1 s2.0 S0017931018324591 Main

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

International Journal of Heat and Mass Transfer 130 (2019) 240–251

Contents lists available at ScienceDirect

International Journal of Heat and Mass Transfer


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

Combined flow and heat transfer measurements of backward facing step


flows under periodic perturbation
Z.Y. Li a, S. Guo a, H.L. Bai b, N. Gao a,⇑
a
School of Aeronautics and Astronautics, Dalian University of Technology, Dalian, China
b
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong

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

Article history: The flow fields and convective heat transfer downstream of a backward facing step (BFS) with a
Received 18 May 2018 ReH ¼ 9630 were investigated using experimental methods. The BFS flow was perturbed using a synthetic
Received in revised form 2 September 2018 jet actuator deployed at the separation point with a range of perturbation frequencies
Accepted 18 October 2018
(f A H=U o ¼ 0:04  0:39) and amplitudes (u0A =U o ¼ 0:3  0:9). The momentum transport across the shear
Available online 31 October 2018
layer appeared to be enhanced by the perturbations, resulting in a shortened recirculation bubble, a
larger curvature in the time-average streamline over the separation bubble and enhanced heat transfer
Keywords:
rate in the attachment and recirculation regions. The distributions of the reattachment length, streamline
Backward facing step
Synthetic jet
curvature, static base pressure, turbulent kinetic energy and convective heat transfer rate all suggested
Active flow control there was a critical perturbation frequency f A H=U o  0:22 where the perturbation frequency matched
Heat transfer the characteristic frequency of the attaching shear layer. Flow changed only slightly with the perturba-
tion amplitude when f A H=U o J0:22 but changed significantly when f A H=U o K0:22.
Ó 2018 Elsevier Ltd. All rights reserved.

1. Introduction separated over a bluff body [13,14] and planar offset attaching jet
[15].
Flow separations over a bluff body, including a backward facing The BFS flow can be significantly modified using external per-
step (BFS), can be found in many cooling and drying applications turbation such as a synthetic jet actuator, or a zero-net-mass-flux
and thus have received a great amount of attentions [1]. Several (ZNMF) jet [30], deployed at the step corner, some of such investi-
unsteady modes were identified in a separated flow, including gations were tabulated in Table 1. The impact of perturbation on
shear layer mode, shedding mode and flapping mode [2]. The shear the BFS flow depends mainly on the perturbation frequency and
layer mode refers to the flow structures appear in the shear layer amplitude. Flow structures in the initial part of the shear layer
due to Kelvin–Helmholtz (K-H) instability and has a typical nondi- appear to lock-in with the periodic perturbation with a frequency
mensional frequency of St h ¼ f h=U o ¼ 0:012 [3] near the separation of St A < 0:8 [18,28], and become more coherent than those in the
location. The shear layer structures pair and merge while evolving unperturbed flow [17,18], resulting in a shortened reattachment
downstream forming larger scale structures with a typical fre- length X r . Chun and Sung [17] argued that the optimal perturba-
quency of St ¼ fH=U o  0:07 to 0.20 [3–8] and a propagation veloc- tion frequency, the frequency at which X r is reduced most effec-
ity of around 0:5U o to 0:65U o over and downstream of the tively, was the shedding frequency of the unperturbed flow.
reattachment region [7]. The merged structure was referred to as However, the optimal frequency scattered, e.g. the optimal St A
the shedding mode and was linked to an absolutely instability orig- was 0.20 to 0.40 [16,17,23–25,31,32] and the optimal St A;h 
inated in the central part of the recirculation region in references 0.007 to 0.025 [3,16–18,20,23] (see Table 1) and did not match
[4,5,9]. Beside the shear layer mode and the shedding mode in the shedding frequency of unperturbed flow (St  0:07  0:20) dis-
the separated flow, evidences revealed that the initial part of the cussed above. By now, there was not a generally agreed theory
shear layer over a BFS flapped up and down periodically with a relating the optimal perturbation frequency to the frequencies of
characteristic frequency one order of magnitude lower than the the unsteady modes in the unperturbed BFS flow, therefore more
shedding frequency [3,7,9–12]. This low frequency flapping motion experimental work are needed.
was also observed in other separating/re-attaching flows, e.g. flow The distributions of Reynolds stresses were also modified con-
siderably by the external perturbations and the amount of modifi-
⇑ Corresponding author. cations were perturbation frequency related. In particular, v 2 along
E-mail address: gaonan@dlut.edu.cn (N. Gao). the shear layer was increased by a factor of 2 to 3 when the flow

https://doi.org/10.1016/j.ijheatmasstransfer.2018.10.077
0017-9310/Ó 2018 Elsevier Ltd. All rights reserved.
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 241

Nomenclature

CP coefficient of the static wall pressure, 2P=qU 2o Uo freestream velocity, m=s


Cl momentum ratio of synthetic jet, su02 A =HU o
2
uA synthetic jet velocity, m=s
E electric voltage, Volt u0A root-mean-square value of the synthetic jet velocity in
f frequency, Hz ejection phase, m=s
fA perturbation frequency, Hz u2 ; v 2 ; uv Reynolds streamwise normal, vertical normal and shear
H step height, m stresses, m2 =s2
h heat transfer coefficient, W=m2 K W width of the facility, m
I electric current, Amp   Xm streamwise location of the maximum Nusselt numbers,
k Turbulent Kinetic Energy (TKE), u2 þ v 2 =2 m
ka thermal conductivity of air, W=m K Xr attachment length, m
N number of independent samples x; y; z spatial coordinate in the streamwise, vertical and span-
Nu Nusselt number, hH=ka wise directions, m
P static wall pressure, Pa y0:1 ; y0:5 ; y0:95 locations with mean velocity of 0:1U o ; 0:5U o and
q_ conv convective heat transfer rate, W=m2 0:95U o , m
q_ elec Joule heating, W=m2 d thickness of the initial boundary layer, m
q_ rad radiation heat transfer rate, W=m2 d displacement thickness of the initial boundary layer, m
R radius of a circle fitted to a section of streamline, m dx vorticity thickness U o =ðdU=dyÞmax , m
ReH Reynolds number, U o H=m  uncertainty
Reh Reynolds number, U o h=m m kinematic viscosity of air, m2 =s
St Strouhal number, fH=U o q density of air, kg=m3
St A nondimensional perturbation frequency, f A H=U o h momentum thickness of the initial boundary layer, m
St A;h nondimensional perturbation frequency, f A h=U o
Tw temperature of heated steel foil, K subscripts
To freestream temperature, K ;o unperturbed flow
T1 ambient temperature, K ; max maximum
s width of the synthetic jet actuator, m
t time, s
U; V stream-wise and vertical components of the local mean
velocity, m=s

