Chemical Engineering Science 54 (1999) 4929}4935

Dynamic #ow behaviour in bubble columns

S. Becker *, H. De Bie , J. Sweeney

Praxair N.V., Nijverheidsstraat 4, 2260 Oevel, Belgium

Praxair, Inc., 777 Old Saw Mill River Road, Tarrytown, NY 10591-6714, USA

Abstract

A detailed experimental study on the time-dependent #ow behaviour of the liquid phase in bubble columns was performed to
provide data for the validation of dynamic #ow models. It focuses on the dynamics of circulation #ows employing laser Doppler
anemometry. In a #at bubble column the velocity time series in lateral and vertical direction are similar to undamped harmonic
oscillations. Amplitude, frequency and phase angle describe them when superimposed turbulent #uctuations are removed by low-pass
"ltering. Lateral and vertical pro"les of these properties prove the local character of the hydrodynamics. In a cylindrical bubble
column the velocity time series are chaotic and not predictable. A broad range of low-frequency oscillations marks the unsteady #ow
behaviour. Nevertheless, regular #ow patterns with periods of 20 s and more are sometimes observed. It is assumed that they are
linked to the radial movement of the bubble hose in the sparger zone.  1999 Elsevier Science Ltd. All rights reserved.

Keywords: Bubble column; Gas}liquid #ow; Hydrodynamics; Flow dynamics; Laser Doppler anemometry

1. Introduction #ow simulations (see e.g. Sokolichin & Eigenberger,

1994; Lapin and LuK bbert, 1994; Delnooij, Kuipers & van
Flow phenomena in bubble column reactors have Swaaij, 1997a) because they were considered to approach
a broad range of characteristic time scales: (I) the turbu- a two-dimensional #ow model, and they facilitate de-
lent #uctuations are in the range of milliseconds and less; tailed experiments. Due to these e!orts the characteristic
(II) the liquid mixing through gross circulation #ows dynamic #ow behaviour of the bubble plume in a locally
needs seconds or minutes; (III) the change of material aerated #at bubble column is qualitatively well-known
properties, e.g. the increase of liquid viscosity during fer- (Becker, Sokolichin & Eigenberger, 1994; Delnooij,
mentation, can take hours. This time-dependent #ow Kuipers & van Swaaij, 1997b) and it is considered as
behaviour has to be taken into consideration for gas} a benchmark for numerical simulations.
liquid processes such as mixing, absorption, chemical
reaction and heat transfer.
This study focuses on the dynamics of circulation #ows 2. Experimental
that are caused by non-uniform gas hold-up distributions
(range II). The objective is to provide detailed data for The main dimensions of the #at bubble column are
the validation of dynamic #ow models. To date their pre- 0.2 m width and 0.04 m depth. The height of the gas/
dictions are mostly compared with visual observations liquid dispersion is 0.45 m. For this study a single-ori"ce
or time-averaged velocity pro"les because of the lack of gas distributor (diameter of the hole: 1 mm) is located in
time-dependent data. the centre axis of the column and the gas #ow rate is 0.8
The experiments were performed in two bubble col- l/min (w "0.17 cm/s).
%
umns with di!erent geometry: a cylindrical vessel and a The cylindrical bubble column has a diameter of
#at apparatus with a rectangular cross-section. Flat bub- 0.295 m. The height of the gas/liquid dispersion is 2.5 m
ble columns have gained a special interest for dynamic (165 l volume). Two perforated plate spargers with 21 (for
w "0.15, 0.5, 1 cm/s) and 89 holes (for w "2 cm/s)
% %
are employed. The plates have a diameter of 0.28 m and
uniformly distributed holes (0.7 mm diameter) in a quad-
* Corresponding author. Tel.: 00-32-14-247652; fax: 00-32-14-581505. ratic pitch of 52 and 25 mm, respectively (free areas: 0.012
E-mail address: stefan}[email protected] (S. Becker) and 0.052%).

0009-2509/99/$ - see front matter  1999 Elsevier Science Ltd. All rights reserved.
PII: S 0 0 0 9 - 2 5 0 9 ( 9 9 ) 0 0 2 1 4 - 6
4930 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935

