Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Contents lists available at ScienceDirect

Journal of Wind Engineering & Industrial Aerodynamics

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

Aerodynamic flow control for a generic truck cabin using synthetic jets
G. Minelli a, *, E. Adi Hartono a, V. Chernoray a, L. Hjelm b, S. Krajnovi
c a
a
Department of Applied Mechanics, Chalmers University of Technology, Gothenburg, Sweden
b
Volvo Trucks AB, Gothenburg, Sweden

A R T I C L E I N F O A B S T R A C T

Keywords: This experimental work presents the achievement in drag reduction with the use of active flow control (AFC) for a
AFC generic bluff body. Experiments were done in the Chalmers University closed loop wind-tunnel at Reynolds
Drag reduction number Re ¼ 5  105 . The Re is based on the undisturbed velocity Uinf ¼ 20 m/s and the width of the model W ¼
Synthetic jet
0:4 m. The model consists of a simplified truck cabin, characterized by sharp edge separation on top and bottom
Truck
edges and pressure induced separation on the rounded vertical side edges. The pressure induced separation re-
Vehicle aerodynamic
Experiments produces the flow detachment occurring at the front A-pillar of a real truck. The investigation of the unactuated
and actuated flow was conducted by means of time-resolved particle image velocimetry (PIV). Loudspeakers were
used as the actuation device. These were characterized before the actuation study, highlighting an interesting
analogy between actuation frequency and jet vortex pair size. The effects of different actuations were evaluated
with hot wire anemometry. The effect of the actuation was studied using phase averaging and modal analysis. A
notable reduction of the side recirculation bubble was observed. The nature of the separation mechanism was
investigated and related to different actuation frequencies spanning the range 1 < F þ < 6:2. As for the Re, the non-
dimensional frequency F þ is based on the undisturbed velocity Uinf and the width of the model W.

1. Introduction control the flow separation by means of acoustic excitement of the

The past two decades have experienced the exponential growth of an
actuation technology known as “synthetic jet”. The main feature of a
synthetic jet lies in the capacity to generate a jet flow using the working
fluid of the flow system. In this way, no complex piping is necessary to
recreate a sinusoidal jet signal able to manipulate the surrounding flow
field. This technique turned out to be versatile for different flow control
applications. Aeronautic researchers were the principle investigators of
this technique. The study pursued in order to enhance the aerodynamic
performance of a post-stall aerofoil pushed the investigators to find a
non-invasive technique that satisfies the following requirements:
 It should be easily adaptable to the flow conditions.
 It should introduce momentum in the boundary layer upstream of
separation.
 It should be embedded in the surface of the aerofoil.
 It should use a limited level of energy.
A synthetic jet is able to meet all these requirements. The mechanical
motion of a solid membrane opens the possibility to easily modulate the
amplitude and frequency of a sinusoidal jet flow. The first attempts to
boundary layer have been reported by different authors (Huang et al.,
1987; Hsiao et al., 1990, 1994). The investigators observed a post-stall
increase in lift, that was later quantified up to 50% (Chang et al.,
1992). The later published works (Seifert et al., 1996; Wu et al., 1998;
Amitay and Glezer, 2002a, 2002b; Smith, 2002; Glezer et al., 2005)
asserted its potential. Several reviews, collecting old and the latest
achievements, were published during the years (Gad-el Hak et al., 1998;
Amitay and Cannelle, 2006; Cattafesta and Sheplak, 2011; Brunton and
Noack, 2015), although, the work proposed in Glezer (2011) identifies an
interesting and essential distinction based on the control frequency.
Here, the author distinguishes the presence of two synthetic jet control
strategies. The first approach employs low frequencies, which interact
with the global flow instability. In this way, the actuation is able to lock
the separated flow motion to its own frequency. The second strategy
employs higher frequencies, out of the range of the so-called “receptive
band”. In this case, as mentioned by the author, the control frequency is
decoupled from the global instability of the base flow. The remarkable
results that are cited have attracted various investigators and opened the
possibility to adapt this control strategy to bluff bodies. Transport vehi-
cles (e.g. trains, trucks, ships and cars) are indeed characterized by
heavily separated flow. Always inspired by previous aeronautic

* Corresponding author.
E-mail address: minelli@chalmers.se (G. Minelli).

http://dx.doi.org/10.1016/j.jweia.2017.05.006
Received 3 February 2017; Received in revised form 19 April 2017; Accepted 20 May 2017

0167-6105/© 2017 Elsevier Ltd. All rights reserved.