was perturbed at the frequency St A ¼ 0:2 [21,32]; u2 in the near The amplitude of a synthetic jet actuator was usually presented
wall region was increased by a factor of 2 in the flow perturbed using the maximum velocity of the actuator, uA;max , or the root
at St A ¼ 0:08 [32]. The differences in the flow field under perturba- mean square (r.m.s.) value of actuator velocity, u0A . The maximum
tions of different frequencies suggested different unsteady modes synthetic jet velocity was usually less than 0:3U o [19–21,23–
of the natural flow were excited, and the details of these modes 25,32] with the ratio of the oscillatory jet momentum to the
 
were not yet clearly understood. momentum deficit associated with the step, C l ¼ ðs=HÞ u02 2
A =U o ,

Table 1
Frequencies and amplitudes of periodic forcing applied to reattaching flows (flow type: BFS = backward facing step, study type: E = experimental, N = numerical, forcing
2
frequency: St A ¼ f A H=U o ; St A;h ¼ f A h=U o , forcing amplitude: uA;max =maximum synthetic jet velocity, u0A = r.m.s velocity of ejection phase, momentum ratio C l ¼ su02
A =HU o ).
 
Flow ReH Study StA St A;h Optimal Synthetic jet forcing amplitude
 
type type StA St A;h

Bhatacharjee et al., 1986 [16] BFS 26,000 E 0.2–0.4 0.35 (0.007)


Hasan, 1992 [3] BFS 11,000 E 0.136–0.98 (0.003–0.021) 0.436 (0.010)
Chun and Sung, 1996 [17] BFS 13,000– E 65.0 0.25–0.275
33,000 (0.010)
Chun and Sung 1998 [18] BFS 1200 E 0.305–0.955 0.477 (0.025)
Yoshioka et al., 2001 [19] BFS 1800–5500 E 0.04–0.3 0.19 uA;max =U o ¼ 0:3; u0A =U o  0:2; C l  2:3  103
Wengle et al., 2001 [20] BFS 1480, 3000 E K3:64 0.18 (0.012) uA;max =U o ¼ 0:04, C l  1:7  104 & 8:4  105
Dejoan and Leschziner 2004 BFS 3700 N 0.20 (0.096) uA;max =U o ¼ 0:3; u0A =U o  0:2
[21]
Liu et al., 2005 [22] BFS E 0.275 (0.013), 1.5 (0.07)
Henning and King, 2007 [23] BFS 25000 E 0.08–0.52 0.3(0.017) u0A =U o 6 0:1; C l 6 4:0  104
Dahan et al., 2012 [24] BFS 2000 N 0.004–5.0 0.2–0.4 uA;max =U o ¼ 0:1  0:3; u0A =U o  0:07  0:2; C l
 1:5  104  1:4  103
Kapiris and Mathioulakis, BFS 6940 E 0.026–0.23 0.16 uA;max =U o ¼ 0:078; u0A =U o  0:055; C l  3:0  104
2014 [25]
Ma et al., 2015 [26] BFS 20000 E 0.30 (0.020)
Ma et al., 2017 [27] BFS 20000 E 0.165 (0.011)
Berk et al., 2017 [28] BFS 41000 E 0.21–1.98 (0.02–0.21) uA;max =U o ¼ 1:67; u0A =U o  1:19; C l  0:014
Dandois et al., 2007 [29] Hump 28,275 N 0.14(0.007), 1.1(0.057) uA;max =U o ¼ 0:5; u0A =U o ¼ 0:36; C l ¼ 0:01
Current BFS 9630 E 0.04–0.39 (0.0009–0.009) 0.22–0.30 u0A =U o = 0.3, 0.5 & 0.9, C l  3:6  103 , 0.01
(0.005–0.007) & 0.024
242 Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251

less than 2:3  103 . Here, s was the width of the synthetic jet actu- et al. [33] measured the flow fields with and without perturbation
ator. Actuations with large C l were also used to control flow sep- using PIV and attributed the heat transfer enhancement to the
arations, e.g. C l was 0.01 in Dandois et al. [29] and 0.014 in Berk large scale coherent structures produced by the perturbation.
et al. [28]. The mean reattachment length X r was found to decrease However, PIV snapshots were only used for visualization purpose
as perturbation amplitude increased for St A K0:8 [17,20,23,24]. and turbulence statistics was not analyzed.
Henning and King [23] found that X r was particularly sensitive to Up to now, there has not been a detailed experimental study on
the perturbation at the optimal frequency St A ¼ 0:30, where X r both the convective heat transfer and turbulence statistics of the
decreased by more than 20% for a small perturbation of periodically perturbed BFS flows. The main objective of the present
uA;max =U o ¼ 0:02 and did not seem to further decrease when large investigation is to provide a set of well controlled experimental
amplitude perturbations of uA;max =U o P 0:1 were applied. Perturba- results for this simple, yet important, flow so that the mechanism
tion with a low frequency St A K0:08, on the other hand, had a much of heat transfer enhancement could be better understood and com-
smaller impact on X r for a similar amplitude as the decrease in X r putational models can be validated. The responses of the flow field
was within 10% in any previous investigation [17,19,23,24]. to perturbations of a range of frequencies and amplitudes were
Reattaching flow downstream of a BFS causes a peak in the con- also used to study the behaviours of the unsteady modes in the
vective heat transfer rate near the mean reattachment point where BFS flow. The experimental methodologies are presented in the
the turbulent fluctuations are large for both a laminar separation next section, followed by the results and discussions and the con-
with ReH ¼ 1200 [33] and a turbulent separation with cluding remarks.
ReH ¼ 28; 000 [34]. The peak in the local heat transfer was found
to shift upstream under the periodic perturbation of St A P 0:20 2. Experimental methodologies
using an array of miniature electromagnetic flaps deployed at the
step corner by Inaoka et al. [33]. In particular, the size of the heat The experiments were performed in an open-return wind tun-
transfer peak was increased by nearly 20% when the flow was per- nel using a variable speed blower with a 2.2kW motor. The air went
turbed at St A ¼ 0:20 with a perturbation amplitude of ao ¼ 0:05H. through a diverging section, a settling chamber (900 mm by
Inaoka et al. [33] examined the effect of perturbation amplitude 900 mm by 750 mm), a 750 mm long contraction section with a
ao at only two perturbation frequencies St A ¼ 0:2 and 1.0, and 9:1 area ratio and a 300 mm by 300 mm by 1000 mm (width,
found the size of the heat transfer peak increased with ao . The height, length) test section, shown in Fig. 1(a). The bottom wall
effect of ao was relatively large when ao K0:1H and became small of the test section has two parts. A 10 mm thick Acrylic plate with
when ao J0:1H for both frequencies examined. Moreover, Inaoka a length of Ls ¼ 300 mm was mounted in the flow direction at the

