Group Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Title: 14.3.4 - Measuring of particle size distributions in stirred tanks
ALL VOLUMES CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00102023/00350
 Material Information
Title: 14.3.4 - Measuring of particle size distributions in stirred tanks Experimental Methods for Multiphase Flows
Series Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Physical Description: Conference Papers
Creator: Maaß, S.
Grünig, J.
Kraume, M.
Publisher: International Conference on Multiphase Flow (ICMF)
Publication Date: June 4, 2010
 Subjects
Subject: particle size analysis
liquid/liquid dispersion
drop size distribution
measurement technique
 Notes
Abstract: An online measurement technique for drop size distribution in stirred tank reactors is needed but has not yet been developed. Different approaches, different techniques have been published as the new standard during the last decade. Three of them (FBRM®, 2D-ORM® and FBR-sensor) are tested and their results are compared with trustful image analysis results from an in-situ microscope. It is clearly shown, that the measurement of drop sizes in liquid/liquid distribution is a major challenge for all tested measurement probes and none provides exact results for the tested system of pure toluene/water compared to an endoscope. Not only the size analysis but also the change of the size over time gives unreasonable results. The influence of the power input on the drop size distribution (DSD) was the only reasonable observation in this study. Therefore image analysis has become a powerful tool for the work with particulate systems, occurring in chemical engineering. A major challenge is still the manual work load which comes with such applications. Additionally manual quantification also generates bias by different observers, which is shown in this study. Therefore the full automation of those systems is necessary. In this work we have introduced a MATLAB® based image recognition tool to count and measure particles in multi phase systems fully automatic. Up to now this software is limited to spherical particles but will be extended to irregular shapes as well. The program has reached hit rates of 95% with an error quotient under 1% and a detection of 250 drops per minute for simple images and hit rates of 40% with an error quotient of under 5% and a detection of 10 drops per minute for some difficult images. A parallelization on different cores in a multi-core PC will increase those calculation times dramatically. Therewith online process observation and control will be possible.
General Note: The International Conference on Multiphase Flow (ICMF) first was held in Tsukuba, Japan in 1991 and the second ICMF took place in Kyoto, Japan in 1995. During this conference, it was decided to establish an International Governing Board which oversees the major aspects of the conference and makes decisions about future conference locations. Due to the great importance of the field, it was furthermore decided to hold the conference every three years successively in Asia including Australia, Europe including Africa, Russia and the Near East and America. Hence, ICMF 1998 was held in Lyon, France, ICMF 2001 in New Orleans, USA, ICMF 2004 in Yokohama, Japan, and ICMF 2007 in Leipzig, Germany. ICMF-2010 is devoted to all aspects of Multiphase Flow. Researchers from all over the world gathered in order to introduce their recent advances in the field and thereby promote the exchange of new ideas, results and techniques. The conference is a key event in Multiphase Flow and supports the advancement of science in this very important field. The major research topics relevant for the conference are as follows: Bio-Fluid Dynamics; Boiling; Bubbly Flows; Cavitation; Colloidal and Suspension Dynamics; Collision, Agglomeration and Breakup; Computational Techniques for Multiphase Flows; Droplet Flows; Environmental and Geophysical Flows; Experimental Methods for Multiphase Flows; Fluidized and Circulating Fluidized Beds; Fluid Structure Interactions; Granular Media; Industrial Applications; Instabilities; Interfacial Flows; Micro and Nano-Scale Multiphase Flows; Microgravity in Two-Phase Flow; Multiphase Flows with Heat and Mass Transfer; Non-Newtonian Multiphase Flows; Particle-Laden Flows; Particle, Bubble and Drop Dynamics; Reactive Multiphase Flows
 Record Information
Bibliographic ID: UF00102023
Volume ID: VID00350
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: 1434-Maass-ICMF2010.pdf

Full Text


7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010


Measuring of particle size distributions in stirred tanks
S. Maaf3*, J. Grujnig* and M. Kraume*
Department of Chemical & Process Engineering, Technische Universitiit Berlin,
Strate des 17. Juni 135, Sekr. MA 5-7, 10623 Berlin, Germany
E-mail: Sebastian. maas siitu-berlin. de



Keywords: particle size analysis, liquid/liquid dispersion, drop size distribution, measurement technique






