1

1
 Coupling mechanical rotation and EHD actuation
in flow past a cylinder
A. Gronskis, J. D’Adamo, G. Artana, A. Camillieri, J.H. Silvestrini

Abstract— Plasma actuators have shown considerable promise as perfomance codes as DNS have been reported in the rel-
flow control devices in different applications. Models for the plasma- ative simple cases of classical coronas discharges [1]. For
induced body force have been derived extensively in the last years.
There is still, however, a lack of clear understanding of the effect of the specific case of the actuators named dielectric barrier
these actuators on the bulk flow. To this end, flow simulations were devices (DBD) different efforts have been reported by dif-
recently performed by some authors with the electro-hydrodynamic ferent researchers [2–4].
(EHD) effects being modelled by a body force in the equations of mo-
tion. The introduction of this force in high performance numerical
Corke et al. (2005) derived a simplified body force dis-
simulation may pose some problems as a consequence of the large dif- tribution and implemented it in the numerical flow solver
ferences between time or spatial scales of the physics of the discharge CFL3D to estimate the aerodynamic coefficients involved
and those of interest for the fluid flow description. In this paper, in the control of leading-edge separation on wing sections.
we try to simplify this problem by looking for a correspondence be-
tween the global flow modifications produced from an EHD actuation Miles et al. (2006) developed a detailed physical model
and those from a moving body surface. Direct numerical simulations for DBD in air, and noted the computational difficulties
(DNS) of flow past a rotating cylinder are performed. The Reynolds of plasma numerical simulation in large scales. It is con-
number based on the cylinder diameter and free-stream velocity of
the flow is 125 (corresponding to the wake-transition regime). Per-
cluded that in general the cell size for the rectangular grid
formed simulations allow to make the comparison with particle im- has to be chosen according to the characteristic plasma and
age velocimetry (PIV) measures of dielectric barrier discharge (DBD) voltage scales and geometrical sizes (in the order of 10 mi-
plasma actuator flow. The results obtained from the projection of crons), that in turns requires an extremely high number
mean velocity fields show that the time-averaged vortical structures
in the two cases are comparable, and seem to indicate the existence of computational points in order to perform calculations
of a critical frequency actuation between 810 and 816.3 Hz for which in large scale geometry (i.e., beyond the domain which is
vortex shedding ceases for the actuator configuration we have chosen next to the electrode configuration). The other difficulty
for this study.
in kinetic simulations is the choice of time step. Because
the time step should allow the resolution of plasma for-
I. Introduction mation and sheath dynamics time (which is of the order of
nanoseconds), this time scale is several orders of magnitude
Electro-hydrodynamic (EHD) actuators for flow control
smaller than the time during which the plasma is producing
with aims to improve aerodynamic design have been re-
an effect on the surrounding air. Therefore both factors be-
ceiving special attention in the last years. Through the
come prohibitive the approach of plasma effects in a large
ionization of flowing air close to the surface of the body,
spatio-temporal scale. A more recently computational ex-
EHD devices produce a modification of the condition of
ploration has been conducted by Visbal et al. (2006) using
the flow at the wall. The charged particles present in
large-eddy simulations approaches at Re ∼ 104 to capture
the ionized air (plasma) experience an electric force that
the effects of the local forcing of the DBD device on tur-
is transmitted through particle collision to the neutrals of
bulent separated flows. They observed that further work
the flowing air. The expression of this force, depends on
is still needed in order to construct detailed models for the
the discharge characteristics and in some cases there can
spatio-temporal distribution of the plasma-induced body
exist a two-way interaction created by electric forces as
forces suitable for incorporation into viscous flow simula-
they elicit a mechanical response that in turns alter the
tions.
fields. In general electromagnetic wave phenomena consid-
We examine the global effect produced by this forcing by
ered have transit times short compared to times of interest
adding an equivalent tangential velocity on the surface of
of the fluid media where it propagates. Implementation of
the body. From this point of view we consider the possible
this force in numerical flow simulations could be used in
paralelism of the global effects produced by a DBD actua-
optimizing the design and placement of plasma actuators
tion and the ones of a moving surface. We concentrate our
for different applications for flow control. Detailed stud-
analysis to the case of disposing actuators on the surface of
ies on how to implement the electric force coupling in high
a cylinder in such a way that its action can be mimethysed
A. Gronskis is with the University of Buenos Aires
to the rotation of a cylinder about its axis. The present
(aleg@fi.uba.ar). work aims to test the adequacy of this approach and to
J. D’Adamo is with the University of Buenos Aires find the correspondence between the cylinder rotational pa-
(jdadamo@fi.uba.ar).
G. Artana is with the University of Buenos Aires and CONICET rameter and the intensity of the EHD actuation. To this
(gartana@fi.uba.ar). end, direct numerical simulations (DNS) of flow past a ro-
A. Camillieri is with the University of Buenos Aires tating cylinder placed in a uniform stream were performed
(cada@fi.uba.ar).
J.H. Silvestrini is with the Pontificia Universidade Catolica do Rio for various tangential velocities using the virtual boundary
Grande do Sul, Brasil (jorgehs@pucrs.br) technique to model the moving surface. The mean veloc-
ity fields obtained in this way were compared with those 4
Current (mA)
coming from the application of the PIV measures of DBD 3 Voltage (kV)

