Accepted Manuscript

Flow past a rapidly rotating elliptic cylinder

H. Lu, K.B. Lua, T.T. Lim

PII: S0997-7546(17)30592-7
DOI: https://doi.org/10.1016/j.euromechflu.2018.08.011
Reference: EJMFLU 3350

To appear in: European Journal of Mechanics / B Fluids

Received date : 20 October 2017

Revised date : 24 August 2018
Accepted date : 26 August 2018

Please cite this article as: H. Lu, K.B. Lua, T.T. Lim, Flow past a rapidly rotating elliptic cylinder,
European Journal of Mechanics / B Fluids (2018),
https://doi.org/10.1016/j.euromechflu.2018.08.011

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to
our customers we are providing this early version of the manuscript. The manuscript will undergo
copyediting, typesetting, and review of the resulting proof before it is published in its final form.
Please note that during the production process errors may be discovered which could affect the
content, and all legal disclaimers that apply to the journal pertain.
Flow past a rapidly rotating elliptic cylinder
H. Lu1, K. B. Lua2,a), and T. T. Lim1
1
Department of Mechanical Engineering, National University of Singapore, 9 Engineering Drive 1, Block EA 07-08,
117576, Singapore
2
Department of Mechanical Engineering, National Chiao Tung University, 1001 Ta Hsueh Road, Hsinchu, Taiwan
30010, ROC

Abstract
Previous studies have shown that the characteristics of flow over a circular cylinder can be
significantly altered by introducing rotation and ellipticity. While the effect of the former has
been extensively studied; the influence of the latter is less explored. In this paper, we report
results of a numerical investigation on the flow past a two-dimensional (2D) elliptic cylinder that
is rapidly rotating at a Reynolds number of 200, covering a narrow range of high rotational
velocity where low-frequency Mode II shedding occurs for a circular cylinder. Results show that
aerodynamic forces and moments acting on a rotating cylinder are highly nonlinear with respect
to variation in rotational velocity and cylinder thickness. Various wake topologies are revealed,
including steady, quasi-steady, periodic and aperiodic flows. At an intermediate rotation rate, a
“hovering vortex” is typically formed on one side of the cylinder, and at high rotation rates,
multiple vortices are generated and rotated with the cylinder. These rotating vortices are highly
related to a plateau in the mean lift over a certain range of cylinder thickness, as well as a sudden
force transition from a mean drag to a considerable mean thrust when the thickness exceeds a
certain threshold value.

1. Introduction
 
Flow past a cylinder is a classical problem of bluff body aerodynamics. Research on this problem
dated back to Bénard (1908) and von Kármán (1911), and extensive studies have been conducted
since then (see the review articles by Williamson 1996 and Zdravkovich 1997). It is well known
that a vortex street is generated in the wake of a stationary cylinder when Reynolds number ( ,

                                                               
a) Author to whom correspondence should be addressed. Electronic mail: engp4324@nctu.edu.tw

1
 
defined as / , where , and are freestream velocity, cylinder diameter and kinematic
viscosity of fluid, respectively) exceeds a critical value of 46 (Sreenivasan et al., 1987), and a
Hopf bifurcation is triggered by flow instability. However, these flow features can be
significantly altered if the cylinder is subjected to a rotary motion about its center (Tokumaru and
Dimotakis, 1991), which serves as a useful application for active flow control.

In the past three decades, numerous studies on a rotating circular cylinder in a uniform cross
flow have been conducted (see for example, Chew et al. 1995, Kang et al. 1999, Stojkovi et al.
2002, Mittal and Kumar 2003, just to name a few). These studies show that the wake structure of
a rotating circular cylinder is highly dependent on a normalized rotational velocity ratio
(defined as /2 , where is the rotational velocity). At = 200, numerical results of
Mittal and Kumar (2003) show that the vortex street becomes more deflected as is increased,
and vortex shedding ceases completely when exceeds 1.9. In addition, there is another narrow
shedding zone (4.4 4.8 according to Mittal and Kumar 2003 and 4.8 5.15
according to Stojkovi et al 2002), where a low-frequency, one-sided shedding occurs. This is
referred to as Mode II shedding by Rao et al. (2015) and has been confirmed in a recent
experiment by Kumar et al. (2011).

Apart from rotation, geometric asymmetry can also alter the characteristics of cylinder flow. A
typical example is the flattening of a cylinder from circular to elliptic geometry. Modi and
Dikshit (1975) found that the introduction of eccentricity would lead to a rapid variation in the
spacing of vortices in the cylinder wake. Such variation is dependent on the angle of attack (AoA)
and thickness ratio (i.e. minor-major axis ratio) of the elliptic cylinder as is shown numerically
by Badr and Kocabiyik (1997) and theoretically by Chandna (1997). Recently, Thompson et al.
(2014a) numerically investigated the flow past a stationary elliptic cylinder with 90° AoA and
varying from 0 (a flat plate) to 1 (a circular cylinder) at = 31.6 ~ 47.2. They found that a
reduction in leads to the generation of a secondary von Kármán-like wake of longer
wavelength on top of the primary wake.