Abstract
An online measurement technique for drop size distribution in stirred tank reactors is needed but has not yet been
developed. Different approaches, different techniques have been published as the new standard during the last decade. Three of
them (FBRM 2D-ORM" and FBR-sensor) are tested and their results are compared with trustful image analysis results from
an in-situ microscope. It is clearly shown, that the measurement of drop sizes in liquid/1iquid distribution is a major challenge
for all tested measurement probes and none provides exact results for the tested system of pure toluene/water compared to an
endoscope. Not only the size analysis but also the change of the size over time gives unreasonable results. The influence of the
power input on the drop size distribution (DSD) was the only reasonable observation in this study.
Therefore image analysis has become a powerful tool for the work with particulate systems, occurring in chemical
engineering. A major challenge is still the manual work load which comes with such applications. Additionally manual
quantification also generates bias by different observers, which is shown in this study. Therefore the full automation of those
systems is necessary. In this work we have introduced a MATLAB" based image recognition tool to count and measure particles
in multi phase systems fully automatic. Up to now this software is limited to spherical particles but will be extended to irregular
shapes as well.
The program has reached hit rates of 95% with an error quotient under 1% and a detection of 250 drops per minute for
simple images and hit rates of 40% with an error quotient of under 5% and a detection of 10 drops per minute for some difficult
images. A parallelization on different cores in a multi-core PC will increase those calculation times dramatically. Therewith
online process observation and control will be possible.


Introduction

Developing adequate technologies for measuring
particle properties, as for instance the particle size, shape,
composition, velocity etc. are still of major interest. Even
though extensive research has been performed towards this
aim in the present and the past century, still much research
is necessary. The methods in most cases of much higher
sophistication than before, still aim at improved accuracy
and higher resolution, and there are, in spite of the almost
countless principles and instruments invented and
developed over the decades, still the odd new or refined
principles of measurement [1].
In this work we want to focus on measuring the particle
size of different multi phase application. There is still a
considerable lack of understanding, and lack of theoretical
interpretation and of experimental evidence of the
differences seen in the correlation of different equivalent
diameters, obtained when analysing the same particle
(regular or irregular shaped), or a distribution of particles,
using different physical principles. Furthermore, there is
still a lack of understanding of the influence of shape on the
results obtained with optical methods [1]. Most methods
applied suffer from the fact, that even though a size
distribution may be obtained in less than a minute, the
highly sophisticated mathematical background is based on
spheres. Aside from that even for spherical drops or bubbles


different authors found unsatisfying results for optical
measurement techniques while the existing shape influence
cannot be taken into account [2, 3]. MaaB et al. [3] are
questioning the reliability of laser based online probes in
general and suggesting rather the use of image analysis with
nowadays available powerful computers. Those are
necessary to overcome the long calculation times for a
reliable particle size distribution.
According to Pacek et al. [4], any technique based on
external plwsical sampling will change the measurement
results. Especially for technical applications with fast
coalescence or reactions such measurements are not useable
as totally different drop size distributions will be measured
for a sample taken out of the system. Even for sampling
times less than one second, the fast coalescence and drastic
change in the f low condition during sampling result in
significant measurement errors.
Additionally problems arise, if the size distribution to
be analysed covers a wide range of sizes. Most rapid
response methods used for particle size analysis only cover
a limited range of sizes without readjusting the instrument.






7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

reliable and accurate [4, 10], the endoscope results have
been used as a standard which against the other techniques
compete.
Accurate experimental results of DSD's are a necessary
base for testing and developing numerical models for
transient behaviour of drop size distributions.


Nomenclature


[m]
[m]
[m]
[-]
[#]
[1/s]
[-]
[-]
[1/m]
[m3]


Particle diameter
Sauter mean diameter
stirrer diameter
Refraction index
Number of particles
stirrer speed
Power number
Cumulative number distribution
Number density distribution
Reactor volume


] 03



E

e-c


Special characters
P [kg/m3]

Subscripts


Dispersed phase fraction
Density


Particle


Abbreviations
autom. Automatically detected particles
DAT Data acquisition time
DSD Drop size distribution
IPP Inline particle size probe
FBRM Focused beam reflectance measurement
FBR Forward-backward-ratio
fps Frames per second
MAT Measurement acquisition time
man. Manually detected particles
ORM Optical reflectance measurement