All experiments are performed with an air/tap water signi"cantly di!ers from the observed instantaneous #ow
system at a temperature between 20 and 253C. The liquid structure.
phase is operated in batch mode. Both columns are made Corresponding to the visual observations the time
of acrylic glass to enable laser Doppler anemometry series of the lateral and vertical velocities have a periodic
(LDA) measurements. A two-component Aerometrics outlook (Fig. 2). High-frequency #uctuations generated
phase Doppler particle analyser is operated in backscat- by turbulence are superimposed to the low-frequency os-
ter mode. The measured data are analysed with Aeromet- cillations. To separate the #ow phenomena with di!erent
rics' Dataview software. Seeding particles are added to characteristic time scales a digital FIR low-pass "lter
increase the data rate (metallic coated glass spheres of 2.6 (cut-o! frequency: 0.5 Hz, "lter order: 200) with a Ham-
g/cm density and 12 lm mean size). It is assumed that ming window function is employed. The processed time
the contributions of bubble burst signals to the averaged series emphasise the periodic #ow behaviour and enable
velocities are negligible. the characterisation of the low-frequency oscillations
with their amplitude, phase angle and frequency (see
example in Fig. 3).
3. Flat bubble column
3.1. Amplitudes
The bubbles are rising in a meandering way (Fig. 1).
The bubble hose is slowly moving in lateral direction The amplitudes of the low-frequency oscillations
(observed period: about 16 s). It nearly touches the side strongly depend on the location in the #at bubble col-
walls. A gross circulation #ow of the liquid phase con- umn. Close to the side wall (x"0.5 cm) the lateral velo-
tinuously changes the #ow direction due to the move- city is dampened through the wall in#uence (Fig. 3). The
ment of the bubble hose. The single-ori"ce sparger amplitudes reach the maximum values in the centre axis
produces a broad bubble size distribution with a mean (x"10 cm). In vertical direction the amplitudes have a
size of about 5 mm. Some "ne bubbles of 1 mm and less local minimum in the centre axis (x"10 cm) where the
circulate with the liquid phase. bubble hose spreads out due to dispersion and nearly no
The time-averaged #ow structure obtained by LDA downward #ow occurs. To compare the amplitudes in
measurements consists of two vortex cells with upward the centre plane of the #at column (z"2 cm) the
#ow in the middle and downward #ow near the walls. It standard deviations p and p of the low-frequency
TV TW

Fig. 1. Lateral movement of the bubble hose in the #at bubble column (gas #ow rate: 0.8 l/min).
 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935 4931

Fig. 2. Time series of the lateral and vertical velocities in an exemplary measuring point of the #at bubble column (y and x are the vertical and lateral
position, respectively).

u(t). The shift *u"2pf*t (in radians) between both time

series depends on the location in the column. In this
study *t is the smaller one of both time lags and is de-
"ned as positive if the oscillation of the vertical velocity
reaches its maximum amplitude before the lateral com-
ponent. E.g., *u is positive for all three graphs in Fig. 3.
A shifted phase angle of 903 corresponds to a time lag of
4 s. In the centre axis (x"10 cm) the shift is small and
the lateral and vertical time series oscillate nearly simul-
taneously (Fig. 5). In x"6 cm the oscillations are signi"-
cantly shifted through the wall in#uence.

3.3. Frequencies and periods

The observed frequency of the lateral bubble move-

ment is about 0.059 Hz (period: 16 s). Applying a fast four-
ier transformation to the measured time series reveals
a characteristic frequency between 0.052 and 0.062 Hz
(periods between 16 and 19 s) for each position. The
deviations might be caused by slightly di!ering operating
conditions because the movement of the bubble hose is
rather sensitive to, e.g., the gas #ow rate.
On an average of all LDA measurements, the lateral
Fig. 3. Low-frequency oscillations in three lateral positions of the #at and vertical oscillations have a period of 17 s (Fig. 6).
bubble column (y"25 cm; x"0.5 cm: close to side wall, x"10 cm: Along the centre axis the characteristic frequency is
centre axis).
doubled because the bubble hose passes these positions
twice per period. For y"41 and 44 cm the bubble hose is
widely spread out in lateral direction and the periodic
oscillations are calculated (Fig. 4). The maximum stan- behaviour of the liquid phase disappears. The sparger
dard deviations are between 8 and 10 cm/s in lateral and design and the bubble size distribution in#uence the
between 10 and 12 cm/s in vertical direction. In the de- dynamics of the laterally moving bubble hose. I.e., a spar-
gassing and re-circulation region the vertical velocities ger with a multi-ori"ce design (diameter of the holes:
decrease and the lateral velocities become more domi- 0.4 mm) has a period of 2 s less for the same gas #ow rate
nant. Dead zones are observed in the bottom corners. of 0.8 l/min.

3.2. Phase angles 3.4. Phase trajectories

The low-frequency oscillations of the lateral and verti- A phase space diagram is a useful tool to describe
cal velocities do not oscillate with the same phase angle simultaneous two-dimensional velocity measurements
4932 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935

Fig. 4. Standard deviations p of the amplitudes of the low-frequency oscillations in lateral (left) and vertical (right) direction.