G. Minelli et al.
achievements, the aerodynamics of vehicles has significantly evolved,
starting from the early 1980s, and this has enhanced their efficiency. The
work of Choi (Choi et al., 2014), Ahmed (Ahmed et al., 1985), Modi
(Modi et al., 1995) and Cooper (2003) are just a few examples of many
achievements this research has produced over time. In the same way, the
manipulation of the flow by an active flow control strategy can produce a
drastic reduction of aerodynamic drag and, as a consequence, a signifi-
cant reduction in fuel consumption as shown in Seifert et al. (2015);
McNally et al. (2015). Several studies have pursued the reduction of bluff
body separated flow regions and a reduction of the aerodynamic drag by
means of different control strategies: from plasma actuators (Vernet
et al., 2015) to pulsating jets (Krajnovi
c et al., 2010; Barros et al., 2016)
and synthetic jets (Ben Chiekh et al., 2013). The present work provides a
novel study, to the knowledge of the authors, in its attempt to adopt the
synthetic jet control strategy, used for aerofoil, to manipulate the flow
separation occurring at the A-pillar of a truck cabin. This experimental
work is part of an ongoing investigation of a control approach of the
separated flow region in trucks. Earlier work in the same project can be
found in Minelli et al. (2016). Looking closer at the flow features of a
truck, there are four main sources of drag due to massive flow separation:
the wheels and underbody, the wake, the gap between the tractor and
trailer, and the front separation, Fig. 1a. This work focuses on the front
separation occurring at the A-pillar of a truck cabin, Fig. 1b. A simplified
model was chosen to reproduce the A-pillar flow separation and to
control its behaviour, Fig. 2 and Table 1, achieving the following goals:
 A characterization of the actuation is reported, with the aim to
investigate the formation of the jet vortex pair at different actuation
frequencies.
 The flow of unactuated and actuated cases is investigated by means of
time-resolved PIV.
 The unactuated flow snapshots are post processed by means of Proper
Orthogonal Decomposition (POD) and Fast Fourier Transform (FFT)
analyses.
 Four different actuation frequencies are chosen following the modal
analysis and the achievements described in Minelli et al. (2016).
 The reduction of the side recirculation bubble is achieved and
described for all actuations.
 The physics defining each actuated case is described (by means of a
phase averaging of the PIV snapshots), corroborating the presence of
a ”receptive band” of frequencies (Glezer et al., 2005).
 The downstream effect of the actuation in the wake of the model is
studied.
2. The experimental set-up
The model consists of a simplified truck cabin, characterized by sharp
edge separation on top and bottom edges and pressure induced separa-
tion on the rounded vertical front edges. The pressure induced separation
reproduces the flow detachment occurring at the front A-pillar of a real
truck cabin. The model was designed to highlight the A-pillar separation,
which was studied in detail. In particular, two of the domains observed
(Fig. 3b) were investigated by means of time-resolved PIV. Fig. 2a and b
shows the model during the assembly stage. Four speakers (WAVECOR
Fig. 1. Main sources of aerodynamic drag for a truck (a). The A-pillar separation and the effect of the actuation (b).
82
Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90
SW182BD01) were employed to reproduce the jet flow described in
Fig. 4b. Once the front and the top lid of the model were assembled, the
speakers were sealed to avoid air leakages. Fig. 2c and Table 1 present the
model and test section's main dimensions. All the dimensions are scaled
by the model width, W ¼ 0:4 m. The model is defined by a square cross-
section, where width and height have the same dimension.
Experiments were carried out in a closed circuit wind-tunnel at
Chalmers University of Technology. The test section dimensions are
3:00  1:80  1:25 m3, and the speed range is 0–60 m/s. The flow tur-
bulence level was within 0:15%. Fig. 2c shows the model placed in the
wind-tunnel test section. Hot wire measurements were made to charac-
terize the jet flow at the actuation slot. The velocity traces were phase-
averaged over 500 blowing and suction cycles. Since the hot wire mea-
surement technique is not able to distinguish the flow direction, a
correction was implemented for the suction part of the cycle. Thus, it was
initially inverted to reflect the correct behaviour of the flow. During this
operation, the wind tunnel was turned off to record only the jet flow at
the slot. In this way, and knowing the actuation frequency, it was easier
to distinguish the flow direction, i.e. the blowing or suction part of the
cycle. PIV images were registered by a monochrome double-frame
SCMOS camera SpeedSense M340 by Dantec with a 2560-pixel by
1600-pixel resolution, 12-bit pixel depth, and 10-μm pixel size. The
camera was equipped with a 105-mm f/2.8 lens from Sigma. The camera
registered image pairs at a 400 Hz frame rate at full resolution in double
frame mode. Flow seeding was achieved with a fog generator and a glycol
based fluid. The Dual Power Nd:YLF LDY300-PIV laser from Litron pro-
vided up to 2  30 mJ at 1000 Hz and a 527-nm wavelength. The laser
was equipped with a laser guiding arm and laser sheet optics. The flow
field area illuminated was 200  400 mm2. Dantec Dynamic Studio 2015
software was used for data acquisition and post-processing. Each data set
included 800 images, which corresponds to a measurement period of 2 s
with a spatial resolution of 0.125  0.156 mm2 per pixel. The vector
calculation was performed in a multi-pass procedure with a decreasing
window size. The initial interrogation window size was 64-pixel  64-
pixel with a 50% overlap and square 1:1 weighing factor for the first two
passes. Finally, three passes were performed with a 32-pixel  32-pixel
window size, 50% overlap and a round 1:1 Gaussian weighing factor. 2D
snapshots of the flow were recorded during the experiments. Velocity
components were recorded in two planar regions at z ¼ H=2, Fig. 3. The
snapshots captured were used for POD and FFT analyses and phase
averaging. Fig. 3 shows the dimensions of the observed domains.
2.1. Modal and frequency analyses
One single POD temporal coefficient can oscillate at different fre-
quencies, while the FFT analysis highlights the area of interest of the
actual frequencies of the flow. It is interesting to compare the two ap-
proaches in order to gain a complete understanding of the flow structures
in terms of both the energy content and characteristic frequencies.
The POD here is made on velocity snapshots sampled with a constant
time step. For example the wall normal velocity component set of snap-
shots is described by vm ¼ vðx; t m Þ at time t m ¼ mΔt, m ¼ 1; …; M with
the time Δt, and a Cartesian coordinate system x ¼ ðx; y; zÞ with unit
vectors ex ; ey ; ez respectively.
As was originally proposed by Lumley (1970), this method is based on
G. Minelli et al. Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Fig. 2. View of the open model (a and b). Four speakers are used as actuators. The motion of their membranes creates a jet flow as described in Fig. 4b. Wind-tunnel test section and the
model in place (c).