Measurement technique
The measurement applications can be divided into
following main groups: sound, laser and photo based
techniques (see Figure 1). The Lasentec FBRM" [5], the
2D-ORM" [6] provided by Messtechnik Schwartz GmbH
the IPP 30 [7] and the FBR-sensor [8] which all give online
and in-situ information and furthermore an in house
developed endoscope technique [9] have been analysed, to
determine an exact online measurement technique for drop
size distributions.

measurement techniques for drop sizes
in liquid/liquid dispersions



in situ external sample analysis


sound systems laser systems Coulter systems image analysis


102 103
specific power input PlV =Po-pN D5 [Wlm3]
VR
Figure 2: Comparison of the steady state Sauter mean
diameter for three different power inputs by varying
the stirrer speed for four measurement techniques [3].

The 2D-ORM" and the endoscope technique show
nearly the same proportionally of the Sauter mean diameter
over power input (represented by the black curves in Figure
2). So it should be possible to use the 2D-ORM" probe in
liquid/liquid dispersions to analyze and control the change
of DSD for different power inputs.
Also the FBRM" shows physical meaning full
behaviour. For higher energy dissipation rates the mean
drop size is decreasing. The exponent of P/V with -0.11 is
stronger then the exponent resulting from the image
analysis. Such proportionalities and even higher values in a
toluene/water system have been reported by Gilbler et al.
[11] at a pH 3. So it seems that the results from the FBRM"
do roughly reflect the influence of P/V on the steady state
Sauter diameter. It has already been presented in literature
that the deviations of the FBRM" to image analysis results
are reproducible and can be correlated with a linear relation
[12].
In Figure 3 the results of the endoscope for a stirrer
speed of 400 rpm are compared with the results of the other
used measurement techniques until a steady state is reached
(after 25 min). Two major results are obvious. Firstly, both
rotating laser probes measure much smaller drop sizes than
the endoscope and the FBR-Sensor. Secondly, it can be seen
that besides the endoscope no measurement technique
indicates the clear transient behaviour with decreasing drop
sizes of the system. The FBR-Sensor and the IPP-30 show
even an increase of drop sizes with an increase of stirring
time.
Summarizing all the challenges and misleading results,
it becomes obvious that choosing the "right" measurement
technique according to the application is of major
importance for reliable particle size distributions. Based on
the experiences of many scientist groups [1, 3, 10] we will


-acoustic -laser diffraction
-electroacoustic -spatial filtering
-light back scattering


-video technique
-endoscope
-Stereomicroscope


Figure 1: Overview of measurement techniques for liq-
uid-liquid systems.

While photo based methods working with image
recognition give accurate values for the drop sizes which are









focus on using in-situ image analysis techniques and work
on the challenges occurring with those applications.






