Group Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Title: 12.3.1 - PIV Measurement of Binary-size Group of Particles Laden in Turbulent Mixing Layer
ALL VOLUMES CITATION THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00102023/00303
 Material Information
Title: 12.3.1 - PIV Measurement of Binary-size Group of Particles Laden in Turbulent Mixing Layer Experimental Methods for Multiphase Flows
Series Title: 7th International Conference on Multiphase Flow - ICMF 2010 Proceedings
Physical Description: Conference Papers
Creator: Li, C.T.
Chang, K.C.
Wang, M.R.
Publisher: International Conference on Multiphase Flow (ICMF)
Publication Date: June 4, 2010
 Subjects
Subject: two-phase flow
PIV
mixing layer
 Notes
Abstract: A methodology combining the modes of particle image velocimetry (PIV) and particle tracking velocimetry (PTV), which are applied separately to the tracer particles (representing the carrier phase) and large particles (representing the dispersed phase), is studied for the velocity-field measurements of binary-size group of particles laden in turbulent mixing layer. A double-discriminating process, in terms of gray level and image-pattern area, together with the median mask technique is developed for the measurement of dispersed-phase velocity. The optimal setting of the parameters including the thresholds of gray level and image-pattern area as well as the size of median mask are determined with the aid of Taguchi method. There are two mechanisms affecting the particles’ dynamics, that is, turbulent dispersion and inter-particle collisions in the turbulent particle-laden flow. It is found that the valid sample number required to attain statistically stationary results of the dispersed phase is increased from 30, 200 to 230 #/cell for the cases with the mass loading ratios of 1, 3 and 5%, respectively, due to various weightings of these two mechanisms in each cases.
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: VID00303
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: 1231-Li-ICMF2010.pdf

Full Text

Paper No 7th International Conference on Multiphase Flow
ICMF 2010, Tampa, FL USA, May 30-June 4, 2010


PIV Measurement of Binary-size Group of Particles Laden in Turbulent Mixing Layer



C. T. Li*, K. C. Changt and M. R. Wang'

Department of Aeronautic and Astronautic, National Cheng Kung University, Tainan, 70101, Taiwan

tvpicalo(gmail.com, kcchang@imail.ncku.edu.tw and wangmr@imail.ncku.edu.tw

Keywords: Two-phase flow, PIV, mixing layer




Abstract

A methodology combining the modes of particle image velocimetry (PIV) and particle tracking velocimetry (PTV), which are
applied separately to the tracer particles (representing the carrier phase) and large particles (representing the dispersed phase),
is studied for the velocity-field measurements of binary-size group of particles laden in turbulent mixing layer. A
double-discriminating process, in terms of gray level and image-pattern area, together with the median mask technique is
developed for the measurement of dispersed-phase velocity. The optimal setting of the parameters including the thresholds of
gray level and image-pattern area as well as the size of median mask are determined with the aid of Taguchi method. There are
two mechanisms affecting the particles' dynamics, that is, turbulent dispersion and inter-particle collisions in the turbulent
particle-laden flow. It is found that the valid sample number required to attain statistically stationary results of the dispersed
phase is increased from 30, 200 to 230 #/cell for the cases with the mass loading ratios of 1, 3 and 5%, respectively, due to
various weightings of these two mechanisms in each cases.


Introduction

Nonintrusive, optical techniques such as phase Doppler
anemometry (PDA) or previously named Phase Doppler
particle analyzer (PDPA) are well-known velocity
measuring tools for two-phase flow field. These techniques
allow the collection of complete (mean and fluctuating)
and accurate velocity and particle size data. However,
these techniques and limited to a temporal, pointwise
measurement. In contrast, the planar light sheet techniques
such as particle image velocimetry (PIV) and Doppler
global velocimetry (DGV) (Komine 1990), which perform
simultaneous multi-point measurements, are extended to
the velocity measurements of two-phase flow fields in
recent years (Schodal et al. 2002, Tower et al. 1999,
Sakakibara et al. 1996, Hassan et al. 1992, Kiger & Pan
2000, Oakley et al. 1997, Rottenkobler et al. 2002, Dreier
et al. 2000). However, since DGV is limited to the
collection of time-averaged velocity data, only PIV can be
applied to the collection of complete (mean and fluctuating)
velocity data and is capable of providing the spatial
differential quantities of turbulence, such as the vorticity
and the dissipation rates of turbulent kinetic energy.
Application of PIV technique on the velocity measurement
of two-phase flow is still a state of the art so far. In
two-phase flow, the acquired image consists of the
carrier-phase (seeded tracers or small particles) patterns
and dispersed-phase (large particles) patterns. Both phases
of the flow have to be studied separately in order to
provide reliable information of local, instantaneous fluid
dynamics of carrier phase and of the properties of
individual dispersed particles. In terms of the difference in