Fig. 1. Schematics of (a) the heat transfer and (b) the flow measurement facilities.
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 243

exit of contraction section. A 10 mm thick Acrylic plate with a coated with candle soot black paint with an emissivity e of
length of 700 mm was offset 25 mm from the first plate in the ver- 0:96  0:01 [35]. The results are presented in terms of Nusselt
tical direction, so that a step with a height (H) of 25 mm was number (Nu ¼ hH=ka ) in this paper. Here, ka is the conductivity
formed. The aspect ratio of the step (W=H) was 12, which was large of air evaluated using the film temperature (averaged between
enough to ensure a two dimensional flow in the central part of the the free-stream and the local wall temperatures). The uncertainty
test section near the mean reattachment point [1]. The plates were of the Nusselt number was estimated using
hold in place rigidly using two plexiglas side walls(10 mm thick,
 2  2  2
325 mm high, 1000 mm long). The measurements were performed Nu q_ conv
ðT w T 1 Þ
for a free-stream velocity of U o ¼ 5:7 m=s. The boundary layer was ¼ þ
Nu q_ conv Tw  T1
tripped at x=H ¼ 12 using a 2 mm diameter steel wire. The initial  2    
velocity profile was measured using a single hot-wire probe at
q_ q_ rad 2 ðT w T 1 Þ 2
¼ elec
þ þ ; ð4Þ
x=H ¼ 1. The boundary layer thickness was d=H ¼ 0:28; the dis- q_ elec q_ rad Tw  T1
placement thickness was d =H ¼ 0:038; the momentum thickness
was h=H ¼ 0:023; the shape factor of the boundary layer is following the approach outlined in Coleman and Steele [37]. In pre-
d =h ¼ 1:64, suggesting this was a transitional boundary layer. sent study, the uncertainties in q_ elec ; q_ rad and T w  T 1 were less than
The Reynolds number was ReH ¼ 9630 based on jet height and 4:5%; 1:0% and 4:6%, respectively, for a 95% confidence inter-
Reh ¼ 220 based on h. The free stream turbulence intensity was val at the location with a minimum T w  T 1 of 12  C, resulting in a
2.5%. The top and the exit of the test section were open to the Nu =Nu of 6:5%. Infrared images were also acquired using a FLIR E5
ambient air. Measurements of wall pressure fluctuations with handhold infrared camera to check the lateral uniformity on the foil.
and without a top wall were compared and the differences were The temperature distributions were uniform from z=H ¼ 1:0 to 1:0
found to be negligible. to within experimental uncertainty.
The heat transfer rate from the wall to the jet was measured by The velocity distributions on the centerline-plane (z ¼ 0) were
replacing the central part of the bottom wall using an electrically characterized using a LaVision 2D particle image velocimetry sys-
heated surface similar to Gao et al. [35,36]. The heat transfer sur- tem (PIV). The field of view (FOV) is shown in Fig. 1(b). Droplets
face was a 700 mm by 78 mm (28H by 3.2H) 0:03 mm thick stain- of vegetable oil with a mean particle diameter of 1 lm were used
less steel foil mounted over a 80 mm wide 10 mm deep cavity in as tracer particles. These particles are introduced to the diverging
the bottom plate. There were a 1 mm wide gap on each side of section of the wind tunnel by using a in-house four-nozzle head
the steel foil which were sealed using tape. One end of the foil droplet generator. A 200 mJ dual head Nd:YAG pulse laser system
was attached to a rigid aluminum holder and the other end (Litron Nano) was used to illuminate the tracer particles. A camera
clamped to another aluminum holder which was attached to a (Highspeedstar) with 1024  1024 pixel resolution was used to
large free weight to tension the foil. The foil was clamped to the capture the images. The time interval between two exposures
holders with machined square aluminum bars that were connected was 180 ls to allow seed particles to travel approximately 6 pixels
to a regulated DC power supply (ATTEN TPR3020S). A 10 mm thick based on the freestream velocity. A total of 2000 image pairs were
styrofoam plate was positioned 10 mm below the foil to form the acquired at a rate of 50 Hz for the unperturbed flow, 47 Hz for the
closed cavity between the plate and the foil. Following [35,36], flow perturbed at 70 Hz and 49 Hz for other perturbed flow. Differ-
the local heat transfer coefficient computed using ence sampling frequencies were used to ensure the samplings were
evenly distributed in every phase of the perturbations. The DaVis
q_ conv q_ elec  q_ rad 8.3 software package was used for image acquisition and post-
h¼ ¼ ; ð1Þ
DT Tw  To processing. Vectors were computed using image cross correlation
with a 24  24 pixel interrogation windows and a 50% overlap.
where the local electrical heating was estimated by
The uncertainties for Reynolds stresses were estimated using
pffiffiffiffiffiffiffiffiffi
IE 2=N where N was the number of independent samples [38]. In
q_ elec ¼ ð2Þ
A this case, N was 400 and 2000 for St A ¼ 0:04 and 0:35, respectively,
and the local radiation heat transfer from the foil was estimated by resulting in a respective uncertainty of 7:0% and 3:2% for a 95%
  confidence level. Uncertainties for the mean velocity as well as ReH
q_ rad ¼ er T 4w  T 41 : ð3Þ were less than 0:2% for a 95% confidence level [39]. The time-
mean static wall pressure at x=H ¼ 1:0, referred to as the static
The current through the foil (I) was measured using a meter on base pressure hereafter, was measured using only a pressure trans-
the power supply, while the voltage drop across the foil (E) was ducer (Omega Engineering PX655) with a resolution of 0:1 Pa and a
measured using a multi-meter (FLUKE 17B) connected to the 1 mm inner diameter pressure tap. The signals from the pressure
mounts. Here, A was the area of the foil between the mounts. transducer were recorded using a PC with NI-6014 data acquisition
The freestream temperature, T o , was measured using a type-K ther- card and a Labview routine. The sampling frequency was 4096 Hz
mocouple located in the upper settling chamber with an accuracy and sampling time was 150 s.
of 0:2  C and was found to be identical to the ambient tempera- A close cavity (300 mm wide, 300 mm long, 10 mm high) was
ture T 1 . The foil temperature, T w , was measured using an infrared formed under the bottom plate using 10 mm thick plexiglas. A
pyrometer (HUAYU YLOA200) located approximately 30 cm loud-speaker (250 mm diameter, 8 X) was mounted under the cav-
directly above the foil and attached to a computer controlled ity. Periodic flow oscillation was achieved from a two dimensional
traversing mechanism. Measurements were performed at the slot with a width of s ¼ 1 mm at the tip of the step, oriented 45°-
streamwise location x ¼ H to 10H with a spatial resolution of H degrees to the free stream. The speaker was driven by a sinusoidal
on the centerline of the facility, z=H ¼ 0. The mean foil temperature signal from a digital signal processor (Texas Instruments
at each location was determined by averaging the temperature TMS320C6713) and amplified by a 150 W power amplifier. The
from 1000 samples measured at 100 Hz. The target diameter, the flow was perturbed at a range of frequencies 10 Hz to 90 Hz, corre-
diameter for the circular region on the foil where the pyrometer sponding to St A ¼ 0:04 to 0.39. Different perturbation amplitudes
evaluated the temperature, was less than 1:5 cm (0:6H). The u0A =U o ¼ 0:3  0:9 were examined for each perturbation frequency,
pyrometer was calibrated using a blackbody infrared calibrator corresponding to a momentum ratio of C l ¼ 0.0036 to 0.024. The
(CEM BX350) with an accuracy of 0:5  C. The top of the foil was amplitude u0A was the root-mean-square value of the ejection
244 Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251