-I endoscope
11I + IPP-30
E,,d ~L-9- FBR-sensor
a~~.. ....l~---- F BRM @
5 1 ~-M- 2D-ORM"
400 rpm, pH 7
~~~ ]I (p = 20%



O 5 10 15 20 25 30 35 40 45 50
time t [min]

Figure 3: Comparison of the resulting transient Sauter
mean diameter of five investigated measurement tech-
niques at constant stirrer speed (including data from
[3]).


Image analysis in multi phase systems

The use of image analysis to size particles like drops,
bubbles, pellets, cells or the like has a long history.
However, there are far more studies analysing off-line than
on-line sampling. For both applications, Junker gives a
broad overview and a detailed description of technical and
historical developments [13]. Leschonski and also Junker
emphasize the necessity of short data acquisition time
(DAT) and additionally a short measurement acquisition
time (MAT). Ideally the MAT is equal to the DAT [14].
The DAT is limited by the used hardware, especially
the camera. Standard cameras today allow a "real-time"
DAT, for example 30 frames/s (fps). Junker [13] gives a
broad overview of selected photographic techniques in
literature, CCD-cameras are the optimum for the effort /
cost ratio. Figure 4 shows an example image taken with
such a standard camera. Different particles can be observed
on this image. Ritter and Kraume [9] suggested a minimum
number of 200 particles building one distribution for a
liquid/1iquid system. Assuming that this is a reliable number
for statistical demands, 75 images like those shown in
Figure 4 are needed for a trustful data base. That means the
DAT would be less then three seconds for the described
application (for technical details see [3] and [9]).
Of course particle systems are very different so are
the number of particles on one image and therewith the
number of frames required for a single measurement. Junker
[13] reports in his studies a variation between 2 to 400
frames. The required number of objects per measurement
becomes important for storage reasons, if all images need to
be reviewed and therewith saved. This challenge should
become insignificant with the ongoing developments in the
area of computer science.


7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010


















Figure 4 Example image of the used frame set for the
results presented in Figure 5.

Particularly for particle sizing applications the MAT is
much greater then the DAT [13]. MaaB et al. [3] mention 30
min MAT for a single data point. This is in the range that
Junker [13] reports based on his literature studies (5 to 60
min). This means the DAT is only 1/600 of the MAT
because of the manual evaluation. An on-line system should
produce data within a process relevant timeframe.
Elsewhere process control is impossible. The manual
evaluation of such images is highly time-consuming
therefore automation of image analysis should be employed
to speed up the MAT at least about one magnitude.
Additionally automated quantification also avoids bias by
different observers.
Figure 5 shows such deviations between different
observers with each other and also with themselves for the
results of the Sauter mean diameter (d32 = C di /Cdl ) on the
left ordinate and number of counted particles on the right
ordinate. Four different participants counted the same frame
set twice, two of them carried out an additional third
quantification.


700
600

500 '

400 o

300

200o

100 a,
E
0 a


one two three four
manual quantification by observer
Figure 5 Analysis of the sensitivity of manual quanti-
fication of the Sauter mean diameter and the number of
observed drops in a single image batch at constant
process parameters.






7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

time consuming it is still not applicable for usage in online
process control.


Image recognition steps
In order to ensure best possible drop detection given
image series are pre-filtered to remove irrelevant and
misleading image data.
Lighting patterns and dirt on the lens given by the optic
system used to take images are the same for every image
and is shown in Figure 6 (b). These can be acquired and
removed from the sequence (Figure 6 (a) and (c)).
Also the contrast of each image is amplified depending
on local surroundings to ensure intensity invariant
recognition of drop borders (Figure 7).


An example image from the chosen analyzed frame set
is given in Figure 4. The image quality is very high due to
the high refraction index of the used system (toluene/water
with n =1.496) and the low dispersed phase fraction
(cp= 2%). That is why almost no overlapping effects occur.
Even with those "easy" images the four different probands
have a deviation of a 5% for the Sauter mean diameter
(presented as the open diamonds in Figure 5) and of a 15%
for the number of particles found in the whole set (presented
as the black squares). Surprisingly these deviations are
increased by repeating the same counting for both, the d32
(a 10%/) and the number of identified particles (A 30%).
Two probands also carried out a third quantification. The
standard deviation of one observer quantifying manually the