size and/or surface properties, the optical signals from the
tracers and large dispersed particles exhibit different
characteristics.
The potential approaches for separating the carrier and
dispersed phases in a two-phase flow are summarized as
follows.
(1) Color: The tracers carry fluorescent dye that emits light
at a wavelength (color) different from that of the
illuminating light sheet. One may use either a filter
together with two synchronized monochrome cameras or a
color camera for taking images. The color techniques are
relatively expensive because they require powerful lasers
(Tower et al. 1999). There is also a high potential cost due
to the need of the large quantities of mono-dispersed
fluorescent particles.
(2) Image intensity (gray level): Discrimination of the
large (dispersed) particles from the small particles (tracers)
is made by setting a threshold of the image intensity (for
example, the largest tracer particles should produce a pixel
intensity of less than 70% of the maximum (Sakakibara et
al. 1996)) in the frame grabber so that any pixel intensities
less than the threshold value are taken as the images of
tracers. To meet this condition, the dispersed particles are
typically 50-100 times larger in size than the tracers so that
the scattered light intensity is of the order of 1000 times
greater for a dispersed particle (Sakakibara et al. 1996).
(3) Spot size (image pattern area): An image threshold with
subsequent discrimination by white spot area is used.
Hassan et al. (1992) pointed out the problem of having a
gray "corona" around bright bubbles and they solved it by
using local threshold 1/r2, where the r denotes the gray
level of brightness. However, this approach is suitable for






Paper No


the dispersed phase of bubbles.
(4) Spatial frequency: A median filter that eliminates
tracers by treating them as high frequency noise is used
(Kiger & Pan 2000).
(5) Spot shape: Complicated spatial filters are used to
either extract or eliminate the spots exhibiting certain
image patterns (Oakley et al. 1997).
(6) Correlation peak properties: Discrimination of the two
phases is made with the peak separation technique
(Rottenkolber 2002) based on a fact that the shape of a
correlation peak is characterized by its height and width.
This approach requires a condition with sufficient velocity
difference between the two phases. In addition, it is hard to
find the universal criteria on how to set the cutoff shape
(height and width) because the correlation peaks are
usually affected by many parameters.
To allow phase discrimination easily, fluorescent seeding
particles were usually used to trace the carrier phase. As
mentioned before, it requires either two synchronized
monochrome cameras or a color camera, which is
expensive from the point of view on cost, in use of the
color techniques. In addition, most of the dispersed phases
are of polydispersed size distribution in nature. In other
words, the multi-phase is composed of the dispersed-phase
elements (particles or droplets per se) with the same
material but different sizes. It usually took advantage by
treating the small dispersed elements with sufficiently
small Stokes numbers as the tracers in the two-phase flow
measurements. In this study, a PIV measuring methodology,
in use of one monochrome camera and same material for
the tracer and dispersed particles, is developed for the
velocity measurements of the particle-laden flows
consisted of binary particle sizes: small and large ones
representing the carrier and dispersed phases, respectively,
in the turbulent mixing layer.
It was reported (Dreier et al. 2000, Li et al. 2009) that
non-uniform seeding tracer distributions were observed
around the interfaces between the shear layer and two free
stream regions of the mixing layer and it led to a difficulty
in PIV measurements. A proper image post-processing
suggested in our previous study (Li et al. 2009) is used to
resolve this non-uniform tracer distribution problem in PIV
measurements.


Experimental Set-up


The experimental facility is a vertically downward,
rectangular, suction-type wind tunnel, schematically shown
in Fig.la, which is composed of the settling chamber,
contraction section, test section, and noise reduction
chamber. The test section of this tunnel with a
cross-sectional area of 150 mmx150 mm is divided into
two independent flow paths by a central splitting plate. A
perforated plate is placed in the upstream of the
honeycomb to generate the required pressure drop for the
low-speed flow path. The trailing edge of the central
splitting plate extends 150 mm into the test section.
Thickness of the central splitter plate at the trailing edge is
0.2 mm. A Cartesian coordinate is selected such that the
transverse coordinate y is positive toward the high-speed
stream and the streamwise coordinate x is positive


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

downward with the origin at the trailing edge of the
splitting plate. Nearly invariant spanwise distributions of
velocity components are observed in the interest of
measured sections. It assures that the two-dimensional
characteristics of the tested mixing-layer flow are
presented. All the results presented hereafter are made in
the downstream streamwise stations where the flows are of
stationary turbulence.
Two groups of Si02 particles are fed, through the particle
seeding devices shown in Fig.lb, into the flows to form
two-phase flows. The small one (ranging from 0.5 to 10
gm with 2.3 grm in mean size), which are capable of
following the carrier fluid motion faithfully, serves for
tracers to represent the carrier phase, while the larger one
(ranging from 100 to 300 ugm with 175 ugm in mean size),
which is remarkably dominated by its own inertia motion,
represents the dispersed phase. The tracers are fed in the
far upstream of both high- and low-speed streams, while


Dispersed phase
(a) ENDING













W " ----1-- TT-
SETTLING
CHAMBER


SUCTIONV CONTRACTION
SECTION


T TEST
'SECTION























i ( T 150 rm 7
) NOISE REDUCTION CHAMBER41




Particles
ou let
(b) a 1




100 mi (inner)
120mm (outer)



entrance 10


l50 rnu


Figure 1: Experimental facilities of (a) wind tunnel
and (b) particle seeding device.





