Group Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Title: 15.2.2 - Two-phase PIV/PTV measurement of bubbly flow across pin fins in a micro-channel
ALL VOLUMES CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00102023/00370
 Material Information
Title: 15.2.2 - Two-phase PIV/PTV measurement of bubbly flow across pin fins in a micro-channel Micro and Nano-Scale Multiphase Flows
Series Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Physical Description: Conference Papers
Creator: Honkanen, M.
Jung, J.
Kuo, C.-J.
Peles, Y.
Amitay, M.
Publisher: International Conference on Multiphase Flow (ICMF)
Publication Date: June 4, 2010
 Subjects
Subject: microscopic particle image velocimetry
fluorescent imaging
bubbly flow
cylinder wake
image analysis
 Notes
Abstract: This paper presents microscopic Particle Image Velocimetry (μPIV) measurements of bubbly flow across two pin fins in tandem in a micro-channel. μPIV-system and fluorescent imaging technique (Akhmetbekov et al. 2010) are utilized to measure fluid velocity field and gas bubbles simultaneously in the wake of the pin fins. Automatic image analysis procedures are presented to recognize the outlines of breaking and coalescing bubbles and to discriminate tracer particles from background objects and bubbles. Irregularly-shaped bubbles are recognized even when they partly overlap with other bubbles. A digital mask (Gui et al. 2003) is generated on top of bubbles and an instantaneous fluid velocity field is computed with a PIV cross-correlation analysis. Bubble size, shape and velocity are measured utilizing a Particle Tracking Velocimetry (PTV) approach. Statistics of bubble dynamics (x, d, u, ReB) and fluid flow quantities with and without bubbles quantify how dispersed bubbles affect the fluid flow across the micro-pins.
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: VID00370
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: 1522-Honkanen-ICMF2010.pdf

Full Text


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



Two-phase PIVIPTV measurement of bubbly flow across pin fins in a micro-channel


M. Honkanen *, J. Jungt, C.-J. Kuot, Y. Pelest, M. Amitayt


Department of Energy and Process Engineering, Tampere University of Technology:
Korkeakoulunkatu 6, FIN-33720 Tampere, Finland
Markus.HonkaneniiAtut.fi
SDepartment of Mechanical, Aerospace and Nuclear Engineering, Rensselaer Polytechnic Institute, Troy, USA


Keywords: Microscopic Particle Image Velocimetry, fluorescent imaging, bubbly flow: cylinder wake, image analysis




Abstract

This paper presents microscopic Particle Image Velocimetry (gLPIV) measurements of bubbly flow across two pin fins in
tandem in a micro-channel. gLPIV-system and fluorescent imaging technique (Akhmetbekov et al. 2010) are utilized to measure
fluid velocity field and gas bubbles simultaneously in the wake of the pin fins. Automatic image analysis procedures are
presented to recognize the outlines of breaking and coalescing bubbles and to discriminate tracer particles from background
objects and bubbles. Irregularly-shaped bubbles are recognized even when they partly overlap with other bubbles. A digital
mask (Gui et al. 2003) is generated on top of bubbles and an instantaneous fluid velocity field is computed with a PIV
cross-correlation analysis. Bubble size, shape and velocity are measured utilizing a Particle Tracking Velocimetry (PTV)


approach. Statistics of bubble dynamics (x, d, u, ReB) and
dispersed bubbles affect the fluid flow across the micro-pins.


Introduction

Microscale gas/1iquid flows are present in liquid cooling
systems of electronics. Flow boiling regimes in
microchannels differ greatly from the ones in macro-scale
(e.g., Cheng et al. 2009, Harirchian and Garimella, 2009). In
our recent study on flow boiling across the pin fins, it was
found that the bubble departure frequency and departure
diameter strongly depend on the position of the nucleation
site (Krishnamurthy and Peles, 2010). Bubbles growing
from cavities on the frontal pin fin surface have a much
higher frequency and much smaller diameter than bubbles
emerging from the rear region. We believe that the different
frequencies are closely related to different naturally
amplified modes associated with the flow across the fins. In
this study, fluid flow across pin fins is investigated in
single-phase flow and in an isothermal bubbly flow to
quantify the effect of dispersed bubbles on the fluid flow.
Experimental methods are necessary to extend current
knowledge at the micro scale. Bubbly flows in
microchannels are typically investigated with digital
camera-based techniques. Singh et al. (2009) utilized
digital compact camera attached to a microscope to study
flow boiling regimes and vapor void fractions by means of
image analysis. A digital high-speed camera has been
utilized to study bubble dynamics in detail (e.g. Harirchian
and Garimella 2009; Vega et al. 2009; Fu et al. 2009). On
the other hand, fluid velocity field close to the bubbles is
obtained with Microscopic Particle Image Velocimetry
(gLPIV) technique (e.g., King et al. 2007; Yamaguchi et al.
2009). In bubbly flows, PIV analysis can be extended to the