Fig. 5. Local di!erences in the phase angles u of the low-frequency Fig. 6. Lateral pro"les of the period ¹ of the low-frequency oscillations
oscillations in lateral and vertical direction. in lateral and vertical direction (averages of the heights y"21, 25 and
29 cm).

(Borchers, Busch, Sokolichin & Eigenberger, 1999). The amplitudes and frequencies. Characteristic structures are
lateral and vertical velocities of the low-frequency oscilla- found in the #at bubble column (Fig. 7). The ellipse for
tions are the variables that specify the instantaneous state x"0.5 cm expresses the strong oscillations in vertical
of the system. The velocity trajectories comprise their direction and the damped oscillations in lateral direction.
 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935 4933

Fig. 7. Phase trajectories of the low-frequency oscillations in three characteristic lateral positions.

Fig. 8. Example of a measured velocity time series (w "0.5 cm/s) in the cylindrical bubble column (top). Auto-power spectral density APSD and
%
autocorrelation prove the existence of a low-frequency oscillation with a period of 20 s (bottom).

The centre of the structure corresponds to the time- #ow along the walls. Against that, the instantaneous
averaged mean velocity. In x"10 cm the vertical velo- #ow structures are marked by the irregular axial move-
city changes sign twice per period. Therefore, the traject- ment of vortical structures through the column
ory has the shape of the number `eighta. In x"6 cm the (Sokolichin & Eigenberger, 1994; Mudde, Groen & van
amplitudes of the low-frequency oscillations are similar den Akker, 1997). They cause low-frequency oscillations
for both components and the shape of the trajectory is that can be found in the power spectrum of the velocity
approaching a circle. time series. I.e., Mudde et al. (1997) measured dominant
frequencies of 0.1}0.2 Hz in bubble columns of 15 and
23 cm diameter.
4. Cylindrical bubble column These low-frequency oscillations are also detected in
this study. However, their dominant frequencies are
In a cylindrical bubble column the time-averaged #ow smaller than 0.1 Hz. As an example Fig. 8 shows the
structure of the liquid phase is proved to be a gross circu- measured time series of the vertical velocity in one posi-
lation #ow with upward #ow in the centre and downward tion (r/R"0.68, h/H"0.64) of the cylindrical bubble
4934 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935

column for w "0.5 cm/s. The auto-correlation function

%
and the auto-power spectral density (ASPD) prove the
existence of a periodic structure with a period of 20 s
("0.05 Hz). The averaged period of the velocity time
series in "r/R"'0.18 is 23 s. In the centre of the column
the oscillations are less pronounced. The low-frequency
oscillations generally consist of a broad range of frequen-
cies between 0.3 and 0.01 Hz. There are usually several
strong frequencies with comparable intensities. A clear
periodic behaviour as shown in Fig. 8 is an exception but
it shows that a regular structure can exist at least for
a period of time.
For further signal analysis the digital FIR "lter (cut-o!
frequency: 0.5 Hz, order: 200) is employed to remove the
turbulent #uctuations. Additionally, averaged velocities
of 3 s intervals are calculated. This second kind of "lter-
ing has the advantage that it is independent of the "l-
ter characteristics. Fig. 9 shows two examples of "ltered
times series. Low-frequency oscillations are perceptible Fig. 10. Standard deviations p of the low-frequency oscillations in
but they are much more irregular than in the #at column. T?
axial direction (low-pass "ltered through averaging intervals of 3 s). The
They cannot be described in the same simple way. The radial positions are r/R"!0.95, !0.68 and 0.
calculation of the Kolmogorov entropy through the soft-
ware RRCHAOS (Schouten & Van den Bleek, 1992}
1995) proves the chaotic character of the velocity time predictable position. The residence time in one position is
series. in the range of seconds and strongly varying.
It is assumed that the low-frequency oscillations are The amplitudes of the axial low-frequency oscillations
closely linked to the radial movement of the bubble hose are increasing with the super"cial gas velocity (Fig. 10).
in the sparger zone. Bubbles are pushed towards the wall The smallest standard deviations r are measured in the
by the downward #owing liquid and structures with column centre. Strong oscillations occur close to the wall
di!erent bubble number densities are created. Fischer, ("r/R""0.95) and in the region where the time-averaged
Kumazawa and Sada (1994) found them at any height of axial velocity changes sign ("r/R""0.68). Their ampli-
a bubble column through local gas hold-up measure- tudes are comparable.
ments. They measured periods of 12}20 s in a column The time-averaged velocities of the tangential com-
with a diameter of 0.1 m. Our own #ow observations of ponent are close to zero for all axial positions above the
the movement of the bubble hose in the sparger zone sparger zone. However, the low-frequency oscillations are
show some regular behaviour but it is de"nitively not also measured in tangential direction. The amplitudes
periodic. The hose moves in radial direction to a certain and the dominant low frequencies are similar to the one
position, stays for some time and then moves to an un- of the axial velocities.