Paper No


the larger particles are fed into the flows at two transverse
positions of y = 10 and 20 mm in the upstream of the
settling chamber of high-speed side (see Fig. la).
A two-dimensional double-pulse PIV system with
cross-correlation estimation, which was manufactured by
Integrated Design Tools (IDT) Inc., is used for the
instantaneous measurements of velocity field. The PIV
system consists of a pulsed diode laser, a high speed
camera with a 50 mm Nikon standard lens, a synchronizing
timing hub, and a personal computer for data acquisition.
The f-number (f#) and magnification factor (M) are set
equal to 1.2 and 8, respectively. The wavelength (1#) of the
diode solid state laser (Model XS-IR-10) is 795 nm, and
the maximum pulse power is 10W. The thickness of laser
sheet is around 1 mm. According to Nyquist theorem
(Bendat & Piersol 2000), the frame rate must be at least
twice as fast as the frequencies of small turbulent eddies
such as that of Taylor-scale eddy to avoid the problem of
aliasing statistics. The frame rate of CMOS camera (Model
X-stream XS-4) for capturing digital images is 3140 pairs/s
(two frames/s) with full spatial resolution of 500x380
pixel2. The minimum spatial resolution of camera is, thus,
estimated to be 0.12 mm/pixel. It shows that the minimum
temporal and spatial resolutions are 3.18x10-4 sec and 0.12
mm, respectively. The particle image diameter is in terms
of the diffraction limited minimum image diameter (df) and
the particle size (dp) as follows.


d,= (FMd)2 +d (1)

where the diffraction limited minimum diameter is defined
as (Adrian 1991):
d =2.44f# (M+l)A (2)


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

Reynolds number (Ree) are used to identify the flow
conditions of the mixing layer and are defined, respectively,
by

R H UL (4)
UH + U,


Re,


UH80


where UH and UL are the free stream velocities at the high-
and low-speed sides, respectively, v is the kinematic
viscosity of air at the room temperature, and 0o is the
momentum thickness closer to the trailing edge of the
splitting plate (i.e. x = 3 mm) defined by


00 (UH -U)(U -UL)
-- (UH -U,)2


For the study of two-phase flows, two more parameters
including the mass loading ratio (a) and the volumetric
fraction of the dispersed phase (Q) are specified. Three
mass loading ratios of 0.01, 0.03, and 0.05 are studied for
investigating the interaction levels between the two phases.
The volumetric fractions (Q) to be investigated are ranged
from 5.39x10-4 to 2.7x10-3. Details of the initial conditions
are summarized in Table 1. The velocity ratio between
high- and low-speed velocities is maintained around R =
0.67 in this study.


The minimum length of Taylor micro scale (/Tx), estii
using the following formulas, is around 2 mm.


, -< 1

x >
/ --


where the symbol < > denotes the ensemble averaging. The
highest frequency of Taylor-scale eddy is estimated to be
around 500 Hz (= u1/ ). As comparing to the minimum
rTx
temporal resolution of 3.18x 104 sec for the employed PIV
system, it, thus, assures the accuracy of turbulent data
obtained in the present experiments. Illumination is made
with a pulse separation period of 70 uts. Our previous study
on the single-phase PIV measurement in the turbulent
mixing layer (Li et al. 2009) revealed that the use of 2200
frames (in other words, 1100 vectors) were able to assure
the stationary results of turbulence statistics. However, at
least 20000 frames are used in the present two-phase study,
and more discussion on this issue will be elaborated later.
Reliabilities of PIV measurements for the single-phase
conditions (Li et al. 2009) were found to be 1.4 % and 6%
for mean and root-mean-squared (r.m.s.) fluctuating
velocities, respectively.
Two flow parameters, that is, the velocity ratio R and


Phase Discrimination Process
mated
Particle-imaging techniques which are applied to the
measurements of fluid velocity can be generally casted in
three modes in terms of the density of tracer particles
(3) images, including particle-tracking, particle-image, and
laser-speckle modes (Adrian 1991, Raffel et al. 2007). In
the case of low image density, the image of individual
particles can be detected


Table 1: Summary of the initial flow conditions (at x = 3
mm) with R = 0.67 at different mass loading ratios
S() 0 (single 1 3 5
phase)

n 9.09x107 5.46x10-6 1.45x105 2.28x 105

n(#/c.c.) 8.04x104 1.47 4.42 7.37

UH (m/s) 8.7 8.6 8.5 8.0

u'HO (m/s) 0.08 0.28 0.345 0.368

u'LO(m/s) 0.042 0.050 0.041 0.049

o0(mm) 1.11 1.31 1.26 0.62
There exist the seeding particles for PIV
measurement even in the single-phase case.