Table 1 X
∞ X
M1
Dimension of the numerical domain scaled by the model width W ¼ 0:4 m. R is the non- v0 ðx; tÞ ¼ bi ðtÞvi ðxÞ≈ bi ðtÞvi ðxÞ þ vres ðx; tÞ: (2)
dimensional radius of the rounded corners and G is the slot span dimension (see Fig. 4a). i¼1 i¼1

K S D L R G
The definition can now be written in a more compact form if we
3 4.5 0.45 0.9 0.05 0.0025 consider that b0 ¼ 1 and v0 ¼ 〈v〉, following (Rempfer and Fasel, 1994),

X
M 1
energy ranking of orthogonal structures computed from a correlation vðx; tÞ ¼ bi ðtÞvi ðxÞ: (3)
matrix of the snapshots. A Singular Value Decomposition (SVD) approach i¼0

is used to conduct the POD analysis on the set of snapshots mentioned. The first and second moments of the POD modes coefficients are:
Note that the snapshot POD method limits the number of POD modes to
M 1. In the present POD, the wall normal velocity component is 〈bi 〉 ¼ 0; 〈bi bj 〉 ¼ μi δij : (4)
decomposed in the mean field, 〈v〉, and the fluctuating part, v0 , as
The energy content of the single mode, Ki , is approximated from the
vðx; tÞ ¼ 〈v〉ðxÞ þ v0 ðx; tÞ (1) mode coefficients as

The fluctuating part is then approximated, by the SVD approach, with 1

Ki ðtÞ ¼ b2i ðtÞ (5)
space dependent modes, vi , and time dependent mode coefficient, bi , as 2

and the total energy, KΣ ðtÞ, is evaluated as

Fig. 3. Sketch of the domains observed in experiments.

Fig. 4. Zoom of the rounded corner and the slot position (a). The jet flow produced by the motion of a solid membrane (b).

83
G. Minelli et al.
X
M1
KΣ ðtÞ ¼ Ki ðtÞ:
i¼1
In the present work, the POD study is made over 800 time steps. The
non-dimensional time step Δt ⋆ between each snapshot was Δt ⋆ ¼ ΔtUinf =
W ¼ 1:2  101 . In this way, the highest non-dimensional frequency, F þ ,
Eq. (9), is defined (according to the Nyquist frequency) by half of the high
speed camera's frame rate as F þ ¼ 4:2. In the present POD, mode 1
represents the mean value of the flow field and contains the largest
amount of energy. Concerning the FFT analysis, a classical approach is
applied to the set of snapshots. The discrete time signal of each grid point
of the planar snapshot is transformed into its discrete frequency domain.
In this way the energy content of each frequency can be plotted, for each
grid point, on the 2D domain. Figs. like 9b and 13 show the energy
content of a chosen frequency in each point of the domain.
2.2. Actuation's parameters
The magnitude of the velocity at the actuation region (G in Fig. 4a),
Uafc , was defined by a sinusoidal jet signal (Fig. 4b) as follows,
Uafc ¼ 0:26Uinf sinðt2πfa Þ
where Uinf is the magnitude of the free stream velocity, and fa is the
chosen actuation frequency. Two non-dimensional parameters describe
the performance of the actuation. The first parameter is the momentum
coefficient, Cμ . It is an indicator of the energy spent for the actuation (I j )
with respect to the energy of the unactuated flow.
 