fluid flow quantities with and without bubbles quantify how


bubble phase by fluorescent imaging technique
(Akhmetbekov et al. 2010: Dulin et al. 2010). In this
technique, fluorescent dye is added to the fluid to visualize
gas-liquid interfaces clearly. Also fluorescent tracer particles
are added to visualize fluid motion. This enables
simultaneous measurement of fluid velocity fields and
bubble properties (size, shape, and velocity) with a single
camera and without utilization of back-light illumination.
This study presents an application of fluorescence imaging
technique to gLPIV measurements to simultaneously measure
the dynamics of gas and liquid phases across the micro pin
fins in a microchannel. New image analysis method is
presented for the recognition of irregularly-shaped bubbles
that partly overlap with each other in the image. Our
application focuses on the wake region of the micro pin fins
and the interactions of dispersed bubbles and fluid flow are
studied.

Experimental Setup

The experimental setup is shown in Figure 1, which consists
of three primary subsystems: the flow section, the
experimental instrumentation, and a data acquisition system.
The flow section delivered both the working fluid (water)
and the nitrogen gas bubbles into the microchannel. The gas
and liquid flow rates and pressures were monitored. The
water in the reservoir was first degassed and then
pressurized by nitrogen gas to ~600 kPa. The
microchannel included two micro pin fins in tandem, 100
pLm apart with a diameter of 50 pLm. The pins were located
in the middle of a 20,000-pLm long, 1,500-pLm wide, and






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


250-pLm deep microchannel. A 50-pLm circular orifice for
gas injection was located at the bottom of the channel 5,000
pLm downstream the channel inlet and 6,200 pmn upstream
from the micro pins. A Pyrex cover sealed the device from
the top and allowed the image-based measurements and
flow visualization. Bubbly flow was visualized through a
microscope (Zeiss AXIO Observer) with a 10x
magnification utilizing a dual-frame Image Pro X 2M
camera. An Nd:YAG double-cavity laser with maximum
frequency of 15 Hz was used as the light source for the
dual-frame camera. To minimize the error and noise
generated by vibration and leveling effects, the experiments
were conducted on an optical table with Melles Griot active
anti-vibration system.


Figure 1: Experimental setup.


Figure 2: a) A fluorescence image of bubbles and tracer
particles in the microchannel close to the two micro-pins. b)
A Mie scattering image of bubbles and tracer particles in a
laser light sheet illumination.

Tracers attached to the top and bottom walls of the channel
appear as blurry spots in the image. Fluorescent dye
produces bright background intensity in the image. Gas
bubbles are darker than the image background and they
have bright outlines due to the increased dye concentration
on gas-liquid interface. Rhodamin dye concentration in
water is over 1000 ppm in order to effectively contaminate
the bubbles. The distance from the bubble injection nozzle
to the field of view is only 6 mm and the fluid velocity is up
to 8 m/s. Dye contaminates the bubble surface in about 1 ms
before entering the field of view. For that reason, high
concentration of Rhodamin dye is necessary to obtain fully
contaminated bubbles, whose fluorescence is clearly visible
in the experimental images. Contamination clearly affects
bubble dynamics, but bubble dynamics are already affected
by the presence of flow tracer particles in the fluid.
Why the bubble surface needs to be contaminated with the
fluorescent dye? Figure 2b shows a Mie scattering image of
bubbles and tracer particles in a laser light sheet
illumination. Laser light is reflected from the bubble surface
overexposing the camera sensor. On the other hand, the
bubble outlines are partly invisible. The detection of
irregularly-shaped bubbles is very difficult in the Mie
scattering image (Fig. 2b), whereas the bubbles in Fig. 2a
can be automatically detected utilizmng the algorithms
presented next in this paper.