Paper No


0 50 100 150 200 250
r (pixel)
Figure 2: Probability density distributions of image gray
level each obtained with the (a) carrier phase, (b) dispersed
phase and (c) two-phase condition for the flow under the
condition of a = 3 %, in a frame covering x = 120 170
mm.


figure 3: (a) Ungmal image, (b) Identilication o1 the
image patterns with the gray levels over 120 and with the
image-pattern areas over 10 pixel2, (c) the image after
deleting the candidates of the dispersed phase, and (d)
image of the dispersed phase after applying a 7x7 median
mask for the flow under the condition of a = 1%, in a
frame covering x = 120 170 mm.


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


and the image corresponding to the same particle
originating from different illuminations can be feasibly
identified. As a result, evaluation of the velocity to each
particle can be implemented by the tracking method. This
situation is referred to as particle tracking velocimetry
(PTV). In the case of high image density, it is impossible
to detect individual images as they overlap in most cases.
The random phase differences between the images of
individual randomly located particles formed the
interference patterns known as laser speckle. This situation
is called laser speckle velocimetry (LSV). The
particle-image mode occurs when the image densities lie
between those of PTV and LSV. The image densities are
now high enough to guarantee that every interrogation spot
in the image field contains many images, but not so high
enough to form speckles. This situation is referred to
particle image velocimetry.
The mean number densities of the small and large particles,
evaluated in the section of high-speed stream at the trailing
edge, are 8.04x104 and 4.42 particles/c.c., respectively,
which are equivalent to 4.63x10 and 6.36x10-
particles/interrogation window, respectively. Here each
interrogation window is constructed with 24x24 pixel2.
Accordingly, the carrier and dispersed phases are of
medium and low image densities, respectively. The PIV
and PTV techniques will be, thus, applied for the
evaluations of the velocities of the carrier and dispersed
phases, respectively, in the study.
To investigate the distribution ranges of gray levels (full
range in 0 255) for the carrier and dispersed phases, three
measurements of gray level distributions for the flows each
being with introduction of the carrier phase merely, the
dispersed phase merely, and both the carrier and dispersed
phases are preliminarily implemented; and their results are
presented in Figs. 2a, 2b and 2c, respectively. It indicates
that the gray levels of the carrier phase (the case with small
particles only) are distributed below the brightness of 120,
while the gray levels of the dispersed phase (the case with
large particles only) are distributed above the brightness of
50. Clearly, there exists an overlapping range in the gray
levels between 50 and 120 in these two cases. Further, the
probability density distribution for the case with both the
carrier and dispersed phases, i.e., Fig. 2c, is not simply a
superposition of the cases with the carrier phase merely
(Fig. 2a) and with the dispersed phase merely (Fig. 2b).
This is because, in consideration of the planar laser sheet is
associated with finite thickness, the out-focus image
patterns of the large particles and the cumulative
brightness from the overlapping small particles along the
same reflected ray of laser light may yield ill image
patterns. Thus, it is insufficient to fulfill image separation
based on the image intensity merely. It is known that the
median mask of smoothing filter, which replaces the value
of a pixel by the median of all the values in the
neighborhood for each work, can effectively eliminate the
ill image patterns. A double phase-discriminating process,
in terms of gray level and image pattern area (or spot shot),
together with the median mask technique, is developed.
The phase-discriminating process is elaborated with an
example case (a = 1 %) as follows. Original image is
presented in Fig. 3a. Two cut-off values (thresholds) in the
gray level and area of image pattern have to be set at the






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


beginning. In view of the maximum brightness of the gray
level distribution of the carrier phase (Fig. 2a), the
threshold of the gray level for the dispersed phase is firstly
set equal to 120. In other words, for those image patterns
whose brightness of the gray level are greater than the
threshold set here (i.e., 120), they are considered as the
candidates of the dispersed phase (shown in Fig. 3b). To
double-identify the candidates of the dispersed phase, a
second criterion in terms of the image-pattern area is
integrated into the phase-discriminating process. Note that
the carrier phase is represented by the small particles
whose sizes are smaller than 10 nm. The size of image
patterns for the particles with 10 gm in size is evaluated to
be around 3 pixels through Eqs.(1) and (2). Therefore, the
cut-off area of the image patterns fir the dispersed phase is
set equal to 10 pixel2.
After deleting the candidates of the dispersed phase from
the original image, this formed image (Fig. 3c) will be
used to evaluate the carrier-phase velocity by PIV method
developed in our previous study (Li et al. 2009). Since the
image pattern shown in Fig. 3b may still consist of some ill
image patterns of the dispersed phase, the median mask of
smoothing filter is next applied to eliminate the ill image
patterns of the dispersed phase. The minimum size of the
dispersed phase is 100 and its corresponding image object
diameter (de) is estimated to be 6.6 pixel through Eqs. (1)
and (2). A median mask of 7x7 pixel2 is, thus, chosen as an
example to eliminate the ill image patterns in Fig.3b, and
its result is shown in Fig.3d. Only two image patterns are
recognized as the dispersed phase in the frame by using the
presently proposed double-discriminating process together
with the median mask technique. However, since the
analysis of the dispersed phase is made with the PTV mode,
few number of image patterns in a frame such as shown in
Fig.3d is acceptable but it, instead, requires a great number
of frames to achieve the statistically stationary results for
the turbulent case, which will be elaborated in the
following.