plasma actuator flow. 2

1
A. Experimental setup 0

A.1 Wind tunnel and cylinder −1

−2
The experimental configuration we settled for this work
−3
consists of a flow around a circular cylinder at low Reynolds
−4
number, Re = 125. The velocity field measurements were
−5
undertaken with the cylinder placed in a closed loop wind
tunnel with a test section of 18x18 cm2 . In our study the −6
0 0.5 1 1.5 2
time (ms)
EHD actuators were mounted on the surface of a poly-
methyl methacrylate cylinder that were 20 mm in diame- Fig. 2. Calculated current and applied voltage as a function of time
ter D. The cylinder was a hollow tube with a 4 mm thick in the geometry of Fig. 1.
wall. The asymmetric DBD plasma actuator used in our
configuration consisted of four pairs of flat aluminium foils
A.2 Velocity field measurements
electrodes disposed in the spanwise direction of the cylinder
surface, in an arrangement as shown in Figure 1. For each The velocity field was measured using a PIV technique.
pair, one of the electrodes was exposed to the air while the These experiments were conducted using a camera with a
other, separated by the dielectric material, was grounded resolution of 640x420 pixels, 1/100 s of time between im-
and was not in contact with the flow. Both electrodes over- ages. A green laser of 150 mW, combined with a rotating
lap by a small amount (of the order of 0.5 to 1 mm) and polyhedric mirror, was used to provide illumination on the
in order to avoid spurious forcing from parts of the air test section.
exposed electrode non coincident with the active border The algorithms used in this work belongs to GPIV soft-
we passivated these parts with teflon tape. By doing this ware [8] which is under GNU General Public License. We
we assured that the discharge was produced as indicated have considered for our images a two step grid refinement,
in Figure 1. The EHD forcing was produced by stressing so the final interrogation size is 16x16 pixels with a 50%
the air exposed electrode with a pulse of decreasing ampli- overlapping.
tude (see Figure 2). The maximum voltage of this pulse The time resolution was suitable to recover the flow dy-
was regulated in such a way that only in the first cycle was namics, in a way that the vortex shedding frequencies are
produced the discarge. The maximum of the peak was kept lower than the acquisition frequency.
constant in our experiments but the frequency of repetition
B. Numerical methodology
of the pulses was varied in the range [0-1000 Hz].
By undertaking this kind of excitation the plasma actuator B.1 General characteristics of Incompact3d code
operation enabled: The Incompact3d code was used to perform the DNS
- excitations not extremely high for the Reynolds num- studies [9–12, 14]. This code solves the continuity and mo-
ber in which the DNS code was validated and mentum equations. These equations are discretized using
- almost ”steady” excitations, as the input driving fre- a finite difference method on a uniform cartesian grid.
quency was well above the fluid response frequency (taking Spatial discretization is performed with a sixth-order
into account that the Strouhal number St ∼ 0.2 and the compact centered finite difference scheme, while temporal
characterisitc length D=0.2 m, the vortex shedding fre- discretization is performed with a third-order low-storage
quency is ∼1 Hz). Runge-Kutta method. The boundary condition on the
cylinder surface is imposed using the virtual boundary tech-
nique.