2 T=2 2
Ij ¼ ρ G∫ Uafc ðtÞdt;
T j 0
Ij
Cμ ¼ : positive part of the actuation cycle, one can notice that the almost si-
1 2
ρWUinf
2
here, ρj ¼ ρ is the fluid density and T is the actuation period. Cμ ¼ 4:9 
104 is low but sufficient to excite the boundary layer and the natural
instabilities that characterize the flow. All the frequencies in the present
work are described in terms of the second non-dimensional parameter,
the reduced frequency (also called the actuation Strouhal number).
f (300 Hz) produces a smaller structures that dissipate faster and propagate
Fþ ¼  :
Uinf W
here, f represents the frequency in Hertz.
Fig. 5. Hot wire probe positioning (a). Peak velocity measured at the centerline of the slot over voltage E at different frequencies (b). Phase averaged centerline velocity during one cycle of
the actuation at different frequencies (c). The dashed line represents an ideal sinusoidal behaviour. F þ ¼ 1 (50 Hz, stars), F þ ¼ 2:1 (100 Hz, circles), F þ ¼ 3:1 (150 Hz, triangles), F þ ¼ 6:2
(300 Hz, squares).
Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90
3. Results
(6)
3.1. Speaker characterization
The time series of the velocity magnitude were measured along the
centreline of the jet slot for different amplitudes and frequencies.
Changing the voltage, E, and the actuation frequency of the speaker input
signal, the velocity peak of the jet (Uafc , Eq. (7)) changes consequentially,
Fig. 5b. The distance between the hot wire probe and the jet orifice is
fixed during these measurements at J ¼ 1:7G (see G in Table 1). A pre-
vious work (Smith and Glezer, 1998) describes the evolution of the
suction and blowing cycles with the distance from the jet orifice. The
authors show that, while the peak magnitude of the induced velocity
stays constant up to J ¼ 5G, the suction minimum drastically decreases,
becoming a small percentage of the peak velocity. As early as J ¼ 2G, the
suction minimum decreases to 10% of the peak velocity. This tendency is
confirmed by the measurements collected in the present work. Fig. 5c
shows the phase averaged velocity during one cycle of the actuation at
different frequencies. In particular, the actuations at four different fre-
quencies emitting the same peak velocity 0:51Uinf (Fig. 5b) were inves-
tigated. The speakers require a higher voltage to emit the same peak
velocity at higher frequency. Thus, the following configurations were
(7) selected for the analysis: 50 Hz at 1.4 V, 100 Hz at 1.4 V, 150 Hz at 2.5 V
and 300 Hz at 6 V.
The trend observed in Fig. 5c is relevant for understanding the roll-up
and advection that define the formation of the vortex pair at the orifice of
the actuator. In contrast to the measurements presented in Smith and
Glezer (1998), here, the position of the probe is kept constant while the
actuation frequency was changed, as illustrated in Fig. 6a. Comparing the
results found in Smith and Glezer (1998) and Fig. 5c, one can observe the
analogy between different blowing cycle shapes. In this study, Fig. 5c, the
velocity peak is shifted in time with increased frequency (Fig. 5c from left
(8) to right). The shape of the velocity curve also changes. Looking at the
nusoidal shape, produced by the first actuation (50 Hz), changes to a
more characteristic peaked shape, second and third actuation (100 Hz
and 150 Hz), and, it comes back to a more regular profile in the fourth
actuation (300 Hz). The positive part of the actuation cycle increases
with increased frequency. These three features were also observed in
Smith and Glezer (1998), where the location of the hot wire probe was
changed, and the frequency was kept constant. These analogies can be
explained by looking at the sketch in Fig. 6. The high actuation frequency
(9)
to a shorter distance. On the other hand, the roll-up vortex pair produced
by lower frequencies (50 Hz) is larger and dissipates over a longer
84
G. Minelli et al. Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Fig. 6. The formation of different roll-up vortex pairs at different actuation frequencies
(a). The hot wire probe is fixed and the actuation frequency varies. The formation of one
roll-up vortex pair for a fixed frequency (b). The hot wire probe moves along with the
evolving vortex pair.