minimum at 0:22KSt A K0:30, agreeing with the reported optimal


perturbation frequency St A  0.20 to 0.40 [16,17,23–25,31,32].
The reattachment length X r decreased significantly when u0A =U o
increased from 0.3 to 0.9 in cases with a low perturbation fre-
quency St A K0:22, in contrast to a smaller decrease in X r in cases
with a high perturbation frequency St A J0:22. It is noted that the
effect of perturbation amplitude on X r in this work was different
to that observed in a Large Eddy Simulation by Dahan et al. [24],
where X r was much more sensitive to u0A in the super-optimal fre-
quency range 0:2KfH=U o K0:4.

3.2. The mean velocity field

The distributions of time-averaged streamwise and vertical


velocities for different perturbation amplitudes (u0A =U o of 0.3, 0.5
and 0.9) and frequencies (St A of 0.04, 0.18 and 0.35) at selected nor-
malized streamwise locations are shown in Fig. 4. The magnitude
Fig. 2. Typical voltage signal applied to the speaker amplifier (dashed line) and the
actuator velocity measured using a single hot-wire probe at the centerline of the of V near x=X r  0:75 was increased by perturbation in every case
slot exit (solid line). The r.m.s. value of actuator velocity in the ejection phase suggesting the perturbed flows approached the wall more rapidly
(shaded area) was used to evaluate the perturbation amplitude, u0A . than the unperturbed flow. The increase in the magnitude of V at
x=X r ¼ 0:75 was particularly large in the flow perturbed at the fre-
phase of the jet velocity (the shaded area in Fig. 2), measured using quency St A ¼ 0:18 and the amount of increase appeared to grow
a single-wire hot-wire probe at the centerline of the slot with zero with the perturbation amplitude. In flows with St A ¼ 0:35, V
free stream velocity. remained unchanged at different perturbation amplitudes.
The curvature of the mean flow over the separation bubble can
be better studied using a time-mean streamline passing through a
3. Results and discussions
location x=H ¼ 0:05 and y=H ¼ 1:0 generated using the stream2
routine of Matlab. A section (0:2 6 x=H  2:2) of this streamline
3.1. The mean reattachment length, X r
was fitted to a circle of radius R using least square method to
obtain its curvature. The radius R=Ro and the time-mean static wall
The normalized reattachment length for different perturbation
pressure measured at x=H ¼ 1:0 for different perturbation frequen-
frequencies and amplitudes (X r =X ro ) are shown in Fig. 3 with the
cies and amplitudes are shown in Fig. 5(a) and (b), respectively.
data from references [17,19,23] included for comparisons. Here,
Here, Ro ¼ 27:5H was the radius for the unperturbed flow and
X r was defined as the location with a 50% forward flow ratio (cf.
the coefficient of time-mean base pressure C P is presented in terms
[10]) at a height of y ¼ 0:05H. The reattachment length for the
of ðC P  C Po Þ=jC Po j with a pressure coefficient for the unperturbed
unperturbed flow X ro  4:2H was in agreement with Hudy et al.
flow C Po ¼ 0:15. The distributions of R and C P suggested
[5,12] where the expansion ratio was also less than 1.1. The nor-
St A ¼ 0:22 was the critical perturbation frequency: R and C P both
malized reattachment length, X r =X ro , was less than unity suggest-
decreased when St A increased from 0.09 to 0.22 and the decreases
ing recirculation region was shortened in every case agreeing
were particularly large from St A ¼ 0:18 to 0.22. Both R and C P
with those reported in Refs. [17,19,23,24]. The reattachment
remained relative unchanged (in the case of u0A =U o ¼ 0:3) or
lengths of each perturbation amplitude were fitted to a parabolic
increased gradually (in the cases of u0A =U o P 0:5) when St A
profile. The profiles exhibited similar features as X r reached its
increased from 0.22 to 0.39. The static base pressure obtained in
a Large Eddy Simulation by Dahan et al. [24] showed a similar
trend to the current measurement, though their critical perturba-
tion frequency was slightly larger.
Following Champagne et al. [40], the vertical spreading rate of
the shear layer was studied using the loci of points at which the
mean velocity was equal to 0.1, 0.5 and 0.95 of the free-stream
velocity, y0:1 ; y0:5 and y0:95 , shown in Fig. 6. The streamwise posi-
tions were normalized using X r . In the region x=X r J0:5; y0:95 and
y0:5 in the perturbed flows with St A ¼ 0:04 were much larger than
those in the unperturbed flow while the corresponding changes in
y0:1 were relatively small suggesting the size of the shear layer, par-
ticularly size the lower half y0:5  y0:1 , was increased significantly
by the perturbation. Moreover, the amount of increases in
y0:5  y0:1 grew with the perturbation amplitude. In the flow per-
turbed at St A ¼ 0:18; y0:1 became less than that in the unperturbed
flow in the region x=X r K0:5, suggesting the perturbed shear layer
spread more quickly toward the wall near the separation point.
Furthermore, y0:1 became smaller as u0A =U o increased indicating
the downward spreading rate increased with u0A =U o . In the flow
perturbed at St A ¼ 0:35, the width of the lower half of the shear
