the gaseous or the liquid phase is increased further, opment of the flow field, a different representation
the spray flow regime [Fig. 10(c)] and the bubble flow of the images has to be chosen. In Fig. 12 the data
regime [Fig. 10(d)] are encountered, respectively. from Fig. 11 and additional other data points are
In order to fully assess the possibilities for the applica- plotted differently. From the individual images in
tion of capacitance tomography for the tomographic Fig. 11, the central radial lines were taken and
imaging of trickle-bed reactors, highly transient flow plotted one after the other. Therefore, if these lines
regimes are desirable. In these flow regimes, the neces- are plotted from bottom to top, the direction of flow
sary time constants for the measurement are small and would be from top to bottom as is the case here.
up to date no information on the gas and liquid distribu- This type of representation has already been chosen
tion are available. Therefore, most of the tomogramms for other tomographic investigations and is usually
were measured for the pulse flow regime or for flow refered to as Eulerian slices (Hallow et al.,
regimes close to the transition into pulse flow. 1993). Given a constant and homogeneous velocity
Two-phase flow in trickle-bed reactors 2119

+
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
\
P

liquid

Fig. 9. Schematic representation of a trickle-bed reactor.

distribution the images are identical to three-dimen- trailing edge of this zone is identical. For an increase
sional representations of the flow field. The color of the gas flow rate, the front of the gaseous zone loses
representation is identical to Fig. 11. The volumetric its plane shape, while the trailing end remains almost
liquid flow rate and the gas flow rate is increased the same. The general shape of the gas-rich zone can
from Figs 12(a)-(d). best be described by the shape of Taylor bubbles in
As was already evident from the individual frame the vertical two-phase flow in simple pipes.
plots in Fig. 11, the pulses passing through the As the gas flow rate is increased, the pulse frequency
column are aerated with bubbles and not of constant and thus the thickness of the liquid zones is decreased.
thickness. The three-dimensional shape and extension With the decrease in the thickness of the liquid
of the pulse is clearly evident and different for all zone, a decrease in the strong three-dimensional
the pulses. This observation cannot be made from the effect is observed. This can be attributed to an
outside of the column. From visual observations increase in the momentum forces inside the column,
from the outside, the pulses appear to be one-dimen- which make a radial dispersion more difficult.
sional. This can be verified by plotting just the An increase in the frequency can also be observed
void fraction values from the wall region, where a when the liquid flow rate is increased. The shape
constant pattern will appear. Focussing on the shape of the gaseous zones becomes more pronounced
of the gas-rich zones inside the column, it can be while the radial distribution also becomes more
noticed that for low gas flow rates the leading and homogeneous.
2120 N. Reinecke and D. Mewes

Fig. 10. Schematic representation of the flow regimes in trickle-bed reactors.

5.2. Integral void fraction increase in the liquid velocity, the liquid holdup in
Using the tomographic images obtained it is also the column also increases. When the flow regime is
possible to calculate an integral void or liquid fraction reached, the liquid holdup in the gas-rich zones is
as a function of time. This is done not from the actual almost constant. Therefore, additional liquid fed
capacitance values measured, but rather from the into the column must be inside the liquid-rich
reconstructed images. This way, the non-linear relation- zones either by increase in size or frequency of the
ship between the liquid fraction and the capacitance pulses. The same can be observed in Figs 13(c) and
values of the individual measurements does not yield (d) where the same plots are given for the smaller
any error for the calculation. particle size. Even though the overall value of the
In Fig. 13, four plots of the liquid holdup as liquid holdup is increased due to a decrease in the
a function of time are shown. In Figs 13(a) and (b) particle size, the general behavior of the system is
the particle size is dp = 10mm while in Figs 13(c) identical above.
and (d) the particle size is dp =3 mm. In Fig. 13(a), In Fig. 14, the average liquid holdup is plotted as
the liquid superficial velocity is kept constant at a function of the superficial gas velocity [Fig. 14(a)]
jF =8.25 x 10m3 m/s and the superficial gas velo- and the superficial liquid velocity [Fig. 14(b)].
city is varied. For an increase in the gas velocity In these plots, the effect discussed above is clearly
the liquid holdup decreases. When the pulse flow visible, with the liquid holdup decreasing for an
regime is reached, the liquid holdup in the liquid- increase in the superficial gas velocity and the liquid
rich zones increases strongly while the liquid hold- holdup increasing for an increase in the superficial
up in the gas-rich zones decreases further. This liquid velocity. When the liquid holdup as a function
effect is due to the depletion of liquid from the gas- of time is integrated, the increase and decrease, respec-
rich zone through pulses traveling over the liquid tively, of the holdup seems to be independent of the
film and accumulating liquid. In Fig. 13(b) the gas flow regime. Even though the pulse flow regime is
superficial velocity is kept constant at jc =O.l2m/s established, the monotony of the curve remains
and the superficial liquid velocity is varied. For an the same.
Two-phase flow in trickle-bed reactors 2121