Image Analysis

Experimental images are analyzed with automatic image
analysis algorithms to measure the fluid flow field and the
size, shape and velocity of gas bubbles. Firstly a
time-average image of the measurement set is computed.
The time-average image presents the stagnant objects such


A commercial microscopic Particle Image Velocimetry
(gLPIV) system (LaVision) is utilized to measure the fluid
flow field and the size, shape and velocity of gas bubbles.
Fluorescent imaging technique (Akhmetbekov et al. 2010) is
applied to measure both fluid velocity field and gas bubbles
simultaneously. The fluid flow is seeded with fluorescent
polyamide tracer particles with a diameter of 0.9 pLm that
visualize the fluid motion. The outlines of gas bubbles are
visualized adding fluorescent dye (Rhodamin B) into the
fluid. Volume illumination is provided by an Nd-YAG laser
and the reflections of laser light are blocked with a
high-pass optical filter. Hence, camera visualizes only the
fluorescent light of tracers and dye solution. Focal plane of
the microscope is set at the middle plane of the
microchannel.
Figure 2a shows an experimental gLPIV-image of a bubbly
flow around micro pins in a microchannel. Fluorescent
tracer particles produce hundreds of bright spots in the
image. Generally in PIV, two tracer particle images with a
short time delay are acquired and an iterative
cross-correlation analysis is computed between the two
images to provide an instantaneous fluid velocity map. In
the bubbly flow case, experimental images contain also
other objects that must be discriminated from flow tracers to
obtain correct results.
Flow is from top to bottom in Fig. 2a. Two micro-pins are
located in the upper middle part of the image. Micro-pins
produce dark and blurry shadows in the image, whereas gas
bubbles have dark inner parts and bright outer rings. Gas
bubbles breakup when they collide with the micro pins.
Daughter bubbles often coalesce downstream of the pins
producing bubbles with the same size range as initially in
the inlet


D*Ln Inlulrruullr)rr~n


Measurement Technique






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

image background.
The outer part of a bubble surface is clearly brighter
than the background.
Bubble outlines have the same brightness as the
image background.
Bubbles are larger than the tracer particles.


as micro-pins, channel walls and the tracers attached to the
channel walls. The unequal lighting of two-cavity Nd-YAG
laser is taken into account computing a time-average image
for both frames of PIV image pairs. Figure 3 shows the
computed time-average image from a sequence of 100
images. Some artefacts of bubble outlines remain in the
time-average image. Background objects (pins and walls)
are recognized from the time-average image with a global
grayscale threshold. Pixels with a grayscale below the
threshold value (i.e. ~200) belong to the background
objects.



















Figure 3: Time-average image of an image set.

Bubble Outline Detection

A heuristic algorithm for automatic bubble outline detection
is developed as follows. Let's look at an example image of
bubbles in Figure 4. Two bubbles can be seen downstream
the micro-pins, and one smaller bubble is located between
the micro-pins. The grayscale profile across the bubble
image reveals the appearance of a bubble in the image.
When the bubble outline is approached horizontally from
left, the image grayscale value first slightly decreases due to
shadowing effect of the bubble. Then, image grayscale
increases above the background level at the bubble outline
and continues to increase reaching its maximum at the point
of first-order reflection on bubble surface. Then, the
grayscale value rapidly decreases, and the values in the
middle part of the bubble image are clearly below the
background grayscale level. The location of bubble outline
does not correspond to the location of maximum grayscale
value or the location of maximum grayscale gradient.
Bubble outlines cannot clearly be seen in the fluorescence
images, but their location can be estimated as the point
where the grayscale value first time exceeds the background
grayscale level. The background grayscale level can vary
temporally and locally due to the presence of bubbles:
Shadowing effect of bubbles locally decreases the
brightness of liquid fluorescence. On the other hand, the
multiple reflections of fluorescent light on bubble surfaces
increase the brightness of liquid fluorescent especially
between closely located bubbles. The varying background
grayscale level must be noticed in the analysis.
In summary, bubbles can be detected from microscopic
fluorescence images based on their general features:
The inner part of a bubble is always darker than the


010 d.lb 10.10 dOi I.m oo~ o.ll
*u~u~


0.ii 020


Figure 4: Experimental image with a grayscale profile.

The difference of the experimental image and the
time-average image (i.e. background image) is shown in
Figure 5. The positive values are shown in red color and the
negative values in blue color. White color represents the
background grayscale level in Fig. 5. Bubbles are clearly
visible, but bubble outlines are very difficult to distinguish.
The sharp regions of the bubble image possess strong
grayscale variation and we may expect that the contrast
between the bubble outline and the image background is
higher in those regions than in the blurred, out-of-focus
regions. Based on this assumption, the bubble outlines are
recognized with a locally adaptive grayscale segmentation
method presented by Niblack (1986). Niblack proposed that
the grayscale threshold value equals the local (15x15 pixels)
grayscale standard deviation value multiplied by a constant
value 0.12. He as well carried out image background
removal prior to the image segmentation.