Fig. 9. Exemplary low-frequency oscillations of the axial velocity close to the wall (left) and in the region where the time-averaged velocity changes sign
(right). The lines are low-pass "ltered time series, the circles mark the averaged velocities of 3 s intervals.
 S. Becker et al. / Chemical Engineering Science 54 (1999) 4929}4935 4935

5. Concluding remarks Indices

A periodic lateral movement of the bubble hose marks a axial direction

the #ow behaviour in a non-uniformly aerated #at bub- t tangential direction
ble column with a sparger located in the centre axis. The x lateral direction
velocity time series in vertical and lateral direction are y vertical direction
similar to undamped harmonic oscillations and can be
characterised by amplitude, frequency and phase angle.
The dynamic #ow behaviour is thus predictable. Flat Acknowledgements
bubble columns are therefore a useful tool for the valida-
tion of dynamic #ow models. Furthermore, this geometry Financial support by the European Commission
enables the visual observation of all #ow regions and (Brite-Euram project `ADMIREa, BE95-2039) is grate-
facilitates non-invasive optical measurement techniques fully acknowledged. The authors would like to thank the
as LDA. project partners for the productive cooperation.
The time-dependent #ow behaviour in a cylindrical
column is chaotic and not predictable. Some regular
structures are detected in the velocity time series in axial References
and tangential direction that lead to a kind of periodicity.
E.g., for w "0.5 cm/s the measured period of a domi- Becker, S., Sokolichin, A., & Eigenberger, G. (1994). Gas}liquid #ow in
% bubble columns and loop reactors: Part 2. Comparison of detailed
nant low-frequency oscillation is 23 s. However, there is
experiments and #ow simulations. Chemical Engineering Science, 49,
no low frequency that is characteristic for the whole col- 5747}5762.
umn. The velocity time series typically contain a broad Borchers, O., Busch, C., Sokolichin, A., & Eigenberger, G. (1999).
range of signi"cant low-frequency contributions. It is Applicability of the standard k-e turbulence model to the dynamic
assumed that the dynamic behaviour is linked to the simulation of bubble columns. Part II: Comparison of detailed
irregular radial movement of the bubble hose in the experiments and #ow simulations, submitted for publication.
Delnooij, E., Kuipers, J. A. M., & Van Swaaij, W. P. M. (1997a).
sparger zone. The low-frequency oscillations of the axial Computational #uid dynamics applied to gas}liquid contactors.
and tangential velocities have signi"cant amplitudes in Chemical Engineering Science, 52, 3623}3638.
any position of the column. Delnooij, E., Kuipers, J. A. M., & Van Swaaij, W. P. M. (1997b).
Dynamic simulation of gas}liquid two-phase #ow: e!ect of column
aspect ratio on the #ow structure. Chemical Engineering Science, 52,
3759}3772.
Notation Fischer, J., Kumazawa, H., & Sada, E. (1994). On the local gas holdup
and #ow pattern in standard-type bubble columns. Chemical Engin-
f frequency, Hz eering and Processing, 33, 7}21.
h vertical position in the cylindrical column, m Lapin, A., & LuK bbert, A. (1994). Numerical simulation of the dynamics
H height of the cylindrical column, m of two-phase gas}liquid #ows in bubble columns. Chemical Engin-
eering Science, 49, 3661}3674.
r radial position in the cylindrical column, m Mudde, R. F., Groen, J. S., & Van den Akker, H. E. A. (1997). Liquid
R radius of the cylindrical column, m velocity "eld in a bubble column: LDA experiments. Chemical
t time, s Engineering Science, 52, 4217}4224.
¹ period, s Schouten, J.C., Van den Bleek, C.M. (1992}1995) RRCHAOS: an inter-
v velocity, m/s active software package for deterministic chaos analysis of non-
linear time series. Delft, The Netherlands: Reactor Research
x lateral position in the #at column, m Foundation.
y vertical position in the #at column, m Sokolichin, A., & Eigenberger, G. (1994). Gas}liquid #ow in bubble
z position in the depth of the #at column, m columns and loop reactors: Part 1. Detailed modelling and numer-
w super"cial gas velocity, m/s ical simulation. Chemical Engineering Science, 49, 5735}5746.
%
Greek letters

u phase angle, radians or degrees

p standard deviation