Jtam-A4.dvi


JOURNAL OF THEORETICAL

AND APPLIED MECHANICS

52, 2, pp. 499-505, Warsaw 2014

COMPARISON OF SINGLE AND X-WIRE MEASUREMENTS OF

STREAMWISE VELOCITY FLUCTUATIONS IN TURBULENT

BOUNDARY LAYER

Artur Dróżdż, Witold Elsner

Czestochowa University of Technology, Institute of Thermal Machinery, Częstochowa, Poland

e-mail: welsner@imc.pcz.czest.pl

The paper discusses the problem ofmeasuring velocity fluctuations of a turbulent boundary
layer using single and X-wire probes. It seems that the difference between the streamwise
fluctuating component of these two probes results not only from spatial resolution, but also
from influence of the wall-normal fluctuating component, which is usually not considered.
It was shown that the vector summing these two components obtained from X-wire probe
gives the shape of fluctuation distribution obtained from a single-wire probe. It implies that
the underestimation of the near-wall peak of streamwise fluctuating component in X-wire
measurements results from disregarded wall-normal fluctuations, which is obviously taken
in the case of a single-wire probe. Moreover, it was shown that the criteria for wire length
i.e. l+ ¬ 20 could not be sufficient to properly estimate the streamwise and wall-normal
fluctuations.

Keywords: turbulent boundary layer, energy spectra, hot-wire spatial resolution

1. Introduction

Themeasurements of small-scale turbulence are highly challenging due to the insufficient spatial
resolution of the probes especially in high Reynolds number flows. According to the common
opinion, a hot-wire anemometry using a single-wire probe is sufficient to resolve the stream-
wise ux velocity component (Hutchins et al., 2009), however, influence of the wall-normal uy
component on a single-wire probe readings is not throughly discussed.One should be aware that
a single-wire probe does not measure the ux component, but the resultant velocity, composed
of the streamwise ux and wall-normal uy fluctuation components.

Most researchers who do measurements in the turbulent boundary layer believe that the
influence of uy component is insignificant and can be ignored, but it is only a simplifying as-
sumption.The comparison of ux fluctuation distributions obtainedwithDNS (DirectNumerical
Simulation) and from a single-wire probe revealed self-similarity in shape, and some differences
in levels are attributed to the uncertainty error (Monty andChong, 2009; Schlatter et al., 2009).
However, from the physical point of view, the negligible small influence of the uy component in
a single-wire readings is not so convincing. Despite the predominant motion of the streamwise
direction, the vortical structure that is present in a turbulent boundary layer acts on the probe
wire inducing the uy velocity component.

DNS study of Lenaers et al. (2012) confirms the presence of the high value of wall-normal
velocity fluctuations, which occasionally occur in the near-wall region and have the magnitude
larger than their local standard deviation. Since the high values were initially observed only
in direct numerical simulations and not in experiments, it was thought that this effect was
not a physical, but rather numerical artifact. The results of research performed by Hutchins
et al. (2009) concerning spatial resolution effects of a single-wire probe on the energy spectra,
are consistent with the study of Lenaers et al. (2012). The authors show that the calculated



500 A. Dróżdż,W. Elsner

“missing energy” (difference in energy between a shorter and longer wire), due to increase of
the probe wire length, is observed on the viscous length scale λ+ ≈ 600 and viscous time scale
τ+ ≈ 60. This time scale is close to the characteristic single event occurrence (τ+ ≈ 20) of the
wall-normal component noticed by Lenaers et al. (2012). Marusic et al. (2010) showed that the
maximum of uy energy spectrum for the same scale (λ

+
≈ 600, τ+ ≈ 60) occurs close to the

wall. However, this coincidence of energy maximum of uy with missing energy of ux measured
by a single-wire probe was not noticed by theMelbourne group (Marusic et al., 2010; Hutchins
et al., 2009). This coincidence mentioned above could result from small scale vortices, which
produce a strong wall-normal component and transfer the energy to the streamwise component
by the mean shear in the near-wall region.