Fig. 3. Distributions of the time-averaged reattachment length of the perturbed
layer (y0:5 -y0:1 ) was increased by the perturbation, while the overall
flow, X r , normalized using that of the unperturbed flow, X ro , for a perturbation width of the shear layer (y0:95 -y0:1 ) remained relatively constant
amplitude of u0A =U o ¼  0.3,  0.5 and M 0.9. implying the impact of perturbation at this frequency was only
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 245

Fig. 4. Profiles of the time-averaged (a–c) streamwise and (d–f) vertical velocities of (soild line) the unperturbed flow and perturbed flows with a perturbation amplitude of
u0A =U o ¼ 0:3  0:5 þ 0:9. The perturbation frequency is St A ¼ (a,d) 0:04 (b,e) 0:18 (c,f) 0:35.

Fig. 5. Distributions of (a) the radius of a circle fitted to a section (0:2 6 x=H  2:2) of the streamline passing through x=H ¼ 0:05 and y=H ¼ 1:0 and (b) time-mean static base
pressure measured at x=H ¼ 1:0.

obvious in the lower part of the shear layer. It is also noted that y0:1 closely associated with the position of the lower side of the shear
collapsed reasonably well for all u0A =U o and St A , e.g. y0:1 =H  0:27 layer.
for St A ¼ 0:04 and y0:1 =H  0:32 for St A ¼ 0:35, suggesting the reat- The thickness of the shear layer can be further evaluated using
tachment location X r determined using 50% forward flow ratio was the local vorticity thickness [14], dx ðxÞ ¼ U o =ðdU=dyÞmax , shown in
246 Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251

Fig. 6. Distributions of y0:1 ; y0:5 and y0:95 of (dash line) the unperturbed flow and perturbed flows with an amplitude of u0A =U o ¼ 0:3  0:5 r 0:9. The perturbation frequency
is St A ¼ (a) 0:04, (b) 0:18 and (c) 0:35.

Fig. 7. For perturbation with St A ¼ 0:04 (Fig. 7, a), dx for all the ‘deep breathing’ motion as described in Refs. [4,5,9]. Moreover, u2
flows increased almost linearly in the region x=X r K1:2, the rate also increased in the middle of the recirculation bubble (y=H  0:9
of increase was larger for perturbation with a larger amplitude, and 0.1 at x=X r  0:5) suggesting the unsteady recirculating flow
due largely to the growth of the upper half of the shear layer as also became stronger due to perturbation.
was shown in Fig. 6 (a). When the flow was perturbed at Reynolds stresses were also increased everywhere in the flow
St A ¼ 0:18 (Fig. 7, b), vorticity thickness was larger for a larger per- perturbed at St A ¼ 0:18, as shown in Fig. 8(b,e,h). In particular,
turbation amplitude in the region x=X r K0:5 due to a larger growth the maxima in u2 and v 2 at x=X r ¼ 0:75 were increased by a factor
rate in the lower side of the shear layer shown previously in Fig. 6
of 3 and 7, respectively, for u0A =U o ¼ 0:9. The large increase in v 2
(b). Vorticity thicknesses in the case of large amplitude, e.g.
was also observed by Yoshioka et al. [19] for a similar St A , and
u0A =U o ¼ 0:9, were only slightly larger than those for a smaller
was attributed to the enhancement of the shear layer structures
amplitude in the region 0:5Kx=X r K0:8; the difference became pro-
by the periodic perturbation. Downstream of the reattachment
nounced again in the region x=X r J0:8, attributed to the growth of
location the increases in Reynolds stresses associated with pertur-
the upper side of the shear layer revealed by the distributions of
bation at St A ¼ 0:18 quickly diminished due to the impact of the
y0:5 and y0:95 . The vorticity thicknesses in the flows perturbed with
wall on the flow structures. Similar to the case with St A ¼ 0:18, per-
St A ¼ 0:35 collapsed reasonable well to that of the unperturbed
turbations at St A ¼ 0:35 also caused large increases in stresses, par-
flow which fitted well to a parabolic profile, as was shown in
Fig. 7(c). ticularly v 2 and uv , due to enhancement of the coherent vortical
structures over the separation region, see Fig. 8(c,f,i). However, the
Reynolds stresses in this case (St A ¼ 0:35) did not vary with u0A =U o
3.3. Reynolds stresses so much as those in the cases with a smaller perturbation fre-
quency suggesting again the generation of vortical structures were
Fig. 8(a,d,g) show the Reynolds normal stresses (u2 and v 2 ) and less sensitive to the perturbation amplitude at St A ¼ 0:35.
shear stress (uv ) with and without perturbations at St A ¼ 0:04. It The effects of perturbation on the turbulence quantities can be
is noted that the scales were different for different stresses. The better illustrated using the maximum local stresses, u2 max ; v 2 max
Reynolds stresses became larger under perturbations in contrast and uv max , shown in Fig. 9. The largest effect of perturbation on
to the unperturbed flow and their magnitude increased with the velocity fluctuation occurred in u2 max at x=X r > 1:0 for
u0A =U o . In particularly, the maximum u2 at x=X r ¼ 1:5 was increased St A ¼ 0:04, associated with the enhanced flapping motion as illus-
by a factor of 2 and 4 for u0A =U o ¼ 0:3 and 0.9, respectively (Fig. 8, trated in the schematic and discussed above. When
a). The large increase in u2 could be associated with a strengthened St A P 0:18; v 2 max increased significantly, likely the results of the