Fig. 8. POD decomposition of the wall normal velocity component. Energy content, from
distance from the slot. Therefore, keeping the hot wire location constant, mode 1 to mode 20. Mode 1, which is the average flow field, contains the 12% of the total
one can observe different cycle profiles for different actuation fre- energy content.
quencies. Bearing in mind the results observed in Smith and Glezer
(1998) and the findings of this study, one can infer that higher actuation layer instability triggers the formation of the second larger structure
frequencies produce smaller roll-up vortex pair. In fact, the blowing cycle instability downstream. In other words, the shear layer vortices define a
curve of 300 Hz actuation (Fig. 5c, squares) shows the typical behaviour shear separation angle and frequency. These two factors determine the
of an already evolved vortex pair, while the other three cases show the formation of more energetic structures downstream. Thus, introducing a
cycle profiles of an early stage formation of a vortex pair, Fig. 5c. By this small disturbance on the shear layer structures can dramatically affect
reasoning, it can be inferred that the actuation frequency plays a crucial the entire separation mechanism. Taking a closer look at Fig. 9, the flow
role in the momentum introduced in the boundary layer, and this is of structures increase size and energy content (Fig. 8) moving from mode 7
major importance in order to find an optimal flow control actuation. to mode 2. Interesting frequencies, describing the motion of the shear
layer and the wake structures, were found in a range between
0:7 < F þ < 3:1. Fig. 9b shows the spatial distribution of the energy of
3.2. The actuation of the flow
three different frequencies that can be associated with the three modes
depicted by Fig. 9a.
The recirculation bubble arising on the side of the model of a generic
The study of the unactuated flow leads to a better understanding of
truck cabin was investigated by means of POD and FFT analyses and
the structures that play a major role in the evolution of the turbulent
phase averaging. The maximum observed frequency is limited to F þ ¼
separation. A correct coupling between the actuation frequency and the
4:2 (200 Hz) by the maximum frame rate of the high-speed camera (see
global instabilities of the base flow is essential for the suppression of the
end of Section 2.1). According to a previous numerical study (Minelli
recirculation bubble. Looking at the unactuated flow separation (Figs. 7b
et al., 2016), this frequency range is sufficient to capture the main flow
and 9), even if the frequency of the side wake structures is lower when
structures and the instability of the shear layer. The averaged flow field of
compared to the shear layer instability, their interaction is of great
the side separation region and the wake is depicted in Fig. 7a. The sep-
importance (Unal and Rockwell, 1988). For this reason, an actuation in
aration of the flow is massive, and no reattachment is observed on the
this range can amplify and eventually lock the unsteadiness of the flow.
side face of the model. Thus, a large wake is observed downstream in the
Thus, four actuation frequencies, in the range of the ”natural” frequencies
model, worsening the overall aerodynamic performance. Fig. 7b shows a
of the flow, are chosen for the actuation. This range is also called
snapshot of the instantaneous flow velocity streamlines captured along
“receptive band” (Glezer et al., 2005) and it represents the spectrum
the model. The small structures characterizing the shear layer instability
region where the actuation can lock the separation mechanism to its own
are visible and their evolution is crucial for the formation of the larger
frequency. Further explanation of this observation is provided by the
wake structures appearing downstream. To clarify this aspect, POD
subsequent figures and discussion. The jet signal emitted by each actu-
modes are plotted and correlated to three distributions of energy of three
ation was described in detail in Section 3.1. As a reminder, lower actu-
frequencies. Figs. 8 and 9 show the modal analysis of the wall normal
ation frequencies produce a larger vortex pair and introduce a larger
velocity component. The modes in Fig. 9a are listed from left to right,
momentum in the boundary layer. Thus, research into an optimum
from the least to the most energetic mode. The visible flow pattern, only
actuation, keeping constant the peak velocity produced at the jet slot
guessed in Fig. 7b, is confirmed by the POD analysis. In particular, the
(Fig. 5b), is carried out. The global effect on the flow field is clearly
small structures describing the shear layer instability are captured by
defined by the actuation. Fig. 10 shows the averaged flow field of the
mode 7, Fig. 9a (first column), while a second, more energetic, instability
unactuated and actuated cases. The impact of the actuation is strongly
is captured by mode 2, Fig. 9a (third column). Now, combining the in-
dependent on the frequency employed. The highest reduction of the side
formations of the instantaneous flow field, depicted in Fig. 7b, and the
recirculation bubble is achieved utilizing the actuation frequency that
decomposition in single modes, Fig. 9a, one can notice that the shear

Fig. 7. Streamwise averaged velocity and velocity streamlines (a). A flow instant described by velocity streamlines, side observed domain (b). Side and rear observed domains (see Fig. 3).

85
G. Minelli et al.
Fig. 9. POD modes of the wall normal velocity component (a). From left to right: mode 7, mode 3 and mode 2. Spatial distribution of the energy of different frequencies (b). From left to
right: F þ ¼ 3:1, F þ ¼ 2:1, F þ ¼ 0:8. Side observed domain (see Fig. 3).
defines the shear layer structures, Fig. 10d. The actuated case at F þ ¼ 2:1
also shows a consistent reduction of the separated flow, Fig. 10c, while
the low frequency and high frequency cases, Fig. 10b and e respectively,
show a weaker global impact on the recirculation bubble. Nevertheless,
all the actuated cases experience a deflection of the shear layer toward
the surface of the model, enhancing the behaviour of the flow to
different extents.
Even though the cases presented in Fig. 10b and e give a similar
reduction of the recirculation bubble, their interaction with the flow field
is fundamentally different. Fig. 11 shows the phase averaged flow field of
four different instants during one actuation cycle. The first three actua-
tions (Fig. 11b, c and d) show the clear presence of a train of vortices
rolling up from the separation point. These train of eddies is defined by
the actuation frequency of each case. The last case, Fig. 11d, diverts from
this behaviour and no presence of recursive structures, advected down-
stream by the control, is observed. As mentioned before, the reduction of
the recirculation bubble produced by this last case is similar to the one
produced by the first one, although it is different in the separation
mechanism. The structures visualized in Fig. 11b are clearly defined by
the actuation frequency, and a new structure arises, from the A-pillar,
every time a new actuation cycle begins. In this case the frequency is low
enough to roll-up, during the blowing period, a vortex of large dimen-
sion. As a consequence, large and periodic vortices are produced. In
contrast, the last case, Fig. 11e, does not present any periodic structures
or train of eddies related to the actuation frequency, but a quite stable
recirculation bubble. Note that Fig. 11e represents the phase averaged
velocity at F þ ¼ 6:2. The recirculation bubble of this case is observed to
move, repetitively in time, at a much lower frequency, F þ ¼ 0:75. Thus,
this case was phase averaged using F þ ¼ 0:75, Fig. 12. This figure shows
Fig. 10. Averaged flow streamlines superimposed over contours of the streamwise velocity. Unactuated flow (a). Actuated flow at F þ ¼ 1 (b). Actuated flow at F þ ¼ 2:1 (c). Actuated flow
at F þ ¼ 3:1 (d). Actuated flow at F þ ¼ 6:2 (e). Side observed domain (see Fig. 3).
86
Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90
the dominant movement of the recirculation bubble. The flow is not
receptive to the actuation frequency any more, and the main frequency of
the recirculation bubble seems reminiscent of one of the ”natural” flow
frequencies depicted in Fig. 9. As analyzed in Section 3.1, one can infer
that the eddies produced by the actuation are too small to introduce
enough momentum in the flow and lock the separation mechanism to the
actuation frequency. Another explanation, as observed in Glezer et al.
(2005), is that the control become frequency independent when the
actuation frequency exceeds the “receptive band” of the unactuated flow.
Nevertheless, the energy introduced by this actuation is enough to
reduce, although only to some extent, the recirculation bubble. In
contrast, the first three actuated cases, Fig. 11b, c and d, have a clear
frequency locking behaviour. In these configurations the separation
shows the periodic formation of a train of structures that moves down-
stream with the actuation frequency. The configuration actuated at F þ ¼
3:1 (Fig. 11d) presents the formation of the smallest periodic structures.
The eddies visualized in this latter case dissipate very early downstream
reducing to the minimum the recirculation bubble.
Fig. 13 shows the space-distribution of the energy of the actuation
frequency of the first three actuated cases. When the flow is actuated
using frequency F þ ¼ 1, a large part of the observed domain is affected by
the control. The large structures visualized in Fig. 11b influence a large
part of the observed domain. As confirmed by Figs. 11b and 13a shows
how the large fluctuations that are introduced evolve downstream,
reaching the wake of the model. The phase averaged plot of the wake
velocity, Fig. 14b, reveals a coherent motion of the wake at the actuation
frequency. In this case the wake flow is receptive to the actuation fre-
quency. For the actuated cases F þ ¼ 2:1 and F þ ¼ 3:1 there is no evi-
dence of organized phase-coherent motion of the wake, exhibiting no
G. Minelli et al. Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Fig. 11. Sequence of phase averaged velocity streamlines during one cycle of the actuation. One actuation cycle (a); the red circle indicates the position of the phase average. Actuated flow
at F þ ¼ 1 (b). Actuated flow at F þ ¼ 2:1 (c). Actuated flow at F þ ¼ 3:1 (d). Actuated flow at F þ ¼ 6:2 (e). Side observed domain (see Fig. 3). (For interpretation of the references to colour
in this figure legend, the reader is referred to the web version of this article.)