same frame set is around five percent. Concluding those
results and the impression from literature an automated,
trustful, online image analysis system ofparticulate systems
was developed and implemented in MATLAB". This
approach is able to analyse multiple batches of images for
various optical systems, to automatically identify the
contour of existing drops or bubbles and classify them
according to their diameter. Such efforts have already been
carried out and published in Literature [15, 16, 17].


Enlgure / Local contrast normalization.


It assu ingratiny us rrtru irriage:
acquisition device (b)


The program starts to search for circular patterns for a
range of possible radiuses. This is the most time consuming
and critical process of the detection. The results are
probabilities for each image pixel to be a centre point for
circles of all given radiuses (Figure 8). The intensities
behave proportional to the centre probabilities, white being
the lowest and the strongest tone (black) being highest
probability.


uriginal witnour naraware
acquisition falsification (c)


Figure 6 Pre-filtering during image recognition.

Scherze et al. [16] were analysing the drop sizes in
multiple emulsions. Due to their set-up and the limitations
in the amount of the dispersed phase in the system, almost
no overlapping occurred and the use of commercial
software to evaluate the drop size was possible. Up to now,
no commercial software for the analysis of highly
concentrated dispersions is available but needed for
industrial relevant applications [17]. More promising are the
works of Alban et al. [15] and Bris et al. [17]. Both groups
work with highly concentrated dispersions in stirred
vessels, like the ones presented in Figure 6 and developed
their own software to evaluate such data. Interestingly both
groups are also using MATLAB" for their image
recognition. Bris et al. [17] are using as the second step,
after the edge detection, the Hough transformation [18] to
evaluate the circles. While Hough transformation is very


Figure 8 Circle centre probability distribution over
the image.

The best possible centre points (these are local
probability maxima) and radiuses are then checked in detail
to finally be classified (Figure 9). This is done by
decomposing the possible drop circle into segments and






7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

in its founding working group. The major challenge is to
develop software which can be adapted to many multi phase
systems.


checking these for local validity to being part of a real drop
(Figure 10 (a) and (b)). Note that thanks to the contrast
enhancement borders can be evaluated with image intensity
invariance. The percentage of features needed to pass as
circles are very much dependent on the optical
characteristics of the system.


Centre probabilities (values over 50%) for
centre of blue marked circle radius
Figure 9 Circle centre probability pre-selection.


Figure 11 Example image taken from Hirschberg et al.
[19] with ten percent dispersed phase fraction. The blue
circles represent 154 automatically detected drops.

Therefore, four different systems were used for the
validation of the developed software:
1. Images from literature
2. Solid/liquid system (glass beds/water)
3. Liquid/liquid system (toluene/water)
4. Liquid/liquid system (n-butdl chloride/water)

To ensure a broad variety of applications, image data
from literature were used to test and engineer the software.
As the reason was for the development the analysis of
liquid/liquid system, such have been chosen in the first run
for validation. Hirschberg et al. [19] are analysing the
mixing efficiency of static mixers in various applications,
also in a liquid/1iquid system with two different dispersed
phase fractions. The results are shown in Figure 11 and
Figure 12.


Border property compliance Intensity Invariant border
for each angle (a) property compliance for
each angle (b)
Figure 10 Drop classification.

In each of the processes parameter optimization is
necessary to save computing time and improve the error
quotient and hit rate. This is achieved by marking a few
drops manually and letting the software extract their optical
characteristics. These include the steepness and regularity
of the drop border as general optical behaviour. These
values vary strongly from system to system but also in one
system solely depending on the drop radius.
The automatic drop recognition is most effective with
high contrast backlit images with a high drop density. The
high contrast ensures a good hit rate and low error quotient
while the density ensures that the computation renders more
drops recognized per minute. Nevertheless it has proven
itself to be very robust and able to compete with the human
eye in complex images. The computation time is nearly
proportional to the pixel- quantity being processed and
depending on the images twice to fifty times faster than a
person counting. The user is able to define thresholds for the
error quotient and the hit rate to force the program into
further parameter optimization.


RESULTS