Results and Discussion

In the proposed image pattern recognition process, there
are three prerequisite parameters including two thresholds
of the gray level and image-pattern area and one
neighborhood range of the median mask. As mentioned in
the preceding section, the maximum area of the image
patterns for the carrier phase (dp < 10gm) was estimated to
be around 10 pixel2. It is reasonable to set the threshold of
image-pattern area for the dispersed phase equal to 10
pixel2. However, there exists an overlapping zone between
the gray-level distributions of the carrier and dispersed
phases (see Fig. 2c). A more considerable amount of the
dispersed-phase candidates would be removed while
adopting a higher threshold of gray level as being
demonstrated in the preceding section. Similar situation
happens to the choice of the neighborhood of median mask.
In consideration of the present particle sizes denoted as the
carrier (dp < 10 gLm) and dispersed (- 175 gLm) phases, the
neighborhood of the median mask should fall into the
range between 3x3 and 11x11. Use of broader
neighborhood of the median mask would remove a more


102


3x3 pixel2
5x5 pixel2
7x7 pixel2
9x9 pixel2


10o I I I, A I , I
80 120 160 200 240
r* (pixel)
Figure 4: Comparison of collected samples per frame
versus the image brightness with four different median
mask under the condition of a = 3% in a frame covering x
= 80 130 mm.

considerable amount of the dispersed-phase candidates.
Nevertheless, achievement of the meaningful (stationary)
statistics for the turbulence requires a certain amount of the
valid samples to be analyzed. Less number of valid
samples of the dispersed phase in a frame requires more
number of frames in the analysis of the turbulent properties
of the dispersed phase. A sensitivity study on the threshold
of gray level, the neighborhood of the median mask, and
the minimum number of the valid samples for achieving
the meaningful statistics of the turbulent dispersed phase is
performed for the particle-loading flow with a = 3%.
Domain of each frame covers the length of 50 mm and
width 65.8 mm. Each frame is divided into 60x60 cells for
the evaluation of the dispersed-phase velocity distribution.
In the PIV mode, the velocity vector positioned in a
specified cell is calculated by implementing the
cross-correlation algorithm within an interrogation window
of 24x24 pixel2. In contrast, the velocity vector positioned
in a specified cell is calculated by those particles passing
through the cell (_ 8x8 pixel2) in the PTV mode. In other
words, the interrogation window used for the PTV mode is
the same as the area of the cell which is smaller than what
was used for the PIV mode. This is because the correlation
coefficients in the PTV mode are preserved at unity value
while those in the PIV mode are always below unity.
In this study, four different thresholds of gray level (80,
120, 160, and 200) and four different neighborhoods of
median mask (3x3, 5x5, 7x7, and 9x9 plxel\) are
investigated. Figure 4 shows the variations of the total
amount of valid samples for the dispersed phase (n),
collected in a frame covering x =80 mm to 130 mm, versus
the image brightness with four different neighborhoods of
median mask. Here nt is a mean value calculated using 100
frames for each test condition. It is clearly observed that
the total number of the valid samples in a frame reduces
with the use of either higher threshold of gray level or
broader neighborhood of the median mask.
In order to determine the proper set of the cell size used for
PTV measurement, four different cell sizes (16x 16, 12x 12,
8x8 and 5x5 pixel2) are tested for the evaluation of the
distributions of mean and r.m.s. fluctuating velocities for
streamwise and transverse components in the streamwise
station of x = 120 mm; and their results are shown in Fig.5.


Paper No






Paper No


10 0 10 20 30
y(mm)


16x 16 pixel2
12x 12 pixel2
8x8 pixel2
5x5 pixel


l0 -0 0 20 30
r y (mm)


SAZA
A AA


CKj 0.5. ka
10 0 10 20 30 10 0 10 20 30
y (mm) y (mm)
Figure 5: Statistics of the U, V, u'.m and v'.m of the
dispersed phase obtained with four various cell sizes in
PTV measurements by using the brightness threshold of
120 gray level and 5x5 median mask under the condition
of a = 5% at x = 120 mm.


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

their results are shown in Fig. 6. It shows that the use of
20,000 frames is sufficient to provide stationary results of
PTV measurements. However, the baseline is set by
choosing the PTV measurement with 100,000 frames.
Fluctuations of the dispersed-phase motion are in principle
stemmed from two mechanisms of turbulent dispersion and
inter-particle collisions. To investigate the required
collected sample number in a cell for assuring statistically
meaningful results of the dispersed-phase velocities, the
ensemble data of the mean and r.m.s. fluctuating
streamwise velocities calculated with various collected
sample numbers under the three a values of 1%, 3% and
5% in the streamwise station of x = 80 mm are made and
presented in Figs.7-9, respectively. Three positions, which
are separately located in the high-speed (y = 20 mm) and
low-speed (y = -10, -12 and -13 mm for the cases of a = 1,
3 and 5%, respectively) free stream regions and in the
shear layer (y = 0 mm), in this station are investigated for
each case of the tested mass loading ratio. Two
observations are made from the results shown in Figs. 7-9.
Firstly, the minimum collected sample numbers (Ns) to
assure the statistically stationary results for the mean
velocity of the dispersed phase are less than those for the
r.m.s. fluctuating velocity of the dispersed phase which is a