Fig. 12. Sequence of phase averaged velocity streamlines during one cycle of the actuation. Actuated flow at F þ ¼ 6:2. The phase average is done over the frequency F þ ¼ 0:75. Side
observed domain (see Fig. 3).

Fig. 13. Spatial distribution of the energy of the frequency content of the wall normal velocity. Spatial distribution of the energy of the actuation frequency for case F þ ¼ 0:8 (a). Spatial
distribution of the energy of the actuation frequency for case F þ ¼ 2:1 (b). Spatial distribution of the energy of the actuation frequency for case F þ ¼ 3:1 (c). The values are normalized by
the maximum energy value of the space-averaged spectrum of the unactuated case. Side observed domain (see Fig. 3).

coupling between the separated shear layer and the wake flow, Fig. 14c
and d. This is in agreement with Fig. 13b and c. These two plots show the
presence of the actuation frequency only in a limited region of the
observed domain. A very regular train of structures is observed (Fig. 11c)
in case F þ ¼ 2:1, and the energy of its frequency is concentrated in the
shear layer of the separation bubble, Fig. 13b. On the other hand, case
F þ ¼ 3:1 presents only a relatively high energy level of the actuation
frequency, contained in a small region close to the separation point,
Fig. 13c. In particular, case F þ ¼ 3:1 shows that the eddies generated by
the wall normal disturbances remain small and dissipate earlier in
comparison to all actuated cases. Overall, the FFT analysis and phase-
averaging show the presence of a ”receptive band” of frequencies that
are able to amplify an unsteady component of the unactuated flow. What
was observed in Glezer et al. (2005), is that, above a certain frequency
limit, the effect of the actuation become frequency independent. The
same trend is observed in the present work. When the actuation fre-
quency is chosen to be in the receptive band, the flow separation is locked
to the actuation frequency. This allows the formation of structures
(which define the dimension of the recirculation bubble) that scale with
the wavelength of the actuation. Choosing a frequency higher than the
upper frequency of the receptive band, the flow separation decouples
from the imposed control, and no presence of the actuation frequency is
observed in the separated flow. Moreover, the main frequency charac-
terizing the motion of the recirculation bubble seems to appear at the
“natural” main frequency, Fig. 9. This frequency independent behaviour
is corroborated by previous findings. In Wu et al. (1998) the investigators

87
G. Minelli et al. Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Fig. 14. Sequence of phase averaged velocity streamlines during one cycle of the actuation. One actuation cycle (a); the red circle indicates the position of the phase average. Actuated flow
at F þ ¼ 1 (b). Actuated flow at F þ ¼ 2:1 (c). Actuated flow at F þ ¼ 3:1 (d). Rear observed domain (see Fig. 3). (For interpretation of the references to colour in this figure legend, the reader
is referred to the web version of this article.)