Abroad evaluation is necessary to ensure the reliability
of the image algorithm software. Often an algorithm is
developed and only optimized according to the application


Figure 12 Example image taken from Hirschberg et al.
[19] with twenty percent dispersed phase fraction. The
blue circles represent 236 automatically detected drops.

These example images have been taken directly from
the publication, therefore the quality is not at its optimum.
However, the results show clearly the strengths of the new
software especially because the algorithm parameter could
kept constant for both process parameters. The calculation
















man. autom.
Qo [-1 I- 5*J
- G [1/mm] I
d32 [plm] 431 449
n [#] 247 556 4


7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

(n =556) than for the manual one (n =247). Those
inconsistencies lead to a difference of 4.7 percent in the
Sauter mean diameter: those values are also given in Figure
14.


times per frame were under two minutes (using a single core
of a Pentium 4 with 2.6 Ghz). Further developments of the
software using multi cores will decrease these times
substantially (factor four to eight).

,100


S80 -
70 -


50 -


30 -
n [#]
S20
.2 manual detection 640
Til 10
5 automatic detection 672
E 0
o 100 200 300 400 5
Particle diameter d, [pm]

Figure 13 Comparison of manual and automatic de-
tected particle size distributions for bimodal glass beds
from a frame set with 82 images; example image in the
graph.

Using a known size distribution of glass beds in water
was the second task to validate the software. A bimodal
population of two different diameter was used with
dmes~ = 200 pLm and dmes~ = 350 pLm. The deviation of the
mean diameters was around + 20%. Figure 13 shows the
comparative results between manual and automated
detected distributions and the number of detected particles n
in the whole frame set of 82 images. Both show almost the
same results and lead the authors to the next challenge of
analysing liquid/1iquid systems.
Here the already introduced "easv" system with a low
dispersed phase fraction was used (toluene/water; cp = 2%;
see Figure 4 as an example image).
After optimizing the parameter set for the algorithm,
quite promising results could be achieved. The cumulative
number distributions Qu(dP) Shown in Figure 14 are almost
equal for the manual (man.) and the automatic (autom.)
detection. They are presented by the dotted curves over the
particle diameter on the left ordinate. For a better
differentiation the number density distribution qu(dP) iS
plotted on the right ordinate, presented as straight lines.
With the following equation (1) the cumulative number
distributions have to be transformed into number density
distributions for a more detailed analysis:


1.0



aT 0.8



0.6



a
E 0.


S0.2
00O E


5.0

4.5,
E
4.0 .E

3.5 0-

3.0


.u,
2.0 >

1.5 C


E
0.5 '

0.0
)0


10


0.0


100


100
particle diameter d, [pm]


Figure 14 Comparison of manual and automatic de-
tected drop size distributions for toluene/water;
cp = 2%.

The last results presented in this work deal with a
highly concentrated liquid/liquid system of n-butdl
chloride/water with cp = 45%. Example images of such a
system are already given in Figure 6. These data were
gained in an industrial project presented in detail in [20] a
stirred tank.

225
n-butyl chloride/water
S200 400 rpm, pH 7; cp = 45 %
3
-0 175


E 150 -~ autom. O O

125 -r mn

S100-

S75


0 20 40 60
time t [min]

Figure 15 Comparison of manual and automatic de-
tected transient Sauter mean diameters for highly con-
centrated dispersions for two different stirrers.

The size of drops in agitated vessels is the result of two
opposing phenomena: drop breakage and coalescence.
When agitation starts, the bulk of the dispersed phase is
pulled down into the continuous phase and is broken into


d Qo(d,)
q o(d ) =
Sdd,


The broadness of the distribution is different and this
becomes obvious by comparing the number density
distribution of the automatic and the manual detection. A
few more bigger and many more smaller particles are
detected by the automatic software compared to a manual
observation. That comes parallel with more than a doubled
number of detected particles for the automatic detection






7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

ing and experimental analysis of PSD measure-
ments through FBRM. Part. Part. Sys. Char., 17
(2000) 167-179.
[6] M. Fricke, A. Voigt, and K. Sundmacher. Droplet
size distribution of technical Macroemulsions:
Variation of Process Parameters. Proc. Int. Cong.
Part. Tech. Nuremberg, 27.-29.03.2007, pp. 4.
[7] D. Petrak, Simultaneous measurement of particle
size and particle velocity by the spatial filtering
technique. Part. Part. Sys. Char., 19 (2002)
391-400.
[8] B. Sachweh, C. Heffels, R. Polke, and M. Riidle.
Light scattering sensor for in-line measurements
of mean particle sizes in suspensions. Proc. 7th
Europ. Symp. Part. Char. Nurembergl998, pp.
635-644.
[9] J. Ritter and M. Kraume, On-line measurement
technique for drop size distributions in liq-
uid/1iquid systems at high dispersed phase frac-