A 2Ok frames
o 40k frmes
A 60k frames
0 80k frames
S 100k frames


0 *'^ o.f


0 10 0 20 30 -10 0 10 20 30
y (mm) y (mm)
Figure 6: Comparison of the transverse profiles of the (a)
mean streamwsie, (b) mean transverse, (c) r.m.s.fluctuating
streamwise, and (d) r.m.s. fluctuating transverse
dispersed-phase velocity components velocity calculated
with different numbers of frames (using the brightness
threshold of 120 gray level and 5x5 median mask) under
the condition of a = 5% at x = 120 mm.

To obtain stationary statistics of the PTV measured data
easily, the cell size should be sufficiently large, whereas
the use of too large cell size could smear the fluctuating
characteristics of turbulent quantities. In consideration of
these two concerns and the results appeared in Fig. 5, the
cell size of 12x12 pixel2 are selected for the following PTV
measurements. To investigate how many valid samples are
required to attain the stationary results of the
dispersed-phase velocity, five different numbers of frames
(20,000, 40,000, 60,000, 80,000, 100,000) are used for the
evaluation of the distributions of mean and r.m.s.
fluctuating velocities for streamwise and transverse
components in the streamwise station of x = 120 mm; and


(a) (b)
9. y 20mm y=0mm
8


Ns =20



Ns=22


Ns =20



Ns=30


y = -10 mm


Ns =20



Ns=30


N N N
Figure 7: Statistical results of Up and (up') versus
number of samples at the positions within the (a)
high-speed free stream region, (b) shear layer and (c)
low-speed free stream region under the condition of the a
= 1% in the streamwise section of x = 80 mm.


y = 20mm



Ns= 100



Ns= 180


y=0mm



Ns= 120



Ns = 200


y = -12mm


Ns= 100



Ns= 150


0 100 150 200 250 50 100 150 2o00 250 0 100 150 200 250
N N N
Figure 8: Statistical results of Up and (up') versus
number of samples at the positions within the (a)
high-speed free stream region, (b) shear layer and (c)
low-speed free stream region under the condition of the a
= 3% in the streamwise section of x = 80 mm.


- k






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


y = 20 mm


Ns=100


Ns= 190


(b)
S y = 0nun


Ns= 100


Ns= 220


y = -13 mm


S Ns=120



Ns= 230


2 50 100 150 200 20 50 5 20 2 100 150 200 250
N N N
Figure 9: Statistical results of U and u'.s versus number
of samples at the positions within the (a) high-speed free
stream region, (b) shear layer and (c) low-speed free
stream region under the condition of the a = 5% in the
streamwise section of x = 80 mm.

high order correlation than the mean velocity for each case
of the investigated a values. Figure 10 presents the relation
of the maximum of Ns in each case (i.e. the suggested Ns
value which is used for the PTV measurement in the case
of a specified a) versus the mass loading ratio (a). It
reveals a trend that Nsmax asymptotes to a constant level
along with the increased a value.The optimal setting of the
parameters used in PTV measurement is searched with the
aid of Taguchi method (Taguchi 1993). Four parameters
including the threshold of gray level, size of median mask,
interrogation


250 r


100


111111111111


pixel
Figure 11: Dependence of the image diameter with the
threshold setting of gray level.

window offset, and the treatment of the image sharpening
are chosen for study. Table 2 summaries the set of
parameters in this study by using the L9 orthogonal array.
Arrangement of L9 orthogonal array means that nine
experiments are capable of identifying the performances of
four parameters simultaneously. For instance, the
performance of the parametric setting of Al is obtained by
averaging the results of No.1 through No.3. Similarly, the
performances of A2 and A3 come out with averaging the
results of Nos. 4-6 and Nos. 7-9, respectively. Here the
performance of the parametric setting is justified by means
of the signal-to-noise-ratio (SNR).
The image diameter of a particle is dependent on the set
threshold of gray level as illustrated in Fig. 11. It reveals
that there exists a coupling effect between the brightness
(gray level r) and the image object diameter (de). To further
examine the effects of the parametric setting on the
accuracy of the evaluation of dispersed-phase velocity, a
series of comparisons of the transverse distributions of U,
V, u'.s and v'.s at the station of x = 120 mm are presented


, ,, ,, I I ,,, i I
0 1 2 3 4 5 6
a(%)
Figure 10: Relationship of the required collected samples
and the mass loading ratios.