found that the response shedding frequency of the near wake becomes
close to the natural shedding frequency of the flow when a frequency, out
of the receptive band, was employed. The independence from the actu-
ation frequency does not drastically compromise the effect of the actu-
ation; the recirculation bubble reduces, and, presumably, a higher
momentum coefficient Cμ is necessary to observe the same results as
obtained for case F þ ¼ 2:1 and F þ ¼ 3:1. This last inference is deducted
considering the results described in Glezer et al. (2005). The in-
vestigators observed a frequency independent behaviour and a
completely attached flow over a stalled aerofoil when the actuation
frequency was chosen one order of magnitude higher than the receptive
band of frequencies. However, the Cμ was one order of magnitude higher
as compared to the present work.
Fig. 15 compares the averaged side and wake flow of the unactuated
case against actuated at F þ ¼ 3:1 and F þ ¼ 1. The flow is drastically
manipulated and the recirculation bubble is nearly suppressed using
F þ ¼ 3:1. Note that the average streamlines of the recirculation bubble
are observed to be significantly different for the mentioned actuations.
Fig. 16 shows the streamwise velocity profiles in the wake of different
cases. The actuation at frequencies F þ ¼ 2:1 and F þ ¼ 3:1 gives a very
similar wake configuration, while actuation F þ ¼ 1 shapes the wake flow
quite differently. The higher absolute value of the streamwise velocity
(Fig. 16) and the elongated recirculation bubble (Figs. 15 and 17) of cases
F þ ¼ 2:1 and F þ ¼ 3:1 suggest the presence of a lower base pressure
region and a higher drag contribution as a direct consequence. The case
actuated at F þ ¼ 1 instead, presents the lowest absolute streamwise ve-
locity (Fig. 16) and the shortest wake (Figs. 15 and 17). This fact makes
us aware that a complete flow control study should integrate both a front
separation treatment and a rear separation treatment, to optimize the
aerodynamic performance of the model.
4. Conclusion
The flow around a truck cabin is characterized by pressure induced
separation occurring at different areas. An experimental investigation
conducted at Re ¼ 5  105 was made to study the effect of an embedded
AFC on the natural separation of the flow. The model of a generic truck
cabin was designed to highlight the A-pillar separation mechanism. The
model consists of a simplified truck cabin, characterized by sharp edge
separation on top and bottom edges and pressure induced separation on

Fig. 15. The overall view of the flow field. Contours of the mean streamwise velocity component with superimposed flow streamlines. From left to right: unactuated flow, actuated flow at
F þ ¼ 1, actuated flow at F þ ¼ 3:1.