Figure 5: Background-subtracted image.






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

template matching technique cannot be applied to
irregularly-shaped bubbles, because the bubble templates
cannot be determined. Vega et al. (2009) presented methods
to locate gas-liquid interfaces from shadowgraphy images
with sub-pixel accuracy. Back-lighted images provide clear
silhouettes of bubbles enabling accurate outline detection.
However, in many experimental cases, such as in our case,
the back-light illumination cannot be arranged. In that case,
the planar fluorescence imaging technique is the best way to
visualize the outlines of irregularly-shaped bubbles.

Detection of Overlapping Bubble Images

Bubbles mostly break when they collide with the micro pin
fins. The wake of the pin fins causes mixing that brings
closely-located daughter bubbles together. Collisions often
lead to coalescence, but in some cases bubbles can overlap
in the image without coalescence taking place. Erroneous
results are obtained, if the overlapping bubbles are detected
as one large bubble. Therefore, the bubbles that belong to a
group of overlapping bubbles need to be recognized
individually. Different types of overlapping object
recognition algorithms have been developed (Meyer 1994:
Pla 1996: Shen et al. 2000; Berg et al. 2002; Honkanen et al.
2005; Brdder and Sommerfeld 2007). Most of these
techniques rely on the accurate detection of bubble outlines.
The recognition of bubbles that overlap in the images is
typically based on the following features:
- Sudden change in the outline direction
- Change in the outline focus level
- Difference between the region outline and region
convex line
Change in the local thickness of the region (Watershed
segmentation)
In the case of planar fluorescence images, the overlapping
bubbles can be recognized based on:
- The bright rings of 1st order reflection
- The dark middle regions of bubbles
We can assume that the inner parts of the bubble regions are
not merged, although the bubble images overlap in the
image. This statement is acceptable, because the dark
middle regions are always circled by the bright rings that
separate them from other bubbles. The middle regions work
as seeds that are enlarged to generate the full segments of
bubbles. Figure 7 shows the segmentation result of the
group of three bubbles. Original image is shown in Fig. 7a.
Fig. 7b shows the segmentation result, where the inner
segments are black and outer segments gray. The outer
segments without inner segment are discarded. The outer
segments that have multiple inner segments are processed
with a Watershed segmentation procedure where each outer
segment pixel is labeled to the closest inner segment. If a
pixel is located at equal distance from several inner
segments, the pixel will be set to zero. This means that the
overlapping parts can belong to only one bubble causing a
small underestimation in measured size of overlapping
bubbles. Fig. 7c shows the final bubble segments.
Overlapping bubbles are labeled with different gray levels.
Bubble outlines are shown in Fig. 7d. The result is correct,
but the accuracy of bubble outline detection is questionable.
Is there room for improvements? Or is back-light
illumination the only way to obtain sub-pixel accuracy in
sizing of irregularly-shaped bubbles?


ragure: O. IVIviauxK seginentlaIO on nlas Uncteu Lt
and the detected bubble outlines are highlighted.


The result of the Niblack segmentation can be seen in Fig. 6
where the detected bubble outlines are highlighted with red
lines. Niblack segmentation is not applied directly to a
background-subtracted image, but first some image
enhancements are required. E.g. the bubble between the
micro-pins in Fig. 4 does not possess enclosed bright outer
ring, but it has two bright arcs on both sides. The caps
between the arcs correspond to the shadows of the
micro-pins and they need to be considered to recognize the
bubble as a whole.
1) A two-dimensional median filter with a 3x3 kernel is
utilized to discriminate tracers from bubbles based on their
size difference.
2) It is noticed that the image background grayscale level
slightly varies temporally due to the changes in dye
concentration. The temporal changes are compensated
leveling the background image to match the mean grayscale
level of the instantaneous image prior to background
subtraction.
3) The negative grayscales in the background- subtracted
image (corresponding to the inner parts of bubbles) are
segmented with a constant threshold value. This threshold
value should be less than the contrast between bubble
images and background and larger than the contrast of
random variations of the background grayscale level. The
detection of dark middle parts is further eased by
highlighting the regions of low local standard deviation. It is
noticed that the liquid phase regions have higher grayscale
variance due to flow tracers than the inner regions of
bubbles.
4) The obtained inner segments are enlarged with a 5x5
sliding maximum filter to close the caps between the dark
inner parts and bright outer parts of bubbles.
5) Niblack segmentation is applied to the positive part of the
background-subtracted image (Fig. 5) and the provided
outer ring segments and the inner part segments are
superimposed to obtain the complete bubble segments. The
remaining holes in the bubble segments are filled.
Sub-pixel accuracy in bubble outline detection clearly
cannot be reached in the case of planar fluorescence images
of distorted, irregularly-shaped bubbles. The appearance of
spherical gas bubbles in planar fluorescence images is
studied in detail by Akhmetbekov et al. (2010) and Dulin et
al. (2010) providing high accuracy in bubble sizing and
positioning by means of template matching technique. The