liquid holdup

Fig. 11. Tomogramms of the two-phase flow in trickle-bed reactors, time plots
2122 N. Reinecke and D. Mewes

liquid holdup top View side View


A-A B-B

nn

a b C d e

v)

o+sensor
0.3 0,3 m/s
10 5 10e3m / s

Fig. 12. Tomogramms of the two-phase flow in trickle-bed reactors, axial plots.
Two-phase flow in trickle-bed reactors 2123

1.0

0,8 - Column diameter d = 1ZOmm Ir =a.3 10-3 m/s


particle diameter d, = 1Omm - j, = 0.06 m/s
z ---- j,=0,16m/s

+ 0.6 -..-..
------ ,G
J,=0,20m/s
= 0.21 m/s -

(4 time t

1.0

Column diameter d - 120mm J6=0.13m/s


0.8 ~ particle diameter d, = 1Omm ___ ],=4.410-‘“l/S -
zi ---- JF=12,510-‘m/S
-..... Jr=16.510~3~/S
’ 0,6
~.- -. JF = 24.5 lo-’ m/s

““0.0 0.5 I,0 I,5 2,0 s 2.5

(b) time I

08 Column diameter d 3 120mm


partule diameter d, = 3mm

z-

4 0.6

z
; 0,4
‘3
superfirml veloches
JF = 8.3 10-3 m/s
f&2
- JG=0.02m/S ---- JG=0.09”,/S
.-..-- J0=0,12m/S -.-.-. JG=0.15”,/S

o,o
I I
o,o 0,s I,0 1,s 2,0 s 2.5

Cc) time I

Fig. 13. Liquid holdup as a function of the gas and liquid superficial velocity and as a function of time.
2124 N. Reinecke and D. Mewes

“OC
Columndmmeterd = 1ZOmm
0,8 particle
diameter d, - 3mm

z7

4 0,6
z
0
G 0.4
S I I I
3 I I I
superflrlalvelocltles
0.2 ~~=0,13m/s

_ ~~=4,41O'm/r ---- ~~=8.310.'m/s

-.---- jr=12.510'm/s ------ ~F=20,9103m/s

I I I
0,o
w 0.5 18 1.5 2.0 s 2,5
(d) time t

Fig. 13. Continued.

G-4 superficial gas velocity j,

1.0
parhcle dtameter
dp=3mm d,=lOmm

0.8 - 0 ~,=0,06m/s . j,=0,13m/s

0 ,‘.O.ll m/r . j,=O.ZOm/s

v I
= 0.6

@I
oo!
’0 5 10
superficial
15
liquid velocity j,
20 25 mm/s 30

Fig. 14. Liquid holdup as a function of the gas and liquid superficial velocity.
Two-phase flow in trickle-bed reactors 2125

(4 time t

I,0
1/s:
k OS
2
.-
: @O
El
.g -0,s
f
Xi
t
8 -1.0
0
2
-1.5

-2,0
0,t
(b) time 7

Fig. 15. Liquid holdup and calculated autocorrelation function.

