Nuclear Engineering and Design 239 (2009) 888898

Contents lists available at ScienceDirect

Nuclear Engineering and Design

journal homepage: www.elsevier.com/locate/nucengdes

Non-intrusive experimental investigation of ow behavior inside a 5 5 rod

bundle with spacer grids using PIV and MIR
Elvis E. Dominguez-Ontiveros, Yassin A. Hassan
Department of Nuclear Engineering, Texas A&M University, College Station, TX 77843, USA

a r t i c l e i n f o a b s t r a c t

Article history: The validity of the simulation results from computational uid dynamics (CFD) is still under scrutiny.
Received 30 November 2008 Some existing CFD closure models for complex ow produce results that are generally recognized as
Received in revised form 25 January 2009 being inaccurate. Development of improved models for complex ow simulation requires an improved
Accepted 26 January 2009
understanding of the detailed ow structure evolution with dynamic interaction of the ow multi-scales.
Thus, the goal of this work is to contribute to a better understanding of presupposed and existent events
that could affect the safety of nuclear power plants. The fundamental phenomena of uid ow in rod
bundles with spacer grids can be elucidated by using state-of-the-art measurement techniques. This
study aims to develop an experimental data base with high spatial and temporal resolution of uid ow
velocity inside a 5 5 rod bundles with spacer grids. The full-eld detailed data base is intended to validate
CFD codes at various temporal-spatial scales. Measurements are carried out using dynamic particle image
velocimetry (DPIV) technique inside an optically transparent rod bundle utilizing the matching index of
refraction (MIR) approach. This work presents full eld velocity vectors and turbulence statistics for a rod
bundle under single phase ow conditions.
2009 Elsevier B.V. All rights reserved.

1. Introduction
Fluid ow around circular cylinders is one of the classical prob-
lems of uid mechanics and has been well studied because of its
common occurrence in many forms and in different applications.
Cylinder-like structures can be found both alone and in groups in
the designs for heat exchangers, cooling systems for nuclear power
plants, offshore structures, power lines, struts, grids, screens, and
cables, in both single and multi-phase ows. A complete under-
standing of the uid dynamics for the ow around a circular
cylinder includes such fundamental subjects as the boundary layer,
separation, the free shear layer, the wake, and the dynamics of vor-
tices. The ow eld of multiple-cylinder congurations involves
complex interactions between the shear layers, vortices and Kar-
man vortex streets (Zdravkovich,1987). The problem is further
complicated by the large number of congurations encountered
in practice, resulting in different ow patterns, and by the effect of
their interactions.
One of the applications of paramount importance in this study is
uid ow in fuel rod bundles of water nuclear reactors. In this type
of nuclear reactor, optimum heat removal from the surface of fuel
elements is the subject of many studies for researchers in order to
determine reactor thermal margin and safety. In these reactors, the
Corresponding author.
E-mail address: Hassan@ne.tamu.edu (Y.A. Hassan).
spacer grids which support the fuel assembly are used as an effec-
tive mixing device by attaching various types of ow deectors.
Several recent works are focused on the development of numerical
simulations that predict the complex behavior of uid ow close to
grid spacers and between fuel assemblies.
The validity of the produced results from computational uid
dynamics (CFD) is still under scrutiny for several applications in
real practical cases. Moreover, the existing models for multiphase
ows produce results that are generally recognized as unreliable.
Development of better models for multiphase simulation requires
an improved understanding of the ow evolution with dynamic
interaction of the ow multi-scales. Therefore, experimental data
is urgently needed for validation of the CFD models.
In order to establish reliable design and performance criteria
for tube bundle models, better velocity data is needed. Simonin
and Barcouda (1988) conducted experiments using laser Doppler
anemometry (LDA) within a specic tube arrangement. Although
they gathered some velocity data in cross-ow over a tube bundle,
the data points were limited and detailed velocity distributions, or
whole ow eld data was not available due to the nature of the LDA
technique. Chang et al. (2008) used 2D-LDA measurements in a 5 5
rod bundle array scaled to be 2.6 times larger than the actual bundle
size. This work focused on the performance and mixing character-
istics of two kinds of spacers with turbulence enhancement vanes.
The used spacer grids were of a typical split and swirl type for pres-
surized water reactors (PWR). The experiments were performed at a
condition of Reynolds number of Re = 48,000 and pressure of 1.5 bar.