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

close to the gas-liquid interface of moving bubbles. It can be
noted that the high-resolution fluid velocity fields can only
be obtained utilizing masked direct correlation method in
the neighborhood of digital masks.
The resulted velocity vector field for the example image (in
Fig. 4) is shown in Figure 8. The mean flow velocity of 6
m/s is subtracted to reveal the details of flow structures in
the wake of the pins and close to the bubbles. Colormap on
left shows the velocity magnitude. Vortical flow structures
are clearly seen in the wake of the micro pins. Bubbles in
the wake elongate along the flow due to hydrodynamic
forces.

Computation of Bubble's Relative Velocity

Computation of bubble velocity is typically based on
consecutive bubble positioning or cross-correlation analysis
(e.g. Uemura et al. 1989; Honkanen and Saarenrinne 2002;
Theunissen et al. 2004; Akhmetbekov et al. 2010) where the
double-frame bubble images are cross-correlated and the
location of the correlation peak is sought. The accuracy of
bubble velocity measurement is the same as the accuracy of
locating the correlation peak, whereas in the case of bubble
positioning methods, the uncertainty in particle positioning
is multiplied by,[ito obtain the uncertainty of the velocity
measurement. However, the accuracy of correlation analysis
attenuates due to bubble deformation and due to the
presence of other objects inside the interrogation window, as
reported by Honkanen (2l1is14. In this study, the bubble
velocity is measured from the displacement of the bubble's
centre point, which corresponds to the centre of the pixels
belonging to the bubble image.
After processing the double-frame images, detecting the
bubbles and computing the fluid velocity field, the results of
both fluid phases are combined. The relative velocity of a
bubble is determined as the bubble velocity subtracted by
the undisturbed fluid velocity. In order to get an estimate of
the undisturbed fluid velocity at the centre point of a bubble,
the surrounding fluid velocity field is linearly interpolated to
the whole field of view and the data point closest to the
bubble centre is taken. The estimate corresponds to the
initial local average fluid velocity surrounding the bubble in
the measurement plane. It is noted that all the turbulence
quantities are measured only in the areas outside the digital
mask, so their accuracy does not depend on the interpolation
scheme.
Figure 9a shows an interpolated fluid velocity field in a
bubbly flow across the pin fins. Fluid velocity field is
presented relative to the mean velocity. On the background
of the vector field, one can see the detected bubbles, their
outlines in two image frames, bubble velocity with red
arrow and fluid velocity with white arrow. Figures 9b and
9d show instantaneous fluid velocity fields and the
corresponding bubble detections and bubble velocities are
shown in Figs. 9c and 9e. The upper bubble in Fig. 9c has
only one velocity vector, because its velocity equals the
fluid velocity with an uncertainty of one image pixel.
In general, bubbles move slower than the surrounding fluid
and the bubble's relative velocity is smaller than the bubble
velocity. The relative Reynolds numbers of bubbles range
between 100 and 500 in the case of cylinder flow Reynolds
number of 300.


c) d)
Figure 7: a) Overlapping bubbles. b) Bubble segments,
where the inner segments are black and outer segments gray.
c) Final bubble segments. d) Result image with detected
bubble outlines shown with red lines.