Table 2: Experimental design through L9 orthogonal array


(multi-objective) under the condition of a
mm


5 % atx = 120


L9 Arrangement
A: r B:medianmask C:off-set D:sharpening
1 1(80) 1(3x3) 1(0) l(no)
2 1 2(5x5) 2(2) 2(Sobel)
3 1 3(7x7) 3(4) 3(Prewitt)
4 2(120) 1 2 3
5 2 2 3 1
6 2 3 1 2
7 3(160) 1 3 2
8 3 2 1 3
9 3 3 2 1


A1


-O o 0 O 20 30
2 (c) (mm)



6
*'"W *


o r 80
A r 120
0 r=160
S baseline


o 10 0 10 20 30
2 (d) Y (mm)



!,C-i


01 o 0 4 20 3
10 0 10 20 30 10 0 10 20 30
y (mm) y (mm)
Figure 12: Comparison of the transverse profiles of the (a)
mean streamwsie, (b) mean transverse, (c) r.m.s.fluctuating
streamwise, and (d) r.m.s. fluctuating transverse
dispersed-phase velocity components velocity calculated
with different brightness threshold under the condition of a
= 5% at x = 120 mm.


Paper No


i~-----~ ~---






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


(b)
o 3x3 pixel2
A 5x5 pixel2
0 7x7 pixel2
S baseline


4 -10 ( 10 20 30 10 10 20 30
2r (c) y(mm) 2r (d) y(mm)


0
/O (B3&^2


0tiB~


10 0 10 20 30 -10 0 10 20 30
y (mm) y (mm)
Figure 13: Comparison of the transverse profiles of the (a)
mean streamwsie, (b) mean transverse, (c) r.m.s.fluctuating
streamwise, and (d) r.m.s. fluctuating transverse
dispersed-phase velocity components velocity calculated
with different neighborhood ranges of median mask under
the condition of a = 5% at x = 120 mm.


p


(b)
o Original
1.5 A Sobel
O Prewitt
x baseline



0.5
0 10 0 10 20 30
2 (d) y(Imm)



0


10 0 10 20 30 10 0 10 20 30
y (mm) y (mm)
Figure 15: Comparison of the transverse profiles of the (a)
mean streamwsie, (b) mean transverse, (c) r.m.s.fluctuating
streamwise, and (d) r.m.s. fluctuating transverse
dispersed-phase velocity components velocity calculated
with different image sharpening filter under the condition
of a = 5% at x = 120 mm.


0 10 0 10 20 30 10 1 0 0 20 30
y (mm) y (mm)
Figure 14: Comparison of the transverse profiles of the (a)
mean streamwsie, (b) mean transverse, (c) r.m.s.fluctuating
streamwise, and (d) r.m.s. fluctuating transverse
dispersed-phase velocity components velocity calculated
with different interrogation window offset under the
condition of a = 5% at x = 120 mm.

in Figs. 12-15, which are evaluated on the basis of 20,000
frames. Comparisons are made by using the baseline case
in terms of the evaluation with 100,000 frames. For
verifying the performance of the optimization process, the
distributions of the SNR according to the multi-objective
optimization of Taguchi method are presented in Fig. 16.
The proper parametric setting in terms of multi-objective
optimization (U, V, u'.s and v'.s) is determined as {Al,
B1, C3, D1}. Thus, the setting with r* = 80, 3x3 median
mask, 4 pixels window offset and without performing any
image sharpening treatments will be used in the two-phase
flow study.


22
SNR
19

16

13

10


Al A2 A3 B1 B2 B3 C1 C2 C3 D1 D2 D3
Parametric setting

Figure 16: Distributions of the SNR obtained by
multi-object optimization of the Taguchi method.


6 (b)


o0 0 10 2 30'
(C) y (mm)


3 optimal setting
x baseline


0 o 0 10 20 30
(d) y mm)


S10 0 10 2 30 10 10 20 30
y (mm) y (mm)
Figure 17: Performance of the optimal setting in terms of
the multi-objective optimaziation for transverse velocity
profiles of (a) mean streamwsie, (b) mean transverse, (c)
r.m.s.fluctuating streamwise, and (d) r.m.s. fluctuating
transverse dispersed-phase velocity components velocity
under the condition of a = 5% at x = 120 mm.


Paper No


P9


2- (b)
0 Ax 0 ixel
L5 A Ax pixel
0 x-4 ie




-1001020to30
2 (d) Y Omo)


"C


11


49~~"4~
X


1.5
-
81






Paper No


(c) y (mm)


(b)
0 optimal setting
x baseline






(d) y (mm)


-10 0 10 20 10 10 20
y (mm) y (mm)
Figure 18: Performance of the optimal setting in terms of
the multi-objective optimaziation for transverse velocity
profiles of (a) mean streamwsie, (b) mean transverse, (c)
r.m.s.fluctuating streamwise, and (d) r.m.s. fluctuating
transverse dispersed-phase velocity components velocity
under the condition of a = 3% and x = 120 mm.


~


Acknowledgements


0 optimal-10000 frames
o optimal-20000 frames
baseline 100000 frames


a 10 20. 30
(d) y (mm)
100000 rmes

20000 frames
1 10000 frames
LtV*WhWrA A- -


-10 0 10 20 30 -10 0 10 20 30
y (mm) y (mm)
Figure 19: Comaprison of the optimal setting in terms of
the 10,000 and 20,000 frames for transverse velocity
profiles of (a) mean streamwsie, (b) mean transverse, (c)
r.m.s.fluctuating streamwise, and (d) r.m.s. fluctuating
transverse dispersed-phase velocity components velocity
under the condition of a = 5% at x = 120 mm.

