You are on page 1of 4

TURBULENT VELOCITY PROFILE MEASUREMENT IN CIRCULAR

PIPE FLOW USING PARTICLE IMAGE VELOCIMETRY TECHNIQUE


7.Takeuchi, "R. Miraghaie, bK.Yuki
' S . Satake, dT.Kunugi, "N. B. Morley

"Mechanicaland Aerospace Eng. Dept.. UCLA


420 WestwoodPlaza, Los Angeles, CA 90095-1597, USA
b
Department of Quantum Science and Engineering, Tohoku University
Aramaki-Aoba.Ada-ku Sendai 980-77, Japan
'Department of Applied Electronics, Tokyo Universiw of Science
2641 Yamazaki,Noda, Chiba, Japan
dDepartmentof Nuclear Engineering, Kyoto UBiversity,
Yoshidu, Sakyo, Kyoto 606-8501, Japan

Abstract: Experimental study of turbulent flow in circular pipe wall decreases considerably. This makes the investigation of
is carried out by using PIV technique as a preliminary the fluid mechanics and heat transfer of high Prandtl number
experiment for studying thermo fluid properties of molten salt, fluid compared to IOW Prandtl number fluid very important
Flibe in particular, which is a high Prandtl number fluid. At this from the design point of view in order to assess the
first stage, water is used as working fluid to obtain reference
data. Measurements of the fully developed turbulent flow in a
applicability of Flibe to high heat flux removal system as well
circular pipe at Re4286 are carried out using PIV system and as the prospect for heat transfer enhancement mechanisms.
near wall flow structure and turbulent statistics as well as
averaged flow field are obtained. Average velocity profile bas 11. EXPERIMENT
good agreement with the DNS data.
The experimental setup consists of liquid loop with circular
pipe test section and optical system for PIV measurement. The
I. INTRODUCTION
schematic view of the experimental apparatus is shown in
The use of molten salt coolants, in particular Fig. 1, Test section is comprised of the three pieces of acrylic
fluorine-lithium-beryllium or FZibe salts, have attractive pipe with inner diameter of 89-56", with a flow straightener
properties for liquid wall and liquid breeder blanket designs in section attached to its inlet. The PIV measurement section is
fusion reactors. The low electrical conductivity of such salts at 60D downstream from the inlet. This is long enough for
reduces the pressure drop associated with magneto- average flow to be fblly developed. Water jacket is attached to
hydrodynamic forces to a nearly negligible level when the test section so as to minimize the distortion of the image
compared to high conductivity liquid metals. But there are inside the pipe due to the curvature of the pipe. The flow rate
still many issues related to fluid mechanics and heat transfer is controlled by the balance between a main throttle valve and
in a blanket design using Flibe. In particular, the low thermal bypass valve. Cooling water line is installed in the loop tank
conductivity and high viscosity of Flibe make it a high F'randtl to compensate the temperature increase due to the pump to
number medium - meaning heat transport ffom a heated wall ensure the temperature to be constant.
into the core of the coolant flow is affected mainly by the
turbulent motion. Loop tank Water jacket Flow Direction
To understand the underlying science and phenomena of
fluid mechanics and heat transfer of Flibe, a series of
experiments as part of the US-Japan Jupiter-I1 collaboration Test section
are in progress. A fluid flow facility utilizing water and Filter
water-based electrolytes as simulants for generic high Prandtl
number, electrically conducting fluids like Flibe has been
constructed at UCLA. The approach involves flow field
measurement and heat transfer experiments using these Flibe
simulant fluids along with modeling and analysis of
fundamental phenomena of turbulence and heat transfer. Pump Flow meter
Experimental results obtained with the particle image Fig. 1 Experimental apparatus for PIV measurement
velocimetry (PIV) technique are acquired and compared to the
simulations to provide fundamental insights for the design of For PIV measurement, the flow is seeded with 20pm
Flibe blankets. polyamide particles with relative density of 1.02. The
Near wall flow structures are especially important because particles are small enough to follow the water flow and its
thermal: boundary layer is much thinner than the momentum relative density is close enough to that of the water to neglect
boundary layer. In the case where the thermal boundary layer the particle sedimentation due to the gravity force. The
lies within the viscous sub-layer, the heat transfer through the
0~7803-7908~X/031$17.00 0 2 0 0 3 IEEE 140
number density of the seeding particles is adjusted so that the 111. RESULTS
particle images have several particles in the 16x16 pixel
Fig.4 shows the vector map of the mean flow velocity
interrogation window. Fig.2 shows the optical configuration obtained by PIV system. The average is taken over 5000
of the PIV system. PIV laser sheet is supplied by Nd:YAG instantaneous flow velocity. The mean flow velocity for
laser with its energy of 50mJ per pulse and its repetition rate
turbulent flow is resolved well in the velocity map.
is 15Hz. The laser sheet passes vertically the half plane of the The vorticity contour is shown in Fig.5. Even though the
pipe after reflected by the mirror and illuminate the seeding
average velocity field shows smooth curve across the pipe, the
particles flowing with the water flow. The reflected image of
vorticity distribution has some fluctuation because the error in
the particles is captured by CCD camera from the side of the the velocity is amplified when the derivative is taken.
pipe.