Fig. 7. Distributions of the local vorticity thickness for a perturbation frequency of St A ¼ (a) 0:04, (b) 0:18 and (c) 0:35.
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 247

Fig. 8. Profiles of the Reynolds shear stress (a–c) u2 , (d–f) u2 and (g–i) uv for (solid line) the unperturbed flow and perturbed flows with an amplitude of u0A =U o ¼ 0:3 0:5
0:9. The perturbation frequency is St A ¼ (a,d,g) 0:04, (b,e,h) 0:18 and (c,f,i) 0:35.

enhancement of the vortical structures in the shear layer, as illus- tudes are shown in Fig. 10 with parabolic profiles fitted to the data.
trated in the schematic. For any perturbation amplitude, the TKE became largest near
The distributions of the maximum Turbulent Kinetic Energy St A  0:22 where the mean flow over the separation bubble had
(TKE) in the whole flow field, denoted using kmax where the largest curvature and the base pressure reached their mini-
 
k ¼ 0:5 u2 þ v 2 , for different perturbation frequencies and ampli- mum value, as indicated in Fig. 5. Moreover, when the perturbation
248 Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251

Fig. 9. Distributions of the maximum local (a, d, g) Reynolds streamwise stress, (b, e, h) vertical stress and (c, f, i) shear stress for a perturbation frequency of St A ¼ (a–c) 0:04,
(d–f) 0:18 and (g–i) 0:35 for (dashed line) the unperturbed flow and perturbed flows with a perturbation amplitude of u0A =U o ¼ 0:3  0:5 þ 0:9.

frequency was St A K0:22, the maximum TKE increased significantly with the perturbation amplitude u0A =U o (Fig. 11, d). The changes in
with the perturbation amplitude while the increase became grad- Nu downstream of the attachment location x=X r J1:0 was small
ual for St A J0:22. Similar trends were also observed in the distribu- under perturbation of St A ¼ 0:04. On the contrary, heat transfer
tions of X r (Fig. 3) and streamline curvatures (Fig. 5, a). was increased everywhere including the regions upstream and
downstream of the reattachment location when the flow was per-
3.4. Convective heat transfer turbed at St A ¼ 0:18 and 0.35 (Fig. 11, e & f). The heat transfer in
the recirculation region x=X r K1:0, appeared to increase with the
The distribution of Nusselt numbers for the unperturbed and perturbation amplitude u0A =U o in flows with St A ¼ 0:18 and 0.35,
the perturbed flows are shown in Fig. 11 in terms of (a–c) x=H where the effect of u0A =U o was more obvious in the case with
and (d–f) x=X r . The Nusselt number was initially small in the recir- St A ¼ 0:18, similar to those observed by Inaoka et al. [33].
culation region in the unperturbed flow and grew to a maximum in The magnitude and location of the maximum Nusselt number,
the attachment region x=H  4:5 (or x=X r  1:1), and then Numax and X m , for a range of perturbation frequencies and ampli-
decreased downstream of the reattachment point. When flow tudes are shown in Fig. 12(a) and (b), respectively. Here, the max-
was perturbed at St A ¼ 0:04; Nu increased in the recirculation imum local Nusselt number was normalized using that of the
region and the amount of enhancement in Nu appeared to increase unperturbed flow, Nuo;max . The time-averaged reattachment loca-
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 249

when u0A =U o was further increased to 0.9. The saturation at a large


amplitude could also be observed in the distributions of X r ; R and
C Po . The amount of increase in Numax with amplitude u0A =U o was
particularly large for the low frequency range St A K0:22, where
the location of the heat transfer peak, X m , also decreased signifi-
cantly with u0A =U o . However, the impact of u0A =U o on Numax and
X m became relatively small when the high perturbation frequency
range, St A J0:22. The differences in the sensitivity of heat transfer
to perturbations of different frequency ranges were similar to
those in X r (Fig. 3), R (Fig. 5, a) and kmax (Fig. 10).