0029-5493/$ see front matter 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.nucengdes.2009.01.009
 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898 889

Fig. 1. Schematic diagram of the experimental hydraulic loop and test section.

They reported the velocity at various axial positions ranging from
one hydraulic diameter to 16 hydraulic diameters. The reported
accuracy of the velocity measurements is about 1.6%. However, the
use of laser Doppler velocimetry (LDV) methods was restricted by
the existence of invisible regions in fuel rod bundles and insuf-
cient spatial resolution for the narrow gaps in rod bundles. Ikeda
and Hoshi (2006) developed a miniaturized LDV system embed-
ded in a fuel cladding. The rod-embedded ber LDV can be inserted
in an arbitrary grid cell instead of a fuel rod without disturbing
the ow. They obtained ow velocity data in a 5 5 rod bundle
typical of a PWR fuel assembly. Measurements were carried out
for a central rod with a pitch-to-diameter ratio P/D = 1.326 with a
hydraulic diameter, Dh = 9.75 mm. The Re number used was 57,000.
The reported uncertainty in velocity data was 2.1%. In their work,
the conditioning grid was without mixing vanes. They showed
results of axial and cross ow velocity. The cross ow data indicated
that positions close to the grid had velocities 40% greater than posi-
tions farther from the grid. The velocity uctuations also increased
as much as 100% at positions far from the grid compared to positions
close to the grid. In the case of the axial-ow velocity, the mean ow
velocity at X = 3.1Dh was smaller than at X = 20.5Dh and uctuations
were greater closer to the grid. It is found that for this type of grid
the ow recovers after a distance of X/Dh = 10 for axial ow velocity.
For the case of cross ow situations the uctuations remain con-
stant after X/Dh = 10 but the mean value of the velocity component
decreases with distance from the grid. Mean cross ow velocity
and turbulence in the fuel bundle are bigger near the spacer grid;
recovery of mean ow velocity and a reduction in turbulence were
observed downstream. These ow behaviors were assumed to be
mainly due to the mixing vanes. The results indicate that the mixing
vane effect has a strong inuence to around X = 10Dh downstream
of the spacer grid. Conner et al. (2005) presents the experimental
results of a 5 5 fuel bundle with spacer grids. Several spacer types
typical of PWRs were tested in an air loop with fully heated rods.
The work focuses on heat transfer measurements using a specially
design thermocouple holder than can be moved axially inside the
bundle. The Re number tested ranged from 15,000 to 37,000 based
on the hydraulic diameter, Dh = 11.77 mm. The results were calcu-
lated based on temperature measurements at discrete positions
inside the bundle. Conner et al. showed an improvement in the heat
transfer after the spacer-grids but did not provide any explanation
about the probable mechanism behind these improvements.
Yang and Chung (1998) have studied the inuence of the spacer
grids on the turbulent mixing within square sub-channel geome-