j
1

't v
Nd:YAG laser Mirror Fig. 5 Vorticity contour of the mean flow field
Fig. 2 Optical configuration for PIV measurement
The mean velocity profile normalized by the centerline
Fig.3 shows the schematic diagram of the data acquisition velocity with respect to the distance from the centerline is
control. The rate of data acquisition is 3.7SHz and time shown in Fig.6. The experimental data is plotted along with
between two pulses on which the cross correlation is taken is the DNS (Direct Numerical Simulation) data by [2]Satake et
1.5ms. The firing of the laser and the timing of capturing the a1 (2000). Since experimental data agrees with DNS, PIV
image are synchronized by the system provided by Dantec. measurement resolves well the mean ffow velocity field of the
turbulent flow.
In Fig.7, mean velocity profile near the wall is converted
Trigger Pulse into the wall coordinate defined by u+=d UT and y+=yuJv to
, Is Laser Pulse 2" Laser Pulse obtain the universal velocity profile independent of Reynolds
number, where y is distance from the wall, U, is friction
velocity and v is kinematic viscosity of the liquid. The
velocity profile in viscous sublayer shifts upward to the linear
profile (1) obtained by theoretical analysis, where DNS data
shows good agreement with the theoretical analysis.

U+ = Y+ (1)
I 1 I 1

2 Particle Map a I Vector Map


Fig. 3: Schematic diagram oftiming control

The PIV ima es are analyzed with the interrogation system


f
described by [ 'Scarano and Riethmuller (1999) which is
provided by Dantec.

r/R
Distance from the centerline
normalized by the radius

Fig. 6 Axial mean velocity normalized by centerline velocity

The reason why the shift occurs is considered that the


curvature of the pipe skew the image in radial direction and
Fig. 4 Vector map ofthe mean flow velocity
the distance from the wall contains considerabIe error. Even
141
though water jacket is attached to the test section to minimize the image due to the pipe curvature has considerable effect as
the distortion, the effect of the difference in light path by the shown earlier. This results. in failure of capture the smalI
thickness of the pipe wall is not eliminated. Commonly motion near the wall.
accepted velocity profile has logarithmic distribution at the
distance larger than 30 in y-t-, which is shown by (2).
U, = 0.251ny+ + 5.5 (2)
The experimental result agrees well with the DNS
although it varies from commonly accepted experimental
correlation. This discrepancy between law of the wall and the
data is discussed and explained to be the difference between
plane channel flow and axisymmetric flow by ["Eggles et a1
(1994). o2.bc-,f
0' 0.2 0.4 0.6
I

0.8
, \
1
r/R
Distance from the centerline
normalized by radius

Fig. 9 Radial component of turbulent intensity

The Reynolds shear stress distribution is shown in Fig.10.


lot
I *y u+=2.51n(yi.f+5. Total shear stress of the fdly developed statistically steady
flow in the pipe must be linear and shown by the straight line
I , ,..I 1 . .,,
of gradient 1. This is because the gradient of the total shear
5 IO 50 IO0 stress must balance the pressure gradient along the axis. In the
Y+=YdV near wall region, the viscous stress plays dominant role and
Distance from the wall Reynolds stress goes to zero with r/R close to one. On the
in wall coordinate . I
other hand, the Reynolds stress plays dominant role and
viscous stress is small in the core region.
Fig. 7 Axial velocity profile scaled into wall coordinate

The root-mean-square values of the fluctuation component


of the velocities normalized by friction velocity, which are
called turbulent intensity, are shown in Fig.8 and Fig.9 and
compared to the DNS data.

r/R
Distance from the centerline
normalized by radius

Fig. I O Reynolds stress and total stress distribution


I , I . I I I .