3.5. Discussions

The agreement in the critical perturbation frequency St A  0:22


for the distributions of reattachment length X r (Fig. 3), streamline
curvature over the recirculation region R (Fig. 5, a), static base pres-
Fig.  10. Distributions of the global maximum turbulent kinetic energy,
sure (Fig. 5, b), maximum turbulent kinetic energy kmax (Fig. 10)

k ¼ u2 þ v 2 =2, in the shear layer for different perturbation frequencies and and the maximum convective heat transfer Numax (Fig. 12) was
amplitudes. not a coincidence. When perturbed at the critical frequency
St A ¼ 0:22, the shear layer structures appeared to lock-in with
the perturbation and were greatly strengthened by the perturba-
tions, X r , were also included in Fig. 12(b) for comparisons. The tion, indicated by the large increase in the vertical Reynolds nor-
maximum local Nusselt number, Numax , increased with St A and mal stress at the positions of the shear layer (Fig. 8, d–f). As a
reached a peak at St A  0:22 for every perturbation amplitude result, the momentum transport across the shear layer was
examined here; Numax became smaller when a perturbation fre- enhanced, the streamline curvature was increased and the size of
quency of St A > 0:22 was used. The critical perturbation frequency the recirculation bubble was shortened significantly. The heat
was again St A ¼ 0:22, in agreement with those found in the distri- transfer near and upstream of the reattachment point was also
butions of X r (Fig. 3), streamline curvatures R and static base pres- enhanced due to an enhanced turbulence level as well as a larger
sure C Po (Fig. 5) and the maximum TKE kmax (Fig. 10). impingement angle of the reattaching flow.
The maximum local Nusselt number, Numax , became larger The results in this paper also showed that X r ; R; kmax and Numax
when u0A =U o was increased from 0.3 to 0.5, but appeared to saturate varied significantly with the perturbation amplitude u0A for pertur-

Fig. 11. Distributions of the Nusselt numbers of  the unperturbed flow and flows perturbed at St A ¼ (a,d) 0:04, (b,e) 0:18 and (c,f) 0:35 and a perturbation amplitude of
u0A =U o ¼  0.3, 4 0.5 and O 0.9. The streamwise positions were normalized using (a–c) H and (d–f) X r .
250 Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251

Fig. 12. Distributions of (a) the maximum Nusselt numbers of the perturbed flows, Numax , normalized by that of the unperturbed flow, Nuo;max and (b) the streamwise
locations of the maximum Nusselt numbers (X m ) for a perturbation amplitude of u0A =U o ¼ 0:3,  0:5 and 4 0:9. The dash lines are the reattachment length (X r ) for u0A =U o ¼
black 0.3, blue 0.5 and red 0.9. The symbol . is the streamwise location for the maximum Nusselt number for the unperturbed flow, X mo ; and O is reattachment length for the
unperturbed flow, X ro . (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

bation frequency St A K0:22, but only varied slightly for high pertur- Acknowledgement
bation frequency St A J0:22. The difference in the response to per-
turbations of different amplitudes was not reported before to the The research was funded by Natural Science Foundation of
authors’ knowledge. The mechanism is not clear by now and more China (91752101, 11572078) and 973 Plan (2014CB744100).
detailed measurement on the flow dynamics are needed. It is also
noted that the perturbation frequency St A ¼ 0:04 appeared to be a References
special case and the responses of the BFS flow were different to
those of other frequencies St A P 0:08. In particular, X r varied [1] J.K. Eaton, J.P. Johnson, A review of research on subsonic turbulent flow
greatly to the changes in u0A (see Fig. 3) while the static base pres- reattachment, AIAA J. 19 (1981) 1093–1100.
[2] R.L. Simpson, Turbulent boundary-layer separation, Annu. Rev. Fluid Mech. 21
sure C P remained relatively unchanged (see Fig. 5, b). The maxi- (1989) 205–232.
mum kinetic energy kmax for St A ¼ 0:04 did not follow the trend [3] M.A.Z. Hasan, The flow over a backward-facing step under controlled
of other perturbation frequencies (see Fig. 10). Furthermore, u2 perturbation: laminar separation, J. Fluid Mech. 238 (1992) 73–96.
[4] D. Wee, T. Yi, A. Annaswamy, A.F. Ghoniem, Self-sustained oscillations and
increased significantly when BFS flow was perturbed at vortex shedding in backward-facing step flows: Simulations and linear
St A ¼ 0:04 (see Fig. 8, a, and Fig. 9, a), suggesting a large variation instability analysis, Phys. Fluids 16 (2004) 3361–3373.
in the streamwise velocity near the wall, likely caused by an [5] L.M. Hudy, A.M. Naguib, W.M. Humphreys, Stochastic estimation of a
separated-flow field using wall-pressure-array measurements, Phys. Fluids
greatly enhanced flapping mode of the shear layer. 19 (2007) 024103.
[6] D.M. Driver, H.L. Seegmiller, J.G. Marvin, Time-dependent behavior of a
reattaching shear layer, AIAA J. 25 (1987) 914–919.
4. Conclusions [7] A.F. Heenan, J.F. Morrison, Passive control of pressure fluctuations generated
by separated flow, AIAA J. 36 (1998) 1014–1022.
[8] A.F. Heenan, J.F. Morrison, Passive control of backstep flow, Exp. Therm. Fluid
Combined measurements of the flow field and convective heat Sci. 16 (1998) 122–132.
transfer rates in BFS flows under perturbations of different fre- [9] C. Chovet, M. Lippert, J.M. Foucaut, L. Keirsbulck, Dynamical aspects of a
quencies (St A ¼ 0:04  0:35) and amplitudes (u0A =U o ¼ 0:3  0:9) backward-facing step flow at large reynolds numbers, Exp. Fluids 58 (11)
(2017) 162.
were performed. The responses of the flow to perturbations were [10] P.G. Spazzini, G. Iuso, M. Onorato, N. Zurlo, G.M. Di-Cicca, Unsteady behavior of
evaluated using turbulence quantities, reattachment length, static back-facing step flow, Exp. Fluids 30 (2001) 551–561.
wall pressure and convective heat transfer rate. The following con- [11] T.M. Farabee, M.J. Casarella, Measurements of fluctuating wall pressure for
separated/reattached boundary layer flows, ASME J. Vib., Acoust., Stress,
clusions can be drawn from the results. First, a critical perturbation Reliab. Des. 108 (1986) 301–307.
frequency St A  0:22 was identified from the responses, agreeing [12] L.M. Hudy, A.M. Naguib, W.M. Humphreys, Wall-pressure-array
with the reported optimal perturbation frequency [16,17,23– measurements beneath a separating/reattaching flow region, Phys. Fluids 15
(2003) 706–717.
25,31,32]. Shear layer structures were enhanced by the perturba- [13] M. Kiya, K. Sasaki, Structure of a turbulent separation bubble, J. Fluid Mech.
tion at the critical perturbation frequency, the reattachment length 137 (1983) 83–113.
and the static base pressure were minimized, while the streamline [14] N.J. Cherry, R. Hillier, M.E.M.P. Latour, Unsteady measurements in a separated
and reattaching flow, J. Fluid Mech. 144 (1984) 13–46.
curvature over the separation bubble, the maximum turbulent
[15] N. Gao, D. Ewing, On the phase velocities of the motions in an offset attaching
kinetic energy and maximum heat transfer from the heated wall planar jet, J. Turbul 9 (2008) 1–21.
to the fluid were maximized. Secondly, the results suggested the [16] S. Bhattacharjee, B. Scheelke, T.R. Troutt, Modification of vortex interactions in
BFS flow responded differently to perturbations in different fre- a reattaching separated flow, AIAA J. 24 (1986) 623–629.
[17] K.B. Chun, H.J. Sung, Control of turbulent separated flow over backward-facing
quency ranges separated by the critical perturbation frequency: step by local forcing, Exp. Fluids 21 (1996) 417–426.
the flow responded significantly to the perturbation amplitude u0A [18] K.B. Chun, H.J. Sung, Visualisation of a locally-forced separated flow over a
for St A K0:22, but only slightly for St A J0:22. The mechanism caus- backward-facing step, Exp. Fluids 25 (1998) 133–142.
[19] S. Yoshioka, S. Obi, S. Masuda, Turbulence statistics of periodically perturbed
ing the differences is not clear by now and required detailed flow separated flow over backward-facing step, Int. J. Heat Fluid Fl 22 (2001) 393–
investigations in the future. 401.
[20] H. Wengle, A. Huppertz, G. Barwolff, G. Janke, The manipulated transitional
backward-facing step flow: an experimental and direct numerical simulation
Conflict of Interest investigation, Eur. J. Mech. B/Fluids 20 (1) (2001) 25–46.
[21] A. Dejoan, M.A. Leschziner, Large eddy simulation of periodically perturbed
separated flow over a backward-facing step, Int. J. Heat Fluid Fl 25 (2004) 581–
The authors declared that there is no conflict of interest. 592.
Z.Y. Li et al. / International Journal of Heat and Mass Transfer 130 (2019) 240–251 251