5.3. Velocities developed sensor system is shown. The sensor consists


Using the data shown in Fig. 13 and the equation of 16 peripherally mounted electrodes, which are
derived in eq. (12), it is possible to calculate the velocity sampled sequentially. The measurement of the
of the pulses traveling through the column. In Fig 15(a) capacitance is done using an AC-bridge and the
the liquid holdup as a function of time is plotted. The reconstruction is done using an iterative ART-algo-
pulses can clearly be distinguished as well as the gas- rithm with a full electrical field calculation. It is shown
rich zones. In Fig. 15(b) the autocorrelation function how the data-processing can be improved for this
proposed in eq. (12) is plotted as a function of the time application and how an additional velocity measure-
step t. A clear maximum can be observed for about ment can be made.
0.1 s. These data were taken with a sensor with The results of the measured tomogramms show the
a slightly longer axial extension of 75 mm. The corre- possibilities of the technique presented. The three-
sponding velocity of the pulses is therefore about dimensional shape of the gas- and the liquid-rich
jPulse = 0.75 m/s. This is also in accordance with the zones inside the column for the pulse flow are visualiz-
data that can be obtained from the literature. ed. It can be seen that this quality of data cannot be
determined from the outside of the column. When the
resulting integral values for the liquid holdup are
6. CONCLUSIONS plotted, it is evident that in the pulse flow regime, the
In this article the possibilities of tomographic liquid is depleted from the gas-rich zones to accumu-
imaging of the two-phase flow in trickle-bed reactors late inside the liquid-rich zones. The overall liquid
using capacitance tomography are discussed. A newly holdup remains unaffected by the flow regime. Using
2126 N. Reinecke and D. Mewes

the data-processing described, it is possible to calcu- Subseries Thermotheraphy: Methods of Hyperther-