The paper, based on the measurements of the turbulent boundary layer with single and
X-wire probes, tries to explain the effect of the wall-normal component in the readings of the
single-wire probe. In particular, we consider comparison of fluctuation profiles and time scales
energy spectra of velocity components.

2. Facility and instrumentation

The experiment was performed in an open-circuit wind tunnel, where the turbulent boundary
layer was developed along the flat plate, which was 2807mm long, 250mm wide and 155mm
high, with a boundary layer thickness of up to 25mm.The test section had two pairs of suction
gaps, located in the channel upstream the test section, aimed to control the two-dimensionality
of the flow. To avoid separation, the leading edge of the flat plate had an elliptical shape. The
tripping of boundary layer, after the leading edge of the flat plate was used in order to obtain a
fully developed turbulence. It was resolved using 2mm cylindrical wire fastened to the plate at
210mmfromthe leading edge,whichallowed onetoobtainavalueof theReynoldsnumber,based
on the friction velocity uτ , equal Reτ ≈ 1000. To accelerate further the breakdown of the large-
scale vortex structures, the strip of coarse-grained sandpaper was placed just behind the wire.
The facility was equipped with the computer-controlled, 2D traversing system (in streamwise
and wall-normal direction). The traverse carriage was driven over a maximum displacement of
180mm by a servo motor. The uncertainty of the drive step was 0.001mm with the smallest
step equal 0.01mm. The wall closest position of the hot-wire probe was determined using the
mirrored image. Further details of the test section were given in Drozdz et al. (2011).

Velocity profiles at the zero pressure gradient region were measured with a single hot-wire
anemometry probe of a diameter d= 3µm and length l= 0.4mm (Dantec Dynamics 55P31).
Those measurements were supplemented with X-wire probe of wire diameter d = 5µm and
length l=1.25mm (Dantec Dynamics 55P61). The probes were combined with the DISA 55M
hot-wire bridge connected to a 14 bit PC card. The acquisition was maintained at frequen-
cy 50kHz with 10 seconds sampling records. For the assumed sampling frequency, the non-
dimensional inner scale representation was f+ ≈ 1. It was consistent with the assumption of
Hutchins et al. (2009), stating that for the proper anemometer/probe response cutoff must be
in the range of f+ > 1/3(t+ < 3).

The mean velocity in the core flow was U∞ ≈ 15m/s and the turbulence intensity was
Tu = 0.4%. The ambient conditions were carefully controlled during the measurements. The
scatter of ambient temperatureat theendof the test sectiondidnot exceed 0.2◦. In thecasewhen
the measured temperature was different from temperature during calibration, the temperature
correction ofCTAvoltagewas used, Jorgensen (2002). At the same time the free-streamvelocity
was monitored by means of a Prandtl tube. The scatter of free-stream velocity was found to
be around 0.2% of the mean value. The convergence of the flow statistics up to 4th order was
checked during preliminary tests. The convergence was achieved after approximately 3.5s while
the acquisition time was equal 10s.



Comparison of single and X-wire measurements... 501

The calibration was performed in situ to eliminate the need for the probe to be moved
between the calibration and measurement stage. This removed the risk of probe misalignment
between the calibration and the measurement, and improved the overall accuracy of the expe-
riments. For X-wire probe, the calibration of yaw response of hot-wire is required, and in the
paper, the approach proposed byWillmarth and Bogar (1977) was applied. For the calibration,
the velocity was selected and the yaw angle was changed from −30 to 30 degrees, while the
corresponding voltages fromwire A and B (EA and EB) being recorded. The process was re-
peated for different velocities, which allowed one to build a voltage-to-velocity conversion map.
A typical calibration map is shown in Figure 1.