TM*
Figure~~~~~ ~ ~ ~~ 8 ntnaeu li eoi fld

Computation~ ~ ~ ~ ofFudVlct il

Commercial~~~~~~~~~~ PI otae(ai .)i tlzdt opt







Fgr8:Isatnosfluid velocity field. fo h ital-skd



pre-processed tracer particle images. Tracer particle image
is obtained subtracting the time-average image from the
original image (in Fig. 4) and adding a digital mask on top
of the regions of detected bubbles and background objects.
The digital mask prevents PIV cross-correlation analysis on
top of masked regions (Gui et al. 2003). A multipass
cross-correlation analysis with a final window size of 32x32
pixels (50% overlap) allows the investigation of fluid flow






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

A single phase flow over a low aspect ratio (H/D-l.4)
cylinder was investigated for its practical importance such
as micro pin fin channel and micro heat sink. From previous
studies (Norberg 1994; Chen et al. 1995; Zovatto and
Pedrizzetti 2001), low aspect ratio of cylinder as well as the
presence of an end plate plays an important role in flow
regime and transition mechanism. As shown in Fig. 10,
critical Reynolds number was delayed from Re=45-50 to
Re--400 due to the low aspect ratio and suppression by walls.
The onset of vortex shedding was observed and it is initiated
from the rear stagnation point which contributed the
increased turbulence intensity at Re=400 as shown in Fig.
11. The results for bubbly flow case with two pin fins and
cylinder aspect ratio of 5 are presented next.













Figure 10: Instant snapshot of a single phase flow field
over a single cylinder a) at Re=200 and b) Re=400.


111
U
41
010
Ud L1

011 IU
U
~" -i
,,




,, I ,,


d)- i - U e)
Figure 9: a) Instantaneous interpolated fluid velocity field.
b) and d) Instantaneous fluid velocity fields. c) and e)
Detected bubbles and their velocities in the corresponding
imaes. The red v cor iu thje ente of ra bubl shows uthe
velocity that is linearly interpolated from the surrounding
flow field.

Results

This paper demonstrates the performance of the presented
Two-phase PIV/PTV measurement technique with a bubbly
flow case with the flow Reynolds number of 300. Two
single phase flow cases are reported for comparison.

Single Phase Flow across Micro Pin Fins

Viscous flow over a cylinder has been extensively studied
and well documented for various Reynolds numbers.
Previous studies either by experimental or numerical
showed that stable flow transits to unstable state at
Re=45-50. As Reynolds number increases, a flow becomes
unstable due to the development of absolute instability at the
trailing edge of recirculation zone while convective
instability grows in the shear layer. As a result, vortex
shedding is initiated from the end tip of recirculation zone
and this mechanism is defined as the laminar vortex
shedding at low Reynolds numbers. Readers can refer to
Williamson (1996) and Zdravkovich (1996) for more details
about cylinder wake behavior.


aJ ~I Ib)
Figure 11: Turbulence intensity field a) at Re=200 and b)
Re=400.

Bubbly Flow across Micro Pin Fins

Bubbles affect the fluid flow and cause additionally mixing
of the flow. This can be seen in Fig. 12 that shows the mean
fluid velocity field and the turbulence intensity field in
bubbly flow across two pin fins in tandem. The magnitude
of turbulence intensity in the wake of the pin fins has
doubled from the single phase flow cases and the wake
region (corresponding to the area of increased turbulence)
has become larger. Recirculation zones of these 50 pmn
micro pin fins are clearly smaller than those of the larger,
100 pLm cylinders.
Bubble dynamics are studied with respect to streamwise
location in the channel ranging from 0.3 mm upstream to
0.6 mm downstream from the lower pin fin, which
corresponds to the location of origin. Bubble detection
frequency as detected samples per image is shown in Fig.
13a. Bubble detection frequency halves at 0.3 mm
downstream from the pins. This reveals that bubbles
coalesce in the wake of the micro pin fins. The upstream
bubbles are often detected only when they have collided
with the upper pin and broken into two or more bubbles. Fig.






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


13b shows the mean bubble diameter +- the standard
deviation intervals of bubble diameter. Bubble size
decreases at the micro pins due to breakage, but soon
increases back to its original value at 0.2 mm downstream
the pins. However, the standard deviation of bubble
diameter increases at the pins and remains large.
The relative Reynolds number of bubbles depends on the
bubble diameter, bubble's relative velocity and liquid
viscosity. Figure 14a shows the mean relative Reynolds
number of bubbles as a function of streamwise position in
the case of flow Reynolds number of 300. The relative
velocity of bubbles is its highest at the pin fins, where
bubbles deform and break into multiple daughter bubbles.
The mean relative Reynolds numbers of about 350 are
obtained on top of the pin fins, whereas elsewhere the
values are below 150. This agrees well with the dynamics of
bubbles: bubbles break only at the micro pin fins and
elsewhere they are only deformed by the flow. Mean
streamwise velocity of bubbles and the +- rms-velocity
intervals are shown in Fig. 14b. It can be seen that the
fluctuations in bubble velocity occur only on top of the pin
fins and in the near wake of the fins, up to 0.3 mm
downstream the lower pin fin. Elsewhere the fluctuations
correspond to those of the carrier flow. Sudden increase of
bubble velocity on top of the pin fins can be explained by
bubble elongation and breakage that shifts the bubble centre
forward. After breakage, bubble velocity continuously
decreases in the near wake of the pin fins, drops below the
mean flow velocity and reaches the fluid flow velocity onlV
at 0.3 mm downstream the pin fins.