0.2 0.4 0.6 0.8 1 The region where the Reynolds stress has peak value
rlR corresponds to the region where the turbulence is generated
Distance from the centerline ([61Tennekes& Lumley, 1972). The Reynolds stress obtained
normalized by radius
by PIV experiment shows smaller peak compared to the DNS
data. Since the PIV measure the average displacement of the
Fig. 8 Streamwise component o f turbulent intensity
particle in the small interrogation window, the fluctuation of
the velocity tends to be smoothed. The Reynolds stress could
PIV data of the streamwise component of turbulent appear to be small when the smoothing by the nature of the
intensity has good agreement with DNS data in the small r/R PIV is significant. To avoid this effect, interrogation window
region but higher peak around r/R=0.9. has to be smaller than the smallest scale of the turbulence;
The radial component of the turbulent intensity obtained by Kolmogorov Microscale. Since the dynamic range of the
PIV agrees with the DNS in the region far from the wall but velocity in PIV system is limited by the number of the CCD
has larger value in near wall region even though this must go elements, decreasing the size of the interrogation window
to zero at the wall by theory. In the near region, distortion of with higher magnification is accompanied by deterioration of
1 42
the highest measurable velocity. This makes the turbulent
measurement by PIV difficult.

IV. coLcLusIoN
PIV measurement has been carried out to study hlly
developed turbulent pipe flow at Reynolds number 5280
based on the bulk mean velocity as a preliminary research of
thermo fluid properties of molten salt, Flibe in particular. The
mean flow property and turbulent statistics obtained by PIV
technique are compared to the DNS (Direct Numerical
Simulation) by Satake et al.
The experimental data of mean flow property show good
agreement with numerical data except the region very close to
the wall. This implies that PIV technique is a suitable method
to measure the flow. In spite of this, there are still some issues
conceming the near wall resolution of the PIV system in
which the distortion of the image exists. The experimental
data of the Reynolds stress also agrees with DNS in core
region but has lower peak around r/R=0.8 even though the
velocity profile agrees well in this region. This implies that
the small scale of turbulence is not fully resolved in this
region and fluctuations are smoothed by cell-averaged
velocity. This must be fixed by employing the smaller
interrogation window with higher magnification so as to have
each interrogation window smaller than the smallest scale of
turbulence.
Although there are some issues concerning the application
of PIV for this experiment, PIV can be applied to the turbulent
pipe flow measurement. These issues will be fixed the future
work.

ACNOWLEGMENT
The authors wish to acknowledge support by the US DOE
Grant No.DE-FG03-86ER52123,and by the Japanese
Ministry of Education, Culture, Sports, Science and
Technology (MEXT) via the Jupiter-2 Collaboration.

REFERENCES
[l] Eggles, J.G.M. et al., “Fully developed turbulent pipe
flow: a comparison between direct numerical simulation
and experiment”, J. Fluid Mech. (1994) ~01.268,175-209
[2] Satake, S., Kunugi, T., Himeno, R., High Reynolds
Number Computation for Turbulent Heat Transfer in a
Pipe Flow, High Performance Computing: Third
Intemational Symposium, ISHPC 2000, Tokyo, Japan,
October 16-18,2000, Proceedings
[3] Scarano, E, Riethmulter, M.L., Iterative multigrid
approach in PIV image processing with discrete window
offset, Experiments in Fluid 26( 1999) 513-523
[4] Particle Image Velocimetry, M. Ragel, E. Willert, Jurgen
Kompenhans, 1998
[ 5 ] Handbook of Particle Image Velocimetry, Visualization
Society of Japan, 2002
[6J Tennekes, H., Lumley, J.L., A First Course in Turbulence,
MlT Press (1972)

143

You might also like