Fig. 1. Calibration results showing calibration points (o), constant flow velocity lines and constant yaw
angle lines determined from the calibration points

3. Results

Spatial averaging due to a large length of the single-wire is known to reduce the near wall
peak of turbulence intensity (Ligrani andBradshaw, 1987), but also it could falsify higher order
moments, like the skewness andflatness factors (Örlü andAlfredsson, 2010). It could also reduce
the frequency of detected burst events as documented by Johansson and Alfredsson (1983).
Ligrani andBradshaw (1987) found two key recommendations for accuratemeasurements, both
became standards for hot-wire design i.e. l+ ¬ 20 and l/d > 200, where l is length of a wire
(in the viscous units l+), while d is wire diameter. To satisfy these conditions, the miniature
probe with length of the wire l=0.4mm and diameter of d=3µm that was characterised by
l/d = 133 was used. The l/d value did not fulfill the recommendation Ligrani and Bradshaw
(1987), however Fig. 2a shows comparison of the fluctuation distributions of the miniature
wire probe (l = 0.4mm) with the standard wire probe of l = 1.25mm and d = 5µm. The
measurements were performed in the region, where uτ had value ≈ 0.78, in order to obtain, for
the miniature single-wire probe, the value of l+ ≈ 19 and to reach the upper limit of Ligrani
and Bradshaw (1987) recommendation. It could be noticed that the magnitude of the near wall
peak increased by 10%and reached value of uu+ ≈ 8,which is typical for the analyzedReynolds
number. It is apparent that the increase of the fluctuation level is greater than the level of the
estimated uncertainty given in Table 1. The confirmation of the effect of spatial averaging due
to longer wire are energy spectra plots recorded at y+ ≈ 15 presented in Fig. 2b, where for
l/d=250, the drop of energy in the high frequency range is observed.
Another inconsistency in streamwise fluctuation distributions results form using different

types of probe. To analyse the problem, the velocity fluctuations from single andX-wire probes
were measured in a different position of the test section, where uτ ≈ 0.63, which allow one to
get l+ ≈ 16 for a single-wire probe and l+ ≈ 50 for X-wire probe. It may be noticed that the
distribution of uu+ (see Fig. 3) obtained bymeans of the single-wire probe reveals a single peak



502 A. Dróżdż,W. Elsner

Fig. 2. The comparison of fluctuating components fromminiature and standard single-wire probes (a)
energy spectra taken for y+ ≈ 15 (b)

Table 1.Uncertainty of HWAmeasurements

Probe Quantity
Viscous Buffer Log Wake
layer layer layer layer

single U 1.5% 5% 2% 2%
wire u′ 1.5% 5% 1% 20%

X-wire U – 1.5% 1.5% 1.5%
V – 1.5% 1.5% 1%
u′ – 3% 1.5% 10%
v′ – 3% 1.5% 6%

located at y+ ≈ 15, which is typical for turbulent boundary layers at zero pressure gradient.
However, this peak location is not reached for X-wire probe because of the large size, which
not allowed penetration of the boundary layer as close as the single-wire probe. It should also
be noticed that the streamwise uu+ and uu+x distributions, respectively for single and X-wire
probes, are clearly different. However, computations of the resultant velocity fluctuations in the

xy plane i.e. uu+xy =
√

uu2x+uu
2
y/u
2
τ using the values obtained with the X-wire probe barely

show identical shape to the streamwise uu+ obtained with the single-wire probe. The slightly
higher values of uu+xy obtained from theX-wire probe could be due to influence of the spanwise
uz component,which slightly increases the readings of theX-wire probe, but does not in the case
of a single-wire, as the spanwise influence is minor due to the same direction as wire axis. On
the other hand, this influence could also be partly attenuated by the larger measuring volume
of the X-wire probe. These results indicate that the readings of the single-wire probe are highly
influenced by uy fluctuations which also suggests that the uu

+ near wall peak have an elevated
value by the impact of the uy component. The influence has to be stronger with the decrease of
wire lengthbecause the uy fluctuations are inducedonlyby small-scales (Marusic et al., 2010). In
order to confirm this influence on scales from a wider range, the energy spectrum using wavelet
transformation was calculated. The analysis was done for all measured points throughout the
boundary layer thickness.