Fig. 2. Test section showing the optical arrangement and PIV system.
890 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
Fig. 3. Measurement system.
try. They analyzed the generation and decay of turbulent energy
from LDA measurement results. Neti et al. (1982) measured the
lateral velocity proles within sub channels using 2D-LDA and
they claimed that the magnitudes were less than 1% of the aver-
aged axial velocity. However, detailed secondary ow features were
unclear. Rehme (1987) performed measurements using hot wire
anemometry (HWA) in order to conrm the existence of lateral ow
pulsations between the sub-channels. It is concluded that the ther-
mal mixing in sub-channel geometry is mostly inuenced by almost
periodic macro scale ow pulsations caused by the ow instability
rather than the secondary ow from the Reynolds stress gradients.
Regarding computational work, several attempts have been
made to model the thermal hydraulic behavior of fuel rod bundles.
Tzanos (2004) simulated the ow in a 7 7 rod bundle typical of a
PWR. The spacers were modeled as plates with no thickness and the
blockage ratio of the spacer was used as controlling parameter. The
simulations were made using various turbulence models with sym-
metric and periodic boundary conditions. Tzanos concluded that far
from components that cause signicant ow deections the agree-
ment of mean velocity predictions with measurements was good.
However, near such components the discrepancy between velocity
predictions and measurements could be signicant. These discrep-
ancies were attributed to shortcomings of the k models. Even
in rod bundles without ow deectors, the turbulence predictions
of standard k models showed signicant discrepancy with mea-
surements. Ikeno and Kajishima (2006) used large Eddy simulation
(LES) coupled with the immersed boundary method in an effort to
represent the effect of ow geometry and compared some of their
results with particle image velocimetry (PIV) focusing on develop-
ing swirl. The Re number for this investigations was Re = 4100. The
Fig. 4. Cross sectional conguration of the rod bundle assembly.
results indicated the effects of: the mixing-vanes which caused the
swirl with large-scale uctuation and enhancement of heat trans-
fer. The produced vortices in the rod gap caused enthalpy mixing
enhancement between channels. The model successfully explained
the mechanism of decaying swirl in the rod bundle: a develop-
ing vortex in the rod gap and the decreasing wake behind the
mixing-vanes promoted the decay of swirl more strongly than in
a pipe. Lee and Choi (2007) focused on the computational analysis
of turbulence intensities, maximum surface temperatures of the rod
bundle, heat transfer coefcients and pressure drops of four kinds
of mixing vanes. The authors used a 17 17 rod bundle since they
considered that vortex size effect may not be represented correctly
with a smaller rod array. Caraghiaur and Anglart (2007) measured
the axial velocity in three different sub-channels of 5 5 rod bun-
dle with spacers using LDV. The results were compared with CFD
predictions using a k and SSG turbulence models. It should be
noted that the spacer did not have any turbulence enhancement
device such as mixing vanes. In this study both turbulence models
over-predicted the rate of the velocity downstream of the spacer.
In addition, the root mean square values of the axial velocity did
not agree with the CFD prediction. Caraghiaur and Anglart con-
cluded that the pressure drop over the spacer was under-predicted
by 2030% in comparison with CFD code calculation. There was a
reasonable agreement between axial velocity distributions down-
stream of the spacer with a larger CFD model i.e. with half of the
bundle geometry to remove the inuence of symmetry conditions
imposed on sub-channel boundaries in the CFD predictions.
Verication and validation (V&V) are the primary means to
assess the accuracy and reliability of computational simulations.
V&V methods and procedures have fundamentally improved the
 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898 891

Fig. 5. A 5 5 spacer grid segment lled with FEP tubes emulating a fuel-bundle. (a) Pipes immersed in water. (b) Pipes immersed in air.

credibility of simulations in several elds, such as aerospace indus-
try. The eld of nuclear reactor safety has placed great emphasis
on developing validation benchmarks. Many of these validations
benchmarks are closely related to the overall behavior for deter-
mination of velocity, temperature, and void fraction among other
parameters. However, detailed information of local velocity and
temperature in full-eld are not available for CFD validation.
The importance of computer simulations in the design and
performance assessment of engineered systems has increased dra-
matically during the last three decades. To have justied condence
in this evolving eld, there should be major improvements in the
transparency and maturity of the computer code used, the clarity of
the physics included or excluded in the modeling, and the compre-
hensiveness of the uncertainty assessment performed. CFD codes