To double-check the above finding for the optimal setting
of PTV measurement, the results of U, V, u'rm and v'.s,
evaluated with the use of 20,000 frames, in the station of x
= 120 mm for two cases of a = 5% and 3% are presented
in Figs. 17 and 18. It shows that the satisfactory PTV
results can be indeed done with the suggested optimal
setting. Figure 19 further compares the results obtained
from 10,000 frames only with those from 20,000 and
100,000 (baseline) frames. It also shows that the
satisfactory PTV results can be obtained but the
measureable range becomes narrow for the results done by
using 10,000 frames due to the restriction of the Nsmax.


This work was sponsored by the National Science Council
of the Republic of China under Contract NSC
98-2221-E006-132.

References

Adrian, R.J., Particle-imaging techniques for
experimental fluid mechanics. Annual Review of Fluid
Mechanics, 23(1): p. 261-304 (1991)

Bendat, J.S. & Piersol, A.G, Random Data: Analysis and
Measurement Processing. Wiley-Interscience Publication,
John Wiley and Sons, New York (2000)

Dreier, T, Blaser, S., Kinzelbach, W., Virant, M., &
Maas, H. G, Simultaneous measurement of particle
and fluid velocities in a plane mixing layer with
dispersion. Experiments in Fluids, 29(5): p. 486-493
(2000)

Hassan, Y, Seeley Jr., Canaan RE, & Tk, B.,
Simultaneous velocity measurements of both components
of a two-phase flow using particle image velocimetry. Int.
J. Multiphase Flow, 18: p. 371-395 (1992)

Kiger, K.T., & Pan,C., PIV technique for the simultaneous
measurement of dilute two-phase flows. Journal of Fluids
Engineering, 122: p. 811 (2000)

Komine,H., System for measuring velocity field of fluid
flow ultizing a laser-Doppler spectral image converter, US
patent 4919536 (1990)

Oakley, T., Loth, E., & Adrian, R., A two-phase cinematic
PIV method for bubbly flows. Journal of Fluids
Engineering, 119: p. 707 (1997)


(a)


2




o 10 2. o 30
(c) y (mm)




~ L~p~aB~P^


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

Conclusions

Application of PIV technique is extended to the velocity
measurements of binary-size group of particles laden in
turbulent mixing layer. In consideration of the particle
number densities of the tracers and large particles in the
interrogation windows of the flow field, PTV and PIV
modes are separately applied to the dispersed and carrier
phases, respectively. A double-discriminating process, in
terms of gray level and image-pattern area, together with
the median mask technique is developed for the velocity
measurement of the dispersed phase. A parametric setting
of the thresholds of image brightness and image-pattern
area being 80 gray level and 10 pixel2, respectively, and
the size of median mask being 3x3 is suggested for the
measurement of the dispersed-phase velocity. There are
two mechanisms affecting the particles' dynamics, that is,
turbulent dispersion and inter-particle collisions in the
turbulent particle-laden flow. In consideration of these two
mechanisms, it is found that the thresholds of the valid
sample number (n*) are 30, 200 and 230 #/cell are required
for the PTV measurements of the particle-laden flows with
a = 1, 3 and 5%, respectively, in this study.






Paper No


Raffel, M., Willert, C. E., Wereley, S. T, & Kompenhans,
J., Particle Image Velocimetry: A Practical Guide. Second
Edition. Springer Berlin (2007)

Rottenkolber, G, Gindele, J., Raposo, J., Dullenkopf,
K., Hentschel, W., Wittig, S., Spicher, U., &
Merzkirch, W., Spray analysis of a gasoline direct
injector by means of two-phase PIV Experiments in
Fluids, 32(6): p. 710-721 (2002)


Sakakibara, J., Wicker, R.B., & Eaton, J.K.,
Measurements of the particle-fluid velocity correlation and
the extra dissipation in a round jet. International Journal of
Multiphase Flow, 22(5): p. 863-881 (1996)


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

Schodl, R., Rohle, I., Willert, C., Fischer, M., Heinze, J.,
Laible, C., & Schilling, T., Doppler global velocimetry for
analysis of combustor flows. Aerospace Science and
Technology, 6(7): p. 481-493 (2002)

Taguchi, G, & Yokoyama,Y, Taguchi Methods: Design of
Experiments. ASI Press. (1993)

Towers, D.P.,Towers, C. E., Buckberry, C. H., & Reeves,
M., A colour PIV system employing fluorescent particles
for two-phase flow measurements. Measurement Science
and Technology, 10(9): p. 824-30 (1999)

Li, C., Chang, K., & Wang, M., PIV measurements of
turbulent flow in planar mixing layer. Experimental
Thermal and Fluid Science, 33(3): p. 527-537 (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