In order to obtain the wavelet transformation of each recorded signal, the Mexican Hat
wavelet function was used. According to Gordeyev (2000), such a wavelet function is the best
choice toperformtheanalysis of single events in the time signal. Iso-conturs of thewavelet energy
spectra E scaled by the square of friction velocity uτ as a function of the y

+ and time scale
τ+ = τu2τ/ν are shown in Fig. 4. To remove the effect of convection velocity, the time scale τ
was used instead of length scale λ, whichwas used byMarusic et al. (2010). The black cross (+)



Comparison of single and X-wire measurements... 503

Fig. 3. The comparison of fluctuating components from single and X-wire probes

Fig. 4. Iso-contours of the energy spectra E/u2τ: effect of the length scale and geometry of the probe:
single-wire l+ =16 (solid); X-wire l+ =50 (dashed) on streamwise energy (a) and comparison of the
streamwise single-wire and wall-normal X-wire energy (b). Contours are from 0 to 2 with the steps

equal 0.2

corresponds to the scale and location of thenearwall peakofvelocity fluctuations.Figure4 shows
the comparison of energy spectra for the single-wire probewith wire length l+ ≈ 16 andX-wire
probeswithwires length l+ ≈ 50. Figure 4a presents the comparison of streamwise components,
while Fig. 4b shows the comparison of streamwise for the single-wire probe andwall-normal for
the X-wire probe components. Dashed lines on both graphs refer to the component measured
by the single-wire probe, which is treated as the reference case. The continuous iso-lines for ux
(Fig. 4a) and uy (Fig. 4b) obtained from theX-wire probe are superimposed for comparison.As
the energy iso-lines are drawn to the same scales, the lower values of ux measured by theX-wire
probe are easily visible (see Fig 4a). Themore interesting, however, is themaximum shift of the
X-wire streamwise energy to higher time scales. For better interpretation, the iso-lines near the
maximum of E/u2τ were drawn by thick lines. This phenomenon is observed mainly for a small
scales range, that is below τ+ ≈ 100. The similar effects for a single-wire probe with different
lengths of the wire were also observed by Hutchins et al. (2009). On the other hand (Fig. 4b),
the location of the maximum of uy energy (solid thick line Fig. 4b) is shifted towards smaller
scales whoose position can be estimated for τ+ ≈ 60. The displacement of the uy maximum
in relation to ux maximum is consistent with the study of Marusic et al. (2010) and results
from the attached eddies hypothesis, where the wall-normal fluctuations will lack a large-scale
component at thewall due to the blocking (Townsend, 1956). It is clear therefore that the energy
maximumof uy is shifted towards the smaller scale in comparison to the streamwise component.



504 A. Dróżdż,W. Elsner

Therefore, it must have an impact on the readings of the single-wire probe. It is worth noting
that the increasedwall-normal component appears in the same time scale as theburstingprocess
(Drozdz and Elsner, 2011) and should result in overestimation of the near-wall peak captured
by the single-wire probe.

In order to demonstrate that the uy component influences the single-wire probe reading, the

resultant fluctuation energy Euxy =
√

E2ux +E
2
uy
compared to the single-wire probe fluctuation

energy was shown in Fig. 5. It is seen now that Euxy has themaximum (solid thick line) for the
scale, which better corresponds to Euxy iso-contours (dashed thick line) obtained for single-wire
probewith respect to the results shown in Fig. 4a. The scale energy redistribution confirms that
the near-wall peak of fluctuation comes from the increase in the small-scale component of uy
near thewall. These results showthat the single-wiremeasurements give not only ux fluctuation,
but rather the resultant of ux and uy velocity components. Furthermore, this indicates that the
near-wall peak of fluctuation obtained by the single-wire probe could be overestimated due to
the influence of the wall-normal component. Moreover, the criteria for wire length i.e. l+ ¬ 20
could not be sufficient to properly estimate the streamwise and wall normal fluctuations.