Fig. 6. Test section showing the optical transparency achieved inside the bundle by matching the refractive index of the FEP rods and water.
892 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
Fig. 7. Horizontal laser sheet passing across the rod bundle without distortion due
to a good matching index of refraction.
should be applied with care to the ows where dominant physics
are claried.
Thus, the ultimate goal of this work is to contribute to a bet-
ter understanding of presupposed and existent events that could
affect the safety of nuclear power plants by using state-of-the-
art measurement techniques that may elucidate the fundamental
physics of uid ow in rod bundle with spacer grids. In particu-
lar, this work concerns the development of an experimental data
base with high spatial and temporal resolution of ow measure-
ments inside 5 5 rod bundles with spacer grids. The data base is
intended to validate CFD codes at various scales. Measurements are
Fig. 8. Multi-scale measurements.
Fig. 9. Top view of rod bundle showing the planes selected for the PIV measure-
ments.
carried out using dynamic particle image velocimetry (DPIV) inside
an optically transparent rod bundle utilizing the matching index
of refraction (MIR) approach (Hassan and Dominguez-Ontiveros,
2008). This work presents results showing full eld velocity vec-
tors and turbulence statistics for the bundle under single phase ow
conditions.
2. Experimental set-up
The experiments were conducted in the test loop of the optical
multi-phase ow research laboratory (OMRL) at Texas A&M univer-
sity (TAMU) allowing for test pressures up to 3 atm. A schematic of
the hydraulic ow loop is shown in Fig. 1. The ow loop consisted of
a variable speed pump, storage tank, ow meter, ow straightener,
mesh, and test section. Water was used as the working uid, and it
owed from the bottom of the test section in an upward direction.
The ow rate was controlled by the rotational speed of the pump
 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
Fig. 10. Typical instantaneous velocity elds.
using a frequency controller and a control valve in the main loop in
order to assure reproducibility of ow rates and constant torque in
the pump impeller. The uid passed through a series of lters that
limited the size of suspended particles in the uid to 50 m during
experiments.
The instrumentation and control stages for the hydraulic loop
contained turbine type ow meters and rotameters. Pressure trans-
ducers were installed along the test section to measure the local
differential pressures across the spacer grids and the loop pres-
sure. The data was collected using a data acquisition unit NI-SCXI
with the 1000/1600/1314 modules for signal conditioning, analog
to digital conversion, and transmission.
A dynamic PIV system was used to measure full velocity elds.
The system was comprised of a high repetition Nd:YaG twin laser,
optics, high accuracy pulse generator, motorized position stages,
and high speed cameras as shown in Fig. 2. The maximum laser
pulse frequency was 20,000 pulses per second with a 527 nm wave-
length and maximum energy of 10 mJ per pulse. Three high speed
cameras with up to 120,000 fps were used to accurately capture the
uid behavior in the test section. A laser sheet with about 1 mm
thickness was produced using a set of optical lenses and mirrors
which impinged perpendicular to the lateral face of the channel.
893
Fig. 11. Average velocity elds.
Polymer seeding particles with diameter distribution of 69 m
and specic gravity of 1.05 were added to the uid in a volume con-
centration of 1.8e2% for the PIV measurements. Data was obtained
at various planes of the test section. Cross-ow measurements were
obtained by using a second laser sheet to form a horizontal plane
parallel to the spacer grid span. The horizontal laser sheet with a
2.5 mm thickness was shaped using a 50% mirror arrangement as
shown in Fig. 3. A third high speed camera was placed in the test sec-
tion with an inclination of 55 in order to capture the data obtained
using the horizontal laser sheet.
2.1. Test section
For observation of the ow structure the test section housed
a square-arrayed 5 5 rod bundle. The test assembly consisted of
24 fuel rods with a diameter of D = 10.25 mm placed inside the
egg-crated spaces of the spacer subsection and one square rod
with a side length of 12.5 mm as shown in Fig. 4. The rods were
arranged in a square array with a P/D = 1.26. This 5 5 rod bun-
dle with spacer grids was installed in the aluminum square duct.
Both ends of the rod bundle were xed with the same type of
spacer grids. The square duct was made of an aluminum verti-
cal channel with dimensions 101.6 mm 101.6 mm 1828 mm. The
channel had four observation windows in order to have optical
894 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
Fig. 12. Average velocity elds with calculated streamlines.
access to the channels interior faces. A set of four spacer grids
were located inside the channel with a distance between them of
508 mm. However, the grid located at the end of the bundle on the
downstream side was 250 mm from the edge of the previous spacer.
The special shallow plastic rods are located inside the spacer grid
in order to simulate the fuel rods. The rods were fabricated using
a uorinated ethylene-propylene (FEP) plastic with dimensions of
10.25 mm OD 10 mm ID 1270 mm long with a 0.07 mm toler-
ance in all dimensions (Dominguez-Ontiveros and Hassan, 2007).
These dimensions followed closely the dimensions of the real fuel
rods used in the UltraowTM spacers (Kraemer et al., 1995). The
rods were sealed at each end with a conical aluminum cap. The
rods were lled with a solution of 5% chlorine in water. This solu-
tion kept the refractive index of water close to its original value of
1.33 and prevented the growth of bacteria and algae in the uid,