B.2 Governing equations

Mass and momentum conservation principles are repre-
sented by the Navier-Stokes equations, which have the fol-
lowing form for an incompressible fluid
~ · ~u = 0
∇ (1)
∂~u
= −∇p~ −ω ~ × ~u + ν∇2 ~u + f~ (2)
∂t
where ν is the kinematic viscosity, p(~x, t) the dynamic
pressure field (P + 12 ρ|~u|2 ), ~u(~x, t) the velocity field and
Fig. 1. Schematic of the EHD actuator ω ~ × ~u). The external force field
~ (~x, t) the vorticity field (∇
f~(~x, t) is used here to generate the virtual cylinder.
B.3 Boundary conditions and Parameters equal to one in a way that kuj k = (uj , uj )1/2 = 1.
Figure 3 shows a schematic view of the boundary con- PIV data is also represented by mean flow fields, ui , cor-
ditions over the inflow, lateral and outflow regions of the responding to each value of DBD frequency actuation fi .
domain. The rotation of the cylinder about its axis in the The projection of these fields over each element of Γ give
counterclockwise direction is controlled by specifying the us a parameter pij = (ui , uj ) which allow us to compare
velocity at the surface boundary points as EHD and mechanical actuation. We established the corre-
spondence where pij reach its maximum value.
~v = (ux cb , uy cb ) = (−vt sinθ, vt cosθ) (3)
D. Results and discussion
θ = arctg (yb /xb ) (4)
For the analyzed range of frequencies of pulse repetition
where vt = α U∞ is the tangencial velocity of the cylinder (0 to 892 Hz) the flow either achieves a steady state or
and (xb , yb ) the coordinates of the boundary points of the reaches a temporally periodic solution. Representations
cylinder with respect to its axis. of the experimental mean flow streamlines are shown in
The simulations were performed using 685x433 grid Figure 4.
points in x and y directions respectively, the domain size
was Lx = 19D, Ly = 12D and the localization of the cylin- 0 Hz
der center was Xc = 8D, Yc = 6D. The grid resolution
was 36 points per diameter. These parameters are based −3

on the validation for the case Re = 200 [9, 14]. −2

−1

y/D
0

3
0 2 4 6 8
x/D
812 Hz

−3

−2

−1
y/D

Fig. 3. Schematic of the boundary conditions 0

C. Coupling mechanical rotation and EHD action 2

As it was disposed, the set of EHD actuators introduce
3
modifications on the flow in a way that is similar to a cylin-
0 2 4 6 8
der rotation. In flow around cylinders the usual compari- x/D
son between experiments and simulations is performed with 892 Hz
Strouhal number. However this coefficient is not convenient
in our analysis as it is not sensitive enough to discriminate −3

between different intensities of rotation. For this reason −2

to compare PIV results from experiments and data issued
from simulation, we perform a projection based on an inner −1
product. Assuming that our flows confined in a region Ω
y/D

0
can be represented in Hilbert spaces H, the product (u, v)
between two elements of H is defined by: 1
Z
2
(u, v) = u(x)v(x)dx (5)
Ω 3

We can identify each flow by its mean flow field u. There- 0 2 4 6 8

x/D
fore, from DNS simulation we construct a functions ensem-
ble Γ with each mean flow field uj corresponding to each Fig. 4. Mean flow streamlines from PIV measures of DBD plasma
value of αj parameter. Every element of Γ has a norm actuator flow at various frequencies of pulse repetition.
The usual Karman vortex street is observed for low values 4
power (watt*10)
of actuations f ≤ fL , where 810 Hz < fL < 816.3 Hz. 3.5 α
equiv
A pair of vortices is shed alternately during each cycle of
shedding and the flow achieves a temporally periodic state. 3

The actuation introduces asymmetry in the strength and 2.5

location of the positive and negative vortices. An increase
in actuation is accompanied by an increase in the upward 2

deflection of the wake and a reduction in its lateral width. 1.5

At a frequency close to 815 Hz the vortex shedding ceases
and the flow achieves a steady state. 1