tions. Chem. Eng. Technol., 23 (2000) 579-582.
[10] B. Hu, P. Angeli, O.K. Matar, C.J. Lawrence, and
GF. Hewitt, Evaluation of drop size distribution
from chord length measurements. AIChE J., 52
(2006) 931-939.
[11] A. Giibler, M. Wegener, A.R. Paschedag, and M.
Kraume, The effect of pH on experimental and
simulation results of transient drop size distribu-
tions in stirred liquid-liquid dispersions. Chem.
Eng. Sci., 61 (2006) 3018-3024.
[12] S. Wollny, R. Sperling, G. Heun, J. Ritter, S.
MaaB, and M. Kraume. Bestimmung von
Tropfengrillenverteilungen zur Charakterisierung
von Riihrorganen hinsichtlich der Partikelbean-
spruchung. Proc. ProcessNet Partikelmesstechnik.
Miinchen, 10.-14.03.2008, pp. 2 (in German).
[13] B. Junker, Measurement of bubble and pellet size
distributions: past and current image analysis
technology. Bioprocess. Biosyst. Eng. 29 (2006)
185-206.
[14] G. Crawley and A. Malcolmson, Online particle
sizing as a route to process optimization. Chem.
Eng. (New York) 111 (21 r 14) 37-41.
[15] F.B. Alban, S. Sajjadi, and M. Yianneskis, Dy-
namic tracking of fast liquid-liquid dispersion
processes with a real-time in-situ optical tech-
nique. Chem. Eng. Res. Des., 82 (21li14,
1054-1060.
[16] I. Scherze, R. Knofel, and G. Muschiolik, Auto-
mated image analysis as a control tool for multi-
ple emulsions. 19 (2005) 617-624.
[17] L.M.R. Br~s, E.F. Gomes, M.M.M. Ribeiro, and
M.M.L. Guimariies, Drop Distribution Determi-
nation in a Liquid-Liquid Dispersion by Image
Processing. Int. J. Chem. Eng., (2009 ) pp. 6.
[18] M. Cohen and GT. Toussaint, On the detection of
structures in noisy pictures. Pattern Recognit., 9
(1977) 95-98.
[19] S. Hirschberg, R. Koubek, F. Moser, and J.
Schock, An improvement of the Sulzer SMX
(TM) static mixer significantly reducing the pres-
sure drop. 87 (2009) 524-532.
[20] S. MaaB, F. Metz, T. Rehm, and M. Kraume, Pre-
diction of drop sizes for PVC-production in slim


small drops. The rate of drop break-up in this stage is much
larger than the rate of drop coalescence. As a result the size
of drops shows an exponential decrease during the transient
stage. The rate of drop coalescence rises with the increasing
number of drops with time during the transient stage.
Eventually a steady state is reached where the rates of drop
break-up and coalescence are balanced. In Figure 15 the
results for transient Sauter mean diameter are shown, for
both the manual and the automatic counting of drops for the
n-butdl chloride/water system with cp = 45%. It shows
clearly that, despite of numerous overlapping of drops in the
images: obviously occurring in such highly concentrated
dispersion compare Figure 6: the automatic algorithm
works very well. It gains almost the same values for the
transient Sauter mean diameter than the manual observation
for the two different stirrer types. The steady state values
after 60 min of mixing have a deviation of 2.5 percent
(d32,man = 104.2 pLm; d32,autom = 101.9 pLm) for the smaller
stirrer and 5.5 percent for the larger one.


Conclusions
Image analysis has become a powerful tool for the
work with particulate systems, occurring in chemical en-
gineering. A major challenge is still the manual work load
which comes with such applications. Additionally manual
quantification also generates bias by different observers,
which is shown in this study. Therefore the full automation
of those systems is necessary. In this work we have intro-
duced a MAILAB" based image recognition tool to count
and measure particles in multi phase systems fully auto-
matic. Up to now this software is limited to spherical parti-
cles but will be extended to irregular shapes as well.
The program has reached hit rates of 95% with an error
quotient under 1% and a detection of 250 drops per minute
for simple images and hit rates of 40% with an error quo-
tient of under 5% and a detection of 10 drops per minute
for some difficult images. A parallelization on different
cores in a multi-core PC will increase those calculation
times dramatically. Therewith online process observation
and control will be possible.


References

[1] K. Leschonski, Particle Characterization, Present
State and possible Future Trends. Part. Char., 3
(1986) 99-103.
[2] D. Greaves, J. Boxall, J. Mulligan, A. Montesi, J.
Creek, E.D. Sloan, and C.A. Koh, Measuring the
particle size of a known distribution using the fo-
cused beam reflectance measurement technique.
Chem. Eng. Sci., 63 (2008) 5410-5419.
[3] S. MaaB, S. Wollny, A. Voigt, and M. Kraume,
Experimental comparison of measurement tech-
nique for drop size distributions in liquid/1iquid
dispersions. accepted manuscript: Exp. Fluids
(March 2010), (2010) pp. 15.
[4] A.W. Pacek, I.P.T. Moore, A.W. Nienow, and R.V.
Calabrese, Video Technique for Measuring Dy-
namics of Liquid-Liquid Dispersion during Phase
Inversion. AIChE J., 40 (1994) 1940-1949.
[5] A. Ruf, J. Worlitschek, and M. Mazzotti, Model-






7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010

reactors part I: Single stage stirrers. submitted to
Chem. Eng. J., (2010).




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - Version 2.9.7 - mvs