which could alter the optical transparency of the stagnant solution
inside the rods. The measurements were taken around the second
spacer grid, with the rst spacer grid serving as a conditioning grid
for the measurement zone. The spacer grids had mixing devices of
the swirl type attached to the edge of the grid.
The velocity components and their uctuations were measured
in the test section at several locations upstream and downstream
Fig. 13. Typical instantaneous velocity elds.
the grid by particle image velocimetry. The local turbulence quan-
tities were integrated over the uid section to obtain the value of
turbulence in different sections. This value could then be used as
boundary condition for CFD simulation. The ow velocities were
measured using varying axial elevations from 0.1 to 6 hydraulic
diameters, with the top edge of the spacer-grid as the reference
point.
2.2. Matching index of refraction
Optical ow studies of complex geometries such as fuel bundles
were shown to be possible by matching the index of refraction of
the solid test section with that of the working uid. When light
passes, through materials with a different index of refraction, the
light bends at the interface of the materials causing distortion or
hidden areas. However, by matching the refractive index of the
materials, the bending of the light was minimized or eliminated,
making the eld of interest accessible and the test section opti-
cally transparent. MIR techniques have been used in past studies
(e.g., Hassan and Dominguez-Ontiveros, 2008). Although match-
ing the refractive index of the solid and liquid was not difcult in
itself, however, other concerns such as uid price, viscosity, density,
toxicity, ammability, compatibility, etc. complicated the selection
process. In this work, optical transparency was achieved by match-
 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
Fig. 14. Average velocity elds.
ing the index of refraction. The immersed plastic rods have a similar
index of refraction as water. The plastic rods are made from FEP
polymer with a refractive index of 1.338, and lled with a solution
of water and chlorine with a refractive index of 1.333. By match-
ing the index of refraction of both materials optical transparency
was achieved. The working uid for the tests was water, therefore
no correction is necessary and the transparency of the test section
was assured.
Fig. 5 shows an emulated hot-rod (red color) embedded in the
central rod position. The rod is optically accessible when the bundle
was submerged in water. The matched refractive index of water and
the FEP pipes allowed for the visualization of the central rod with-
out considerable disturbances from the rod-layers surrounding it.
Fig. 6 presents another pictorial of the achieved transparency in the
setup. The vertical channel was partially lled with water to demon-
strate the immediate clearness and accessibility of any point inside
the grid spacer. The unobstructed path of the blue laser light sheet
as it passed through the rod bundle is demonstrated in Fig. 7. It was
observed that light was not signicantly distorted demonstrating
the optical transparency obtained by this technique. As indicated
by Hassan and Dominguez-Ontiveros (2008), the wall-thickness of
the cylindrical rod must be chosen to achieve an optimal compro-
mise between rod strength and optical transparency. Consequently,
895
the FEP shallow rods were fabricated with a thin wall for trans-
parency and lled with a solution of water and chlorine to enhance
rod strength and prevent bacteria growing.
3. Results
The measurements for axial velocity at the investigated region
were performed at several planes using a multi-scale approach.
This approach consisted of performing measurements on various
viewing areas of the test section. The scales were selected based on
the physical constrictions and the desired resolution of the velocity
elds. Fig. 8 shows the different scales used in this work. The rst
scale examined the ow behavior in the planes that cover the ve
rods and its respective sub-channels which are denoted by the yel-
low dashed line. At this scale the interaction between sub-channels
and general characteristics of the ow were captured. The equiva-
lent viewing area of the cameras for this scale was 70 mm 60 mm.
The second scale is denoted by the area marked with a solid red
line in Fig. 8. This scale captured the ow characteristics of a
central sub-channel with an equivalent camera viewing area of
15 mm 12 mm. At this scale the ow evolution in the sub-channel
could be observed. The third scale is denoted by the red dashed
line in Fig. 8 and it has an equivalent viewing area of 5 mm 4 mm.
This scale was intended to resolve the local ow behavior, such as
boundary layer thickness, turbulence energy production, turbulent
energy dissipation, eddy size, coherent structures identication,
among other important parameters of interest.
The data collection process included the measurement of every
scale at various planes inside the 5 5 bundle. The selected planes
were parallel to the frontal channels window and they varied in
distance from the wall to the center of the bundle by changing the
position of the optical set-up and cameras using automatic motor-
ized positioning stages as shown in Fig. 9. Fig. 10 shows the PIV
measurement result of the axial velocity vectors in the investigat-
ing region downstream and upstream the spacer grid. The velocity
elds show the whole velocity distribution in the region from 0 to
1.2Dh for both cases.
3.1. Velocity eld around spacer
Several different techniques were used to analyze the behav-
ior of the channel ow. Velocity statistics were obtained from PIV
measurements in the xy plane. These velocity statistics were deter-
mined by ensemble averaging over the number of velocity elds
using Eq. (1),