88
G. Minelli et al.
Fig. 16. Averaged streamwise velocity at different locations along the recirculation bubble: x1 =W ¼ 1:125, x2 =W ¼ 1:250, x3 =W ¼ 1:375, x4 =W ¼ 1:500. F þ ¼ 1 (stars), F þ ¼ 2:1
(circles), F þ ¼ 3:1 (triangles). The solid line represents the reference unactuated flow.
Fig. 17. Topology of the flow for two different actuation frequencies: F þ ¼ 3:1 (dashed
line) and F þ ¼ 1 (solid line).
the rounded vertical front edges. Four loudspeakers, driven by a variable
frequency signal, were installed and sealed in the model to recreate a
synthetic jet flow at the actuation slot. The actuation was characterized
for various frequencies and voltages of the driving signal by recording the
emitted velocity at the jet orifice. In particular, four actuation frequencies
were studied and compared. An interesting analogy was found between
frequency and dimension of the vortex pair that was generated.
The main frequencies and coherent structures of the unactuated flow
were analyzed by means of POD and FFT, quantifying the ”receptive
band” of the shear layer and near wake structures in a range between
0:7 < F þ < 3:1. Thus, the actuation of the flow was conducted using four
different frequencies: F þ ¼ 1, F þ ¼ 2:1, F þ ¼ 3:1 and F þ ¼ 6:2. All the
actuations were able to deflect the shear layer toward the side surface
and reduce the size of the recirculation bubble. The first three actuation
frequencies have been seen to lock the side separation to the actuation
frequency. The fourth one diverged from this behaviour, corroborating
the ”receptive band” theory, which leads to a frequency independent
actuation (also called by different authors “control by virtual surface
modification”). In the latter case the phase-average study does not show
any lock-in features and the flow tends to behave according to its natural
frequencies. A coherent motion of the wake at the actuation frequency
was observed only for the actuated case F þ ¼ 1. A higher base pressure
and a longer wavelength of coherent structures was observed for cases
F þ ¼ 2:1 and F þ ¼ 3:1 when compared to F þ ¼ 1. This fact suggests to
consider the implementation of an extra AFC at the rear end of the model
to enhance the behaviour of the wake, and have a complete optimization
of the flow.
Acknowledgements
This work is funded by the Swedish Energy Agency (Reference
89
Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90
number: 2013-003608, Project number: 37739-1) and supported by
Volvo Trucks.
References
Ahmed, S.R., Gawthorpe, R.G., Mackrodt, P.A., 1985. Aerodynamics of road and rail
vehicles. Veh. Syst. Dyn. 14, 319–392.
Amitay, M., Cannelle, F., 2006. Evolution of finite span synthetic jets. Phys. Fluids 18.
Amitay, M., Glezer, A., 2002. Role of actuation frequency in controlled flow reattachment
over a stalled airfoil. AIAA J. 40, 209–216.
Amitay, M., Glezer, A., 2002. Controlled transients of flow reattachment over stalled
airfoils. Int. J. Heat Fluid Flow 23, 690–699.
Barros, D., Bor
ee, J., Noack, B.R., Spohn, A., Ruiz, T., 2016. Bluff body drag manipulation
using pulsed jets and Coanda effect. J. Fluid Mech. 805, 422–459.
Ben Chiekh, M., Ferchichi, M., Michard, M., Guellouz, M.S., B
era, J.C., 2013. Synthetic jet
actuation strategies for momentumless trailing edge wake. J. Wind Eng. Ind.
Aerodyn. 113, 59–70.
Brunton, S.L., Noack, B.R., 2015. Closed-loop turbulence control: progress and challenges.
Appl. Mech. Rev. 67.
Cattafesta, L.N., Sheplak, M., 2011. Actuators for active flow control. Annu. Rev. Fluid
Mech. 43, 247–272.
Chang, R.C., Hsiao, F.-B., Shyu, R.-N., 1992. Forcing level effects of internal acoustic
excitation on the improvement of airfoil performance. J. Aircr. 29, 823–829.
Choi, H., Lee, J., Park, H., 2014. Aerodynamics of heavy vehicles. Annu. Rev. Fluid Mech.
46, 441–468.
Cooper, K., Truck aerodynamics reborn - lessons from the past, SAE Technical Paper
Series (2003).
Gad-el Hak, M., Pollard, A., Bonnet, J.P., 1998. flow control: fundamentals and practices,
number v. 53. In: Flow Control: Fundamentals and Practices. Springer-Verlag.
Glezer, A., 2011. Some aspects of aerodynamic flow control using synthetic-jet actuation.,
Philosophical transactions. Ser. A, Math. Phys. Eng. Sci. 369, 1476–1494.
Glezer, A., Amitay, M., Honohan, A.M., 2005. Aspects of low- and high-frequency
actuation for aerodynamic flow control. AIAA J. 43, 1501–1511.
Hsiao, F.-B., Shyu, J.-Y., Liu, C.-F., 1990. Control of wall-separated flow by internal
acoustic excitation. AIAA J. 28, 1440–1446.
Hsiao, F.-B., Shyu, R.-N., Chang, R.C., 1994. High angle-of-attack airfoil performance
improvement by internal acoustic excitation. AIAA J. 32, 655–657.
Huang, L., Maestrello, L., Bryant, T., 1987. Separation control over an airfoil at high
angles of attack by soundemanating from the surface. In: 19th AIAA, Fluid Dynamics,
Plasma Dynamics, and Lasers Conference, Fluid Dynamics and Co-located
Conferences. American Institute of Aeronautics and Astronautics.
€
Krajnovi
c, S., Osth, J., Basara, B., 2010. LES study of breakdown control of A-pillar vortex.
Int. J. Flow Control 2, 237–258.
Lumley, J.L., 1970. Stochastic Tools in Turbulence - Applied Mathematics and Mechanics.
Academic Press, New York.
McNally, J., Fernandez, E., Robertson, G., Kumar, R., Taira, K., Alvi, F., Yamaguchi, Y.,
Murayama, K., 2015. Drag reduction on a flat-back ground vehicle with active flow
control. J. Wind Eng. Ind. Aerodyn. 145, 292–303.
Minelli, G., Krajnovi
c, S., Basara, B., Noack, B.R., 2016. Numerical investigation of active
flow control around a generic truck a-pillar, flow. Turbul. Combust. 97, 1235–1254.
Modi, V.J., Hill, S.S., Yokomizo, T., 1995. Drag reduction of trucks through boundary-
layer control. J. Wind Eng. Ind. Aerodyn. 54–55, 583–594.
Rempfer, D., Fasel, H.F., 1994. Evolution of three-dimensional coherent structures in a
flat-plate boundary layer. J. Fluid Mech. 260, 351.
Seifert, A., Darabi, A., Wyganski, I., 1996. Delay of airfoil stall by periodic excitation.
J. Aircr. 33, 691–698.
Seifert, A., Shtendel, T., Dolgopyat, D., 2015. From lab to full scale Active Flow Control
drag reduction : how to bridge the gap ? J. Wind Eng. Ind. Aerodyn. 147, 262–272.
G. Minelli et al. Journal of Wind Engineering & Industrial Aerodynamics 168 (2017) 81–90

Smith, D.R., 2002. Interaction of a synthetic jet with a cross ow boundary layer. AIAA J.
40, 2277–2288.
Smith, B.L., Glezer, A., 1998. The formation and evolution of synthetic jets. Phys. Fluids
10, 2281–2297 (1994-present).
Unal, M.F., Rockwell, D., 1988. On vortex formation from a cylinder. Part 1. The initial
instability. J. Fluid Mech. 190, 491–512.
€
Vernet, J.A., Orlü, R., Alfredsson, P.H., 2015. Separation control by means of plasma
actuation on a half cylinder approached by a turbulent boundary layer. J. Wind Eng.
Ind. Aerodyn. 145, 318–326.
Wu, J.-Z., Lu, X.-Y., Denny, A.G., Fan, M., Wu, J.-M., 1998. Post-stall flow control on an
airfoil by local unsteady forcing. J. Fluid Mech. 371, 21–58.

90