Fig. 5. Iso-contours of the energy spectra E(uxy)= (E(u)
2+E(v)2)1/2 component across the boundary

layer thickness – effect of the probe: X-wire l+ =50 (solid); single-wire l+ =16 (dashed). Contours are
from 0 to 2 with the steps equal 0.2

4. Conclusions

The results showed that the difference between the streamwise fluctuating componentmeasured
with the single and X-wire probes results not only from spatial resolution but also from the
influence of the wall-normal fluctuating component, which is usually not considered. It was
shown that thevector summingthese two components, obtained fromtheX-wire probe, gives the
shape of fluctuation distribution obtained from the single-wire probe. To confirm this influence,
the energy spectra using wavelet transformation were calculated. It was shown that the near-
wall peak of single-wire fluctuations is the result of both streamwise and wall-normal small-
scale components of velocity fluctuations. It implies that the underestimation of the near wall
peak of streamwise fluctuating component in X-wiremeasurements results from not taking into
consideration the wall-normal fluctuations, which are obviously taken in the case of the single-
wire probe.Moreover, it was shown that the criteria for the wire length, i.e. l+ ¬ 20, could not
be sufficient to properly estimate the streamwise and wall-normal fluctuations.

Acknowledgements

The investigation was supported by National Science Centre under Grant No. DEC-2012/07/

B/ST8/03791.



Comparison of single and X-wire measurements... 505

References

1. Drozdz A., Elsner W., 2011, Detection of coherent structures in a turbulent boundary layer
with zero, favourable and adverse pressure gradients, Journal of Physics: Conference Series, 318,
6, 062007

2. Drozdz A., Elsner W., Drobniak S., 2011, Application of VITA technique for detection of
the organized structures present in a turbulent boundary layer under an adverse pressure gradient,
Archives of Mechanics, 63, 2, 183-199

3. Gordeyev S., 2000, POD, LSE andWavelet decomposition: Literature Review,Tech. Rep.

4. HutchinsN., NickelsT.B.,Marusic I., ChongM.S., 2009,Hot-wire spatial resolution issues
in wall-bounded turbulence, Journal of Fluid Mechanics, 635, 103

5. Johansson A.V., Alfredsson P.H., 1983, Effects of imperfect spatial resolution on measure-
ments of wall-bounded turbulent shear ows, Journal of Fluid Mechanics, 137, 409-421

6. Jorgensen F.E., 2002, How to Measure Turbulence with Hot-Wire Anemometers – a Practical
Quide

7. LenaersP., LiQ.,BrethouwerG., SchlatterP.,OrluR., 2012,Rarebackowandextreme
wall-normal velocity uctuations in near-wall turbulence,Physics of Fluids, 24, 3, 035110

8. Ligrani P.M., Bradshaw P., 1987, Spatial resolution and measurement of turbulence in the
viscous sublayer using subminiature hot-wire probes,Experiments in Fluids, 5, 407-417

9. Marusic I., Mathis R., Hutchins N,, 2010, High Reynolds number effects in wall turbulence,
International Journal of Heat and Fluid Flow, 31, 3, 418-428

10. Monty J.P., Chong M.S., 2009, Turbulent channel ow: comparison of streamwise velocity data
from experiments and direct numerical simulation, Journal of Fluid Mechanics, 633, 461

11. ÖrlüR., AlfredssonP.H., 2010,On spatial resolution issues related to timeaveragedquantities
using hot-wire anemometry,Experiments in Fluids, 49, 1, 101-110

12. Schlatter P., Orlu R., Li Q., Brethouwer G., Fransson J.H.M., Johansson A.V.,
AlfredssonP.H.,HenningsonD.S., 2009,Turbulentboundary layersup toReθ =2500studied
through simulation and experiment,Physics of Fluids, 21, 5, 051702

13. TownsendA.A., 1956,Thepropertiesof equilibriumboundary layers,Journal of FluidMechanics,
1, 6, 561

14. Willmarth W.W., Bogar T.J., 1977, Survey and new measurements of turbulent structure
near the wall,Physics of Fluids, 20, S9-S21

Manuscript received August 7, 2013; accepted for print November 19, 2013