N
Ai (X)
i=1
A= (1)
N
where A is the quantity being averaged, X is the position vector, and
N is the total number of realizations.
The result from Eq. (1) was a two-dimensional average vec-
tor eld. The uctuating velocity elds were obtained using the
Reynolds decomposition for each vector in each instantaneous
velocity eld using the following expression,
u = U + u (2)
where U is the mean velocity and u is the velocity uctuation.
The mean streamwise velocity determined in this fashion was
subtracted from each instantaneous velocity eld to obtain the
uctuating velocity elds using Eqs. (3) and (4). These uctuating
velocity elds were then ensemble averaged to obtain the stream-
wise and normal intensities as well as the Reynolds shear stresses
896 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898

Fig. 15. Turbulence intensities and Reynolds stresses below and above the spacer grid for Re = 6500.

using Eq. (5). Structure and intensity of the downstream turbulence are impor-
tant for the promotion of heat transfer on the rod surface. As shown
1
N
in Fig. 10, the velocity eld along the rod surface was captured rea-
ui (x, y) = [ui (x, y) U(x, y)] (3)
N sonably by the PIV measurements. The ow elds are presented for
i=1
both positions: downstream and upstream the spacer grid. These
1
N instantaneous velocity elds show non-uniform direction due to
vi (x, y) = [ui (x, y) U(x, y)] (4) the swirling motion of the uid after having passed through the
N
i=1 spacer. There is a change in the maximum velocity attained in the
channel of 43% with respect to the maximum measured down-
1   
N
  stream from the spacer. To obtain the mean value the data was
u v (x, y) = ui (x, y) vi (x, y) (5)
N averaged from more than 45,000 instantaneous vectors. Axial evo-
i=1 lutions of the mean streamwise velocity are shown in Fig. 11. As
Instantaneous velocity vectors are shown in Fig. 10. The swirl shown in Fig. 11 the ow was accelerated through the sudden nar-
was caused by the ow around the mixing-vane. The swirl contains rowing channel spacer. The position of local maximum velocity was
large-scale uctuation, which affected the downstream turbulence. located at the center of the channel with a distance from the spacer
 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898 897