, A


a 17



E013

0 1
E01
-,0


-0 2 0 0 2


0 4 0


b) ~~channel x-locatlo nmm
Figure 13: a) Bubble detection frequency with respect to
streamwise location in the channel. b) Mean bubble
diameter and +- standard deviation intervals.


mean bubble Reynolds number


100'
-0.4


-0.2 0 0.2
channel x-location [mm]


0.4 0.6


mean +/- rms x-velocity of bubbles


-0.4 -0.2 0 0.2 0.4 0.6
b) channel x-location [mm]
Figure 14: a) Mean relative Reynolds number of bubbles
and b) mean streamwise velocity of bubbles with respect to
streamwise location in the channel.


Figure 12: a) Average velocity field and b) turbulence
intensity field in bubbly flow at Re= 300.


a as


3j 0 06


o 04
S0 03
0 02
0 0 01

_CO 4 -0 2 0 0 2 0 4 0
channel x-location [mm]
mean +/- rms diameter









Conclusions


This paper presented a two-phase PIV/PTV measurement
technique to study bubbly flow across pin fins in a
microchannel. The measurement technique with planar
fluorescence imaging was successfully applied to
irregularly-shaped bubbles that deformed, broke and
coalesced in the close vicinity of micro pin fins. A novel
image analysis method was presented to recognize
irregularly-shaped bubbles from fluorescence images even
though bubbles partly overlapped with each other in the
image.
The presented image analysis method detects the bubbles
based on the dark inner region and bright outer ring of
bubbles. It is computationally fast and easy to implement. It
separates the overlapping bubbles with a kind of Watershed
approach where each outer segment pixel is labeled to its
closest inner segment. Therefore, the performance of
overlap detection does not depend on the accuracy of bubble
outline positioning. Bubble outlines are located with a
locally adaptive Niblack segmentation, which is robust
against noise and interference of other objects such as tracer
particles close to the bubble outline. However, sub-pixel
accuracy in irregularly-shaped bubble outline positioning
cannot be achieved. Nevertheless, the measurement
technique is superior to Mie scattering imaging of bubbles
and it provides means to simultaneous measurement of the
fluid velocity field and the properties of dispersed bubbles
without utilization of back-light illumination. Future work is
directed to improve the accuracy of bubble velocity
measurement by means of a tailor-made cross-correlation
approach.

Acknowledgements

The work was supported by the Office of Naval Research
(Program Officer Mark Spector) and the Academy of
Finland. The microfabrication was performed in part at the
Comnell NanoScale Facility (a member of the National
Nanotechnology Infrastructure Network) which is supported
by the National Science Foundation under Grant
ECS-0335765, its users, Comnell University and industrial
affiliates.

References

Y. K. Akhmetbekov, S. V. Alekseenko, V M. Dulin, D. M.
Markovich, K. S. Penaunin. Planar fluorescence for round
bubble imaging and its application for the study of an
axisymmetric two-phase jet. Exp Fluids (2010), DOI:
10. 1007. silr l I US-IIII' s.17 97 -0

E.H. Berg van den, A.G Meesters, C.A. Kenter, W. Schlager.
Automated separation of touching grains in digital images
of thin sections. Computers&Geosci Vol. 28 pp.179-190
(2002).

D. Brdder, M. Sommerfeld. Planar shadow image
velocimetry for the analysis of the hydrodynamics in bubbly
flows. Meas Sci Technol Vol. 18 pp.2513-2528 (2007).

P. Cheng, G Wang, X. Quan. Recent Work on Boiling and
Condensation in Microchannels. J Heat Transfer Vol. 131:


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

Nr. 043211. Doi: 10.1115/1.3072906 (2009).