For the same Reynolds of our experiments (Re=125) flows 0.5

past a cylinder with various α values between 0 and 4 were
computed. All the flow patterns and vortical structures 0
0 200 400 600 800 1000
in mean flows from the two sources are in good agree- frequency (Hz)

ment with each other. Even though in our simulations Fig. 6. Equivalent values of α for the different DBD actuations
the Strouhal numbers are well reproduced [9], shorter re-
circulation bubbles were however determined and we have
associated this result to wall effects not taken into account Further investigations are necessary in order to improve
in our boundary conditions. Even though these small dif- this correspondence by refinements on numerical simula-
ferences we take advantage of the similarities of pattern tions. The correspondance has also to be validated in other
flows to determine the equivalent value of α for the differ- situations like those in which actuation does not involve the
ent DBD actuations. Figure 5 shows the result of the pro- whole surface of the body considered. Another important
jection described above. Based on the projection criteria, remaining issue is the analysis of the incidence of three di-
we could construct the Figure 6 that enables to determine mensional effects that may be present in the wake dynamics
these values for the different cases analysed. The analysis in other regimes.
of these figures reveals that the stronger correlations could
be achieved in the steady cases and that the increase in the References
value of equivalent rotation parameter increases in a quite [1] A. Soldati and S. Banerjee. Turbulence modification by large-
scale organized electrohydrodynamic flows. Phys. Fluids 10(7),
similar way than the electrical power of the discharge. 1742-1750, (1998).
[2] T.C. Corke and M.L. Post. Overview of plasma flow control:
0.75 concepts, optimization and applications. AIAA Paper 2005-0563,
0 Hz 2005.
505 Hz [3] J.P. Boeuf and L.C. Pitchford. Electrohydrodynamic force and
680 Hz aerodynamic flow acceleration in surface dielectric barrier dis-
0.7 727 Hz charge. J. Appl. Phys. 97, 103307 (2005).
778 Hz
[4] G.I. Font and W.L. Morgan. Plasma discharges in atmospheric
812 Hz
857 Hz
pressure oxygen for boundary layer separation control. AIAA Pa-
0.65 892 Hz per 4632, 2005.
[5] A.V. Likhanskii, M.N. Shneider, S.O. Macheret and R.B. Miles.
Pij/Ni

Modeling of interaction between weakly ionized near-surface plas-

mas and gas flow. AIAA Paper 2006-1204, 2006.
0.6 [6] M.R. Visbal, D.V. Gaitonde and S. Roy. Control of transitional
and turbulent flows using plasma-based actuators. AIAA Paper
2006-3230, 2006.
0.55
[7] J.P. Boeuf, Y. Lagmich, Th. Callegari and L.C. Pitchford. Elec-
trohydrodynamic force and acceleration in surface discharges.
AIAA Paper 2006-3574, 2006.
[8] G. van der Graaf, GPIV software, http://gpiv.sourceforge.net
0.5 [9] A. Gronskis, J. D’Adamo, G. Artana and J. Silvestrini. Direct
0 0.5 1 1.5 2 2.5 3 3.5 4 numerical simulation of flow past a rotating cylinder. In Anais da
α
5a EPTT 2006 - ABCM, Rio de Janeiro, Brasil, Set. 25-30, 2006.
[10] Lamballais, E., Silvestrini, J., 2002. Direct numerical simulation
Fig. 5. Normalized projection of PIV-DBD mean fields over DNS of interactions between a mixing layer and a wake around a cylin-
functions ensemble Γ der. J. Turbul. 3, 028.
[11] Silvestrini, J.H., Lamballais, E., 2002. Direct numerical simula-
tion of wakes with virtual cylinders. Int. J. Comp. Fluid Dynam.
II. Conclusion 16 (4), 305-314.
[12] Ribeiro, P. Desprendimento de vortices e controle em esteira de
In this work, it is shown that the mechanism of cylinder cilindros por simulacao numerica direta. Master’s thesis, Instituto
rotation allow us to establish a correspondence between the de Pesquisas Hidraulicas - IPH/UFRGS, 2002.
[13] J. Silvestrini and E. Lamballais. Direct numerical simulation of
wake modifications from an EHD actuation and those from oblique vortex shedding from a cylinder in shear flow. In Proc.
a moving body surface. 3rd International Symposium on Turbulence and Shear Flow Phe-
The mean structural features of this complex flows ob- nomena, Sendai, Japan, 2003.
[14] Vitola, M., 2006. Influencia de um contorno plano sobre o de-
served experimentally through PIV measures of DBD sprendimento de vortices ao redor de um cilindro circular. PhD
plasma actuator flow are satisfactorily recovered numeri- thesis, Instituto de Pesquisas Hidraulicas - IPH/UFRGS, 2006.
cally through DNS of flow past a rotating cylinder.