edge X = 6Dh . The averaged data shows a uniform direction prole

of the velocity eld after having passed the spacer which elucidates
the effect of averaging. On the other hand, the velocity prole down-
stream of the spacer grid showed a similar prole to that found in
shear ow along no-slip wall conditions. One of the advantages
of the DPIV obtained measurements was the direct comparison of
instantaneous data with CFD which was previously done using an
averaging approach.
The plane at a distance of X = 0.3Dh intersected with the mixing
vanes, where the ow was the wake ow behind the mixing-vane,
a local acceleration in the narrow gap between the rod and the
tip of mixing-vane is shown in Fig. 12. Streamtraces were calcu-
lated from the instantaneous and averaged velocity elds in an
effort to visualize the ow behavior below and above the grid.
Fig. 12 shows the streamtraces of the velocity averaged velocity
elds. It was observed that the deection of the ow was caused
by the mixing vanes. The results showed the complex inuence of
the spacer grid in the sub-channel formed between the wall sur-
faces of two adjacent rods. The results of 2D velocity elds showed
a considerable reduction in the axial velocity component of the
velocity vector at the center of the channel. This change in mag-
nitude is compensated by an increase in the radial component
of the velocity vector. Flow visualization demonstrated swirling
immediately above the spacer grid. Figs. 13 and 14 show the instan-
taneous and averaged velocity elds respectively downstream and Fig. 16. Vorticity modication below and above the spacer grid for a Re = 6500. The
upstream the spacers in the second plane measured. As previ- velocity eld spans for about 1Dh from the edge of the spacer.
ously explained in this section, the PIV measures were performed
at various planes using motorized positioning stages. The results
shown in Figs. 13 and 14 correspond to the plane 2 illustrated in
Fig. 9. Since the rods are optically transparent when the water ter was divided by the spacer edge. Turbulence models based on a
is owing through the channel, the instantaneous velocity elds developed turbulent ow were fundamentally inapplicable to these
could be obtained. It was observed that measured vectors in this ows.
plane had strong direction changes caused by the mixing vanes
compared to the uniform vector direction seen downstream from
the grid. Once again, the averaged velocity elds are illustrated 3.2. Vorticity modication
in Fig. 14. Axial evolutions of turbulence intensities and Reynolds
stresses are shown in Fig. 15. The turbulence intensity was based Vorticity could be calculated using the average, the instanta-
on the deviation from the mean velocity. Vortex cores were iden- neous, or the uctuating velocity eld. One of the advantages of
tied using the Galilean decomposition approach. In the Galilean PIV measurements is that it offered information about the vorticity
transformation the total velocity is represented as the sum of a eld from the spatial distribution of velocity. In this case, since mea-
constant convection velocity, Uc , plus the deviation as shown in surements were carried out in the xy plane, only one component
Eq. (6) (spanwise) of the vorticity vector could be obtained. Therefore, the
expression for instantaneous spanwise vorticity is
u = Uc + uc (6)
In an effort to locate vortex cores, Adrian et al. (2000), selected  
u v
different fractions of the centerline velocity as the convection veloc- z = (7)
y x
ity implemented in the Galilean transformation technique. The
resulted vortex cores and length scale are illustrated in Fig. 15
with red circles. The averaged uctuating velocity vectors are The derivative calculation was carried out using central dif-
overlapped in the contour plot showing the magnitude of the ference scheme. Substituting the uctuating components of the
turbulence intensities. This gure shows the reduction in vor- velocity in Eq. (7), the uctuating spanwise vorticity was obtained.
tex cores as a direct consequence of the spacer grids inuence Likewise, the substitution of the average velocity components
on the ow behavior. The swirl decreased with distance from yielded the average vorticity eld. In turbulent ows, the existence
the mixing-vanes. Predicting this decrease rate is important to of a high average vorticity is well known. This high level of vortic-
evaluate heat transfer at the rod surface. The mean axial veloc- ity was caused by the extension or stretching of vortex laments
ity decreased with an increase in turbulence when passing the (Tennekes and Lumley, 1974). Vortex stretching, has been regarded
spacer grid. The axial ow recovered after several diameters from as the cause of the high rate of dissipation associated with turbu-
the grid by settling down the ow eld with distance from the lent motions. The effect of the spacer grid was a modication of the
grid. boundary layer caused by the mixing vanes which seemed to pro-
The results indicate that the ow eld around the spacer duce an attenuation of the vorticity magnitude. Although vorticity
included a large-scale unsteadiness and an undeveloped turbu- was always present in the ow, there seemed to be a counteract-
lent ow. The unsteadiness was caused by a separation above ing phenomenon that suppressed the stretching in the ow. Fig. 16
the mixing vane attached to the spacer. The separation caused a shows the average vorticity elds obtained for single phase ow