[22] Y.Z. Liu, W. Kang, H.J. Sung, Assessment of the organization of a turbulent [31] F.W. Roos, J.T. Kegelman, Control of coherent structures in reattaching laminar
separated and reattaching flow by measuring wall pressure fluctuations, Exp. and turbulent shear layers, AIAA J. 24 (1986) 1956–1963.
Fluids 38 (4) (2005) 485–493. [32] S. Yoshioka, S. Obi, S. Masuda, Organized vortex motion in periodically
[23] L. Henning, R. King, Robust multivariable closed-loop control of a turbulent perturbed turbulent separated flow over a backward-facing step, Int. J. Heat
backward-facing step flow lars henning and rudibert king, J. Aircr. 44 (2007) Fluid Fl 22 (2001) 301–307.
201–208. [33] K. Inaoka, K. Nakamura, M. Senda, Heat transfer control of a backward-facing
[24] J.A. Dahan, A.S. Morgans, S. Lardeau, Feedback control for form-drag reduction step flow in a duct by means of miniature electromagnetic actuators, Int. J.
on a bluff body with a blunt trailing edge, J. Fluid Mech. 704 (2012) 360–387. Heat Fluid Fl 25 (2004) 711–720.
[25] P.G. Kapiris, D.S. Mathioulakis, Experimental study of vortical structures in a [34] J.C. Vogel, J.K. Eaton, Combined heat transfer and fluid dynamic measurements
periodically perturbed flow over a backward-facing step, Int. J. Heat Fluid Fl 47 downstream of a backward-facing step, J. Heat Transfer 107 (1985) 922–929.
(2014) 100–112. [35] N. Gao, H. Sun, D. Ewing, Heat transfer to impinging round jets with triangular
[26] X. Ma, R. Geisler, J. Agocs, A. Schröder, Investigation of coherent structures tabs, Int. J. Heat Mass Tran. 46 (2003) 2557–2569.
generated by acoustic tube in turbulent flow separation control, Exp. Fluids 56 [36] N. Gao, C.Y. Ching, D. Ewing, J.W. Naughton, Heat transfer and flow
(2015) 46. measurements of a planar offset attaching jet with a co-flowing jet, Int. J.
[27] X. Ma, R. Geisler, A. Schröder, Experimental investigation of separated shear Heat Mass Tran. 78 (2014) 721–731.
flow under subharmonic perturbations over a backward-facing step, Flow [37] H.W. Coleman, W.G. Steele, Experimentation and Uncertainty Analysis for
Turbul. Combust. 99 (1) (2017) 71–91. Engineers, second ed., John Wiley & Sons, New York, NY, 1998.
[28] T. Berk, T. Medjnoun, B. Ganapathisubramani, Entrainment effects in periodic [38] J.S. Bendat, A.G. Piersal, Engineering Application of Correlation and Spectral
forcing of the flow over a backward-facing step, Phys. Rev. Fluids 2 (2017) Analysis, Wiley, New York, 1993.
074605. [39] B. Wieneke, PIV uncertainty quantification from correlation statistics, Meas. Sci.
[29] J. Dandois, E. Garnier, P. Sagaut, Numerical simulation of active separation Technol. 26 (2015) 074002.
control by a synthetic jet, J. Fluid Mech 574 (2007) 25–58. [40] F.H. Champagne, Y.H. Pao, I.J. Wygnanski, On the two-dimensional mixing
[30] L.N. Cattafesta, M. Sheplak, Actuators for active flow control, Annu. Rev. Fluid region, J. Fluid Mech. 74 (1976) 209–250.
Mech. 43 (2011) 247–272.

You might also like