J. H. Chen, W. G Pritchard, S.J. Tavener, "Bifurcation for
flow past a cylinder between parallel planes," J. Fluid Mech.
Vol. 284, 23 (1995)

V.M. Dulin, D.M. Markovich, K.S. Penaunin. Round bubble
imaging technique based on planar fluorescence. Meas Sci
Technol (2010), personal communication.

X. Fu, P. Zhang, H. Hu, C.J. Huang, Y. Huang, R.Z. Wang.
3D visualization of two-phase flow in the micro-tube by a
simple but effective method. J Micromech Microeng Vol. 19,
085005 (2009).

L. Gui, S.T. Wereley, Y.H. Kim. Advances and applications
of the digital mask technique in PIV experiments. Meas Sci
Technol Vol. 14 pp. 1820-1828 (2003).

T. Harirchian, S.V. Garimella. Effects of channel dimension,
heat flux, and mass flux on flow boiling regimes in micro-
channels. Int. J. Multiphase Flow Vol. 35 pp. 349-362
(2009)

M. Honkanen, P. Saarenrinne. Turbulent bubbly flow
measurements in a mixing vessel with PIV. 11th Int. Symp
on Appl of Laser Tech to Fluid Mech. Lisbon, Portugal,
8.-11.7.2002, Paper 3.2.

M. Honkanen. PTV methods for overlapping bubble images.
Proc. of Gala-Konferenz: Lasermethoden in der
Strdmungsmesstechnik. Karlsruhe. Paper 39 (21 r'4).

M. Honkanen, P. Saarenrinne, T. Stoor, J. Niinimilki.
Recognition of highly overlapping ellipse-like bubble
images. Meas Sci Technol Vol. 16 pp. 1760-1770 (2005).

C. King, E. Walsch, R. Grimes. PIV measurements of flow
within plugs in a microchannel. Microfluid Nanofluid Vol. 3
pp. 463-472 (2007).

S. Krishnamurthy & Y. Peles. Flow Boiling on Micropin
Fins Entrenched Inside a Microchannel--Flow Pattemns and
Bubble Departure Diameter and Bubble Frequency. J Heat
Transfer Vol. 132, 041007 (2010).

F. Meyer. Topographic distance and watershed lines. Signal
Process. Vol. 38 pp. 113-125 (1994).

W. Niblack. An Introduction to Digital Image Processing, pp.
115-116, Prentice Hall (1986).

C. Norberg. An experimental investigation of the flow
around a circular cylinder : influence of aspect ratio. J. Fluid
Mech. Vol. 258, 287 (1994).

F. Pla. Recognition of partial circular shapes from
segmented contours. Computer Vis & Image Underst Vol.
63 pp. 334-343 (1996).

L. Shen, X. Song, M. Iguchi, F. Yamamoto. A method for
recognizing particles in overlapped particle images. Pattern
Recognit Lett Vol. 21 pp. 21-30 (2000).






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


S.G. Singh, A. Jain, A. Sridharan, S.P. Duttagupta, A.
Agrawal. Flow map and measurement of void fraction and
heat transfer coefficient using an image analysis technique
for flow boiling of water in a silicon microchannel. J.
Micromech. Microeng Vol. 19, 075004 (2009).

R. Theunissen, A. Stitou, M. L. Riethmuller. A Novel
Approach to improve the Accuracy of PTV Methods. 12th
Int Symp on Appl of Laser Tech to Fluid Mech. Lisbon.
Paper 27.1. (21 r'4).

T. Uemura, F. Yamamoto, K. Ohmi. A high speed algorithm
of image analysis for real time measurement of two-
dimensional velocity distribution. Flow Visualization ed B
Khalighi, M Braun, C Freitas (New York: ASME) pp.
129-133 (1989).

E.J. Vega, J.M Montanero, J. Fernandez. On the precision of
optical imaging to study free surface dynamics at high frame
rates. Exp Fluids Vol. 47 pp. 251-261 (2009).

C. H. K. Williamson. Vortex dynamics in the cylinder wake.
Annu. Rev. Fluid Mech. Vol. 28, 477 (1996).

M. M. Zdravkovich, Different modes of vortex shedding :
An overview. J. Fluids and Structures Vol. 10, 427 (1996).

L. Zovatto and G. Pedrizzetti, Flow\ about a circular
cylinder between parallel walls," J. Fluid Mech. 440, 1
(2001).

E. Yamaguchi, B.J. Smith, D.P. Gaver. Micro-PIV
measurements of the ensemble flow fields surrounding a
migrating semi-infinite bubble. Exp Fluids Vol. 47 pp.
309-320 (2009).




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