rapid increase of the pressure drop against inclination of the mix- below and above the spacer grid. It was observed that the close to
ing vanes. The undeveloped turbulence appears in the inlet of the the wall there is a pronounced decrease in the vorticity as the ow
spacer, where the developed main ow in the sub-channel cen- passes through the spacer grid.
898 E.E. Dominguez-Ontiveros, Y.A. Hassan / Nuclear Engineering and Design 239 (2009) 888898
4. Conclusions
The effect of the spacer grid was a modication of the boundary
layer caused by the mixing vanes which produced an attenuation
of the vorticity magnitude. Although vorticity was always present
in the ow, there seemed to be a counteracting phenomenon that
suppressed the stretching in the ow. There is a change in the max-
imum velocity attained in the channel of 43% with respect to the
maximum measured downstream of the spacer. The averaged data
showed a uniform direction prole of the velocity eld after passing
the spacer which elucidates the effect of averaging on the results. In
the other hand, the velocity prole downstream of the spacer grid
shows a similar prole to those found in shear ow along a no-slip
wall.
One of the advantages of the DPIV obtained measurements is the
direct comparison of instantaneous data with CFD which was previ-
ously done using an averaging approach. The results indicated that
the ow eld around the spacer included a large-scale unsteadi-
ness and an undeveloped turbulent ow. The unsteadiness was
caused by a separation above the mixing vane attached on the
spacer. The undeveloped turbulence appeared in the inlet of the
spacer, where the developed main ow in the sub-channel cen-
ter was divided by the spacer edge. Turbulence models based on a
developed turbulent ow are fundamentally inapplicable to these
ows.
Successful use of PIV in addition to matched index of refrac-
tion was used for full eld velocity elds in a complex geometry
without ow disturbance. The results showed that turbulence evo-
lution can be observed using the DPIV technique. Good quality
data with high spatial and temporal resolution open the pos-
sibility of CFD benchmarking and validation among important
quantication of the turbulent structures present in the ow above
spacer grids. The experimental results from this work may be
useful for the development of a turbulence model for applica-
tions such as sub-channel geometries, where the ow is highly
non-isotropic.
References
Adrian, R.J., Christensen, K.T., Liu, Z., 2000. Analysis and interpretation of instanta-
neous turbulent velocity elds. Experiments in Fluids 29, 275290.
Caraghiaur, D., Anglart, H., 2007. Measurements and CFD Predictions of Velocity,
Turbulence Intensity and Pressure Development in BWR Fuel Rod Assembly
with Spacers. Proceedings of the 12th International Meeting on Nuclear Reactor
Thermal Hydraulics (NURETH12), Pittsburgh, PA, September 30October 4.
Chang, S.K., Moon, S.K., Baek, W.P., Choi, Y.D., 2008. Phenomenological investigations
on the turbulent ow structures in a rod bundle array with mixing devices.
Nuclear Engineering and Design 238, 600609.
Conner, M.E., Smith, L.D., Holloway, M.V., Beasley, D.E., 2005. Heat Transfer Coef-
cient Testing in Nuclear Fuel Rod Bundles with Mixing Vane Grids, Water Reactor
Fuel Performance Meeting, Kyoto, Japan, October 26.
Dominguez-Ontiveros, E.E., Hassan, Y.A., 2007. Jet ow mixing measurements in the
vicinity of rod bundles using PIV. Transactions of the American Nuclear Society
Winter Meeting, Washington, DC, November 1115.
Hassan, Y.A., Dominguez-Ontiveros, E.E., 2008. Flow visualization in a pebble bed
reactor experiment using PIV and refractive index matching techniques. Nuclear
Engineering and Design 238, 30803085.
Ikeda, K., Hoshi, M., 2006. Development of rod-embedded ber LDV to measure
velocity in fuel rod bundles. Journal of Nuclear Science and Technology 43,
150158.
Ikeno, T., Kajishima, T., 2006. Decay of swirling turbulent ow in rod-bundle. Journal
of Fluid Science and Technology 1, 3647.
Kraemer, W., Proebstle, G., Uebelhack, W., Keheley, T., Tsuda, K., Kato, S., 1995. The
ULTRAFLOW spaceran advanced feature of ATRIUM fuel assemblies for boiling
water reactors. Nuclear Engineering and Design 154, 1721.
Lee, C.M., Choi, Y.D., 2007. Comparison of thermo-hydraulic performances of large
scale vortex ow (LSVF) and small scale vortex ow (SSVF) mixing vanes in
17 17 nuclear rod bundle. Nuclear Engineering and Design 237, 23222331.
Neti, S., Eichhorn, R., Hahn, O.J., 1982. Laser Doppler measurements of ow in a rod
bundle. Nuclear Engineering and Design 74, 105116.
Rehme, K., 1987. The structure of turbulent ow through rod bundles. Nuclear Engi-
neering and Design 74, 105154.
Simonin, O., Barcouda, M., 1988. Measurements and Prediction of Turbulent Flow
Entering a Staggered Tube Bundle. Proceedings of the 4th international sympo-
sium on applications of laser anemometry to uid mechanics, Lisbon, Portugal,
July.
Tennekes, H., Lumley, J.L., 1974. A First Course in Turbulence. MIT Press, Cambridge,
Massachusetts.
Tzanos, C., 2004. Computational uid dynamics for the analysis of light water reactor
ows. Nuclear Technology 147, 181190.
Yang, S.K., Chung, M.K., 1998. Turbulent ow through spacer grids in rod bundles.
Journal of Fluids Engineering ASME 120, 786791.
Zdravkovich, M.M., 1987. The effects of interface between circular cylinders in cross
ow. Journal of Fluids and Structures 1, 239261.