late the velocity of the pulses. mia Control, ed. M. Gautherie, pp. 35-111. Springer,
Berlin.
Fordham, E. J., Hall, L. D., Ramakrishnan, T. S.,
Acknowledgements Sharpe, M. R. and Hall, C. (1993) Saturation gradi-
The author wish to thank the ‘Deutsche Forschungs- ents in drainage of porous media: NMR imaging
gemeinschaft’ for the funding of the project. measurements. A.I.Ch.E. J. 39, 1431-1443.
Gladden, L. F. (1994) NMR in chemical engineering.
NOTATION
Chem. Engng Sci. 49, 3339-3408.
Gladden, L. F. (1995) Industrial aplications of NMR
A area, m2
imaging. Chem. Engng J. 56, 149-158.
c capacitance, (As/l/ = F) Halow, J. S., Fasching, G. E., Nicoletti, P. and Spenik,
D electrical displacement, As/m2 L. (1993) Observations of a fluidized bed using
E electrical field density, V/m capacitance imaging. Chem. Engng Sci. 48,643-659.
Q charge, As =C Ikeda, T., Kotani, K., Maeda, Y. and Kohno, H.
s sensitivity, dimensionless (1983) Preliminary study on application of X-ray
b axial sensor length, m CT scanner to measurement of void fraction in
d diameter, m steady state two-phase flows. J. Nucl. Sci. Technol.
20, l-12.
h correction function, dimensionless
Isaksen, 0. and Nordtvedt, J. E. (1992) Capacitance
i electrode combination, dimensionless
tomography: reconstruction based on optimization
j superficial velocity, m/s theory. In Proceedings of ECAPT’92, Manchester,
k iteration step, dimensionless U.K., Process Tomography-A Strategy for Indus-
n number of electrodes, exponent, dimen- trial Exploitation-1992, eds M. S. Beck et al.,
sionless pp. 178-189.
t time, s Kantzas, A. (1994) Computation of holdups in
V velocity, m/s fluidized and trickle-beds by computer-assisted to-
X coordinate, m mography. A.I.Ch.E. J. 7, 1254-1261.
coordinate, m Klug, F. and Mayinger, F. (1993) Novel impedance
Y
measuring technique for flow composition in multi-
phase flows. In Proceedings of ECAPTP3, Karls-
Greek letters
ruhe, Germany, Process Tomography-A Strategy
@ autocorrelation function, dimensionless
,for Industrial Exploitation-1993, eds M. S. Beck
M fraction, dimensionless et al., pp. 152-155.
& permittivity, As/Vm Maxwell, J. C. (1892) A Treatise on Electricity and
E void fraction, dimensionless Magnetism. Clarendon Press, Oxford.
cp potential, V McKee, S. L., Williams, R. A. and Boxman, A. (1994)
T time step, s Development of solid-liquid mixing models using
tomographic techniques. In Proceedings oj
Subscripts and superscripts ECAPT’94, Oporto, Portugal, Process Tomogra-
phy --A Strategy for Industrial Exploitation-1994,
0, 1,2 numbering
* eds M. S. Beck et al., pp. 342-353.
dimensionless
Melli, T. R., De Santos, J. M., Kolb, W. B. and
A differential
Striven, L. E. (1990) Cocurrent downflow in net-
F liquid works of passages. Microscale poots of macroscale
G gas flow regimes. Ind. Engng Chem. Res. 29,2367-2379.
M measured Melli, T. R. and Striven, L. E. (1991) Theory of two-
P particle phase cocurrent downflow in networks of passages.
R calculated Ind. Engng Chem. Res. 30, 951-969.
Mewes, D. and Fellhiilter, A. (1993) Observation of
mixing phenomena in wind tunnels by optical to-
REFERENCES mography. In Proceedings of ECAPT’93, Karls-
Abdullah, M. Z., Dyakowski, T., Dickin, F. J. and ruhe, Germany, Process Tomography-A Strategy
Williams, R. A. (1993) Observation of hydrocyclone for Industrial Exploitation-1993, eds M. S. Beck
separator dynamics using electrical impedance to- et al., pp. 289-292.
mography. In Proceedings of ECAPTP3, Karls- Parker, D. J., Hawkesworth, M. R. and Beynon, T. D.
ruhe, Germany, Process Tomography-A Strategy (1994) Process applications of emission tomogra-
for Industrial Exploitation-1993, eds M. S. Beck phy. In Proceedings of ECAPT’94, Oporto, Portu-
et al., pp. 93-96. gal, Process Tomography-A Strategy for Industrial
Boddem, M., Reinecke, N. and Mewes, D. (1994) Exploitation-1994, eds. M. S. Beck et al., pp.
Measurement of two-dimensional phase distribu- 478-489.
tion using a wire-mesh sensor. In Proceedings of Reinecke, N. and Mewes, D. (1994) Resolution
ECAPTP4, Oporto, Portugal, Process Tomogra- enhancement for multi-electrode capacitance
phy-A Strategy for Industrial Exploitation-1994, sensors. In Proceedings of ECAPTP4, Oporto,
eds M. S. Beck et al., pp. 155-162. Portugal, Process Tomography-A Strategy for
Bolomey, J. C. and Hawley, M. S. (1990) Noninvasive Industrial Exploitation-1994, eds M. S. Beck et al.,
control of hyperthermia. In Clinical Thermology, pp. 50-61.
Two-phase flow in trickle-bed reactors 2121

Reinecke, N. and Mewes, D. (1995a) Improvement Wang, M. and Dickin, F. J. (1993) Signal phase shift
of linearity and resolution of multielectrode and compensation in resistive electrical impedance
capacitance sensors for the tomographic visualiz- tomography. In Proceedings of ECAPT’93,
ation of transient two-phase flows. In Proceedings of Karlsruhe, Germany, Process Tomography--A
the Symp. on Flow Visualization and Image Process- Strategy for Industrial Exploitation-1993,
ing of Multiphase Systems, ASME, Fluids Engineer- eds M. S. Beck et al., pp. 193-196.
ing Division, Hilton Head, South Carolina, USA, Wang, M., Dickin, F. J., Williams, R. A., Waterfall,
eds W.-J. Yang et al., pp. 259-266. R. C. and Beck, M. S. (1994) Electrical resistance
Reinecke, N. and Mewes, D. (1995b) Calculation of tomography on metal walled vessels. In Proceedings
interface velocities from single plane tomographic of ECAPTP4, Oporto, Portugal, Process Tomogra-
data. In Proceedings of ECAPTP4, Oporto, phy-A Strategy for Industrial Exploitation--1994,
Portugal, Process Tomography-A Strategy for In- eds M. S. Beck et al., pp. 183-170.
dustrial Exploitation-1994, eds M. S. Beck et al., Xie, C. G., Reinecke, N., Beck, M. S., Mewes, D. and
pp. 41-50. Williams, R. A. (1995) Electrical tomography tech-
Reinecke, N., Boddem, M. and Mewes, D. (1995) niques for process engineering applications. Chem.
Improvement of tomographic reconstructions by Engng J. 56, 127-133.
time-correction of sequential measurements. In Xie, C. G., Huang, S. M., Hoyle, B. S., Lenn, C. P. and
Proceedings of ECAPTP4, Oporto, Portugal, Beck, M. S. (1992) Transputer-based electrical
Process Tomography-A Strategyfor Industrial Ex- capacitance tomography for real-time imaging of
ploitation---1994, eds M. S. Beck et al., pp. 381-392. oilfield flow pipelines. In Proceedings of ECAPTP2,
Reinecke, N., Buchmann, M., Petritsch, G. and Manchester, U.K., Process Tomography-A Strat-
Mewes, D. (1996) Enhancement of resolution for egy fbr Industrial Exploitation-1992, eds M. S.
reconstruction algorithms for limited view absorp- Beck et al., pp. 281-294.
tion spectroscopy. International Seminar on Op- Yang, W. Q. and Scott, A. L. (1992) Low value
tical Methods and Data Processing in Heat and capacitance measurements for process tomography.
Fluid Flow. IMECHE, London, U.K. In Proceedings of ECAPT’92, Manchester,
Seiraffi, M. A. (1993) Ultaschall-Computertomog- U.K., Process Tomography--A Strategy for Indus-
raphie zur Messung von Zweiphasenstriimungen. trial Exploitation-1992, eds M. S. Beck et al.,
Fortschrittsberichte VDI S(327), Diisseldorf. pp. 74-81.
Simons, J. R. S., Seville, J. P. K., Broadbent, C. J., Yang, W. Q. and Scott, A. L. (1993) Tomographic
Bridgwater, J., Parker, D. J., Beynon, T. D. and techniques for process design and operation, CEC
Hawkesworth, M. R. (1993) The study of particle Brite Euram Computational mechanics publica-
motion in fluidised beds using positron emission tions.
tomography. In Proceedings of ECAPT’93, Karls- Yang, W. Q., Strandos, S. K., Scott, A. L., Beck, M. S.
ruhe, Germany, Process Tomography-A Strategy and Johansen, G. A. (1994) A high frequency digital
for Industrial Exploitation-1993, eds M. S. Beck signal generator for capacitance tomographic imag-
et al., pp. 261-264. ing systems-requirements and realisation, In Pro-
Sweeney, D. W. (1972) Interferometric measurement ceedings of ECAPTP4, Oporto, Portugal, Process
of three-dimensional temperature fields, Disserta- Tomography-A Strategy for Industrial Exploita-
tion, University of Michigan, Michigan. tion-1994, eds M. S. Beck et al., pp. 182--193.
Toye, D., Marchot, P., Crine, M. and L’Homme, G. Yu, Z. Z., Peyton, A. J., Beck, M. S., Conway, W. F.
(1996) Modelling of multiphase flow in packed beds and Xu, L. A. (1993) Imaging systems based on
by computer-assisted X-ray tomography. Meas. Sci. electromagnetic tomography (EMT). Electron. Lett.
Technol. I, 436-443. 29, 625-626.