The above studies focus primarily on a stationary elliptic cylinder. Thus, a question arises: What
if geometric asymmetry through ellipticity is combined with a rotational motion? Compared to a

2
 
rotating circular cylinder, substantially fewer studies have been conducted on flow over a
rotating elliptic cylinder. Following the pioneering work by Lugt and Ohring (1977) and Iversen
(1979), Lua et al. (2010) conducted digital particle image velocimetry (DPIV) measurements on
a rotating elliptic cylinder with = 0.125 and = 0.417 ~ 2.5. Their results show that when >
1 (i.e. tip velocity of the cylinder exceeds the freestream velocity), a “hovering vortex” is
generated on one side of the cylinder for at least a few half-cycles of rotation. This hovering
vortex was also found in subsequent numerical studies by Hu (2013) and Naik et al. (2017, 2018).
In their simulation of two-dimensional (2D) rotating elliptic cylinder with = 0.5, 1.0 and 1.5
and = 0.1 ~ 1.0 at = 100, Naik et al. (2017) also observed merging of multiple trailing edge
vortices in the wake, which may affect the circulation of the hovering vortex. In a follow-up of
their earlier experiments (Lua et al. 2010), Lua et al. (2018) recently conducted similar numerical
studies with = 0.0625 ~ 1.0 at = 0 ~ 2.5. Their results show that reducing cylinder thickness
leads to a reduction in lift and a non-monotonic change in drag. Interestingly, they also found
that due to the competition between a prevailing negative pressure on one side of the elliptic
cylinder and a suction force from the hovering vortex on the opposite side, drag reaches a local
minimum at about = 0.375, and becomes negative at = 2.5; in other words, thrust is
generated. Although earlier studies have contributed much to our knowledge of the flow
characteristics of a rotating elliptic cylinder in cross flow, it remains unclear as to how variations
in the thickness ratio for values higher than 2.5 affect the flow topology and lift and drag
characteristics, especially in the Mode II shedding range mentioned earlier. The desire to address
this issue prompted us to extend the numerical study by Lua et al. (2018) to values above 2.5
for a range of ellipticity. Apart from its intrinsic interest, the findings from the present study may
be beneficial to the understanding of flow control and autorotating leaves/seeds (McCutchen
1977, Alexander 2004).

2. Methodology
 
The present numerical work investigates the flow past a rotating elliptic cylinder, which rotates
impulsively with a high angular velocity (see Fig. 1). For the cylinder, the thickness ratio is
defined as the ratio of the semi-minor axis and the semi-major axis . Throughout the study,
ranges from 0.0625 to 1.0; equivalently, the shape of the cylinder changes from a thin eclipse to

3
 
a circle. The veloccity ratio , on the othher hand, is defined aas / ( is the ffreestream
velocityy). In the preesent study, four mely 3.0, 4.00, 4.6 and 5.0, bearing
valuues are conssidered, nam
in mindd that in our earlier num
merical studyy (see Lua eet al (2018)), a range of
o value off less than
3.0 had already beeen investigaated. At thee present vvalues, a rootating circuular cylinderr typically
reaches a steady-sttate (i.e. tim
me-independdent lift, drrag and flow
w structurees) except aat = 4.6,
where vvortex sheddding with a large waveelength occuurs. The low
west value of = 3.0 is slightly
above tthe upper bbound valuee of 2.5 coonsidered inn our previoous work (Lua
( et al. 2018). In
additionn, as previoously mentiooned in thee introduction, Mittal and Kumarr (2003) repported the
existencce of a seccond sheddding regimee in 4.4 4.8. Therefore,
T tthe case off = 4.6
represennts such a reegime. The Reynolds nnumber is deefined as / , where 2 is the
characteeristic lengtth and is the
t fluid kinnematic visscosity. Forr all the casses considerred in this
paper, is maintaained at 200. The laterral and streeamwise forrces are terrmed as liftt and drag
respectively, their ddirections arre shown inn Fig. 1. Thee correspondding lift andd drag coeffficients (
and ) are determiined using tthe followinng equationss:

1 (1)
2

1 (2)
2
where is the fluidd density.

4
 
FIG. 1. Schematics of the currrent problem m of flow paast a rotatin
ng elliptic cyylinder with
h different
thicknesss ratios. Note that the ccylinder rotaates countercclockwise.

The conntinuity equaation and N

Navier-Stokees equationss are solved by the finitte-volume-bbased CFD
packagee ANSYS ○R F
FLUENT, vversion 16.2 on Natioonal Univerrsity of Sinngapore (NU
US) High
Perform
mance Compputing (HPC
C) platform
m. Since thee current prroblem invoolves rotatioon of wall
boundarries, a rectaangular com
mputationall domain w
with 60L inn x and 40L
L in y (seee x and y
directionns in Fig. 1)) is divided into two zoones, i.e. a rootational zone and a staationary zonne (see Fig.
2). The rotational zone,
z whichh encompassses the cylinnder, is of circular
c shappe with a diiameter of
20L, and the structtured cells iinside this zone
z are preescribed witth a constannt rotationall velocity.
maining unsstructured ccells in the computatioonal domainn constitute the stationnary zone,
The rem
which is connectedd to the rotaational zonee through ooverlapping circular intterfaces. Thhe moving
mesh innside the rottational zonne is also terrmed as Slidding Mesh; essentially,, the rotatioon is taken
into connsideration by an Arbbitrary Lagrrangian-Eulerian (ALE
E) approachh, and the governing
equationns are still ssolved in an inertial fram
me.

5
 
FIG. 2. Computational domain and mesh. is the diameter of the rotational zone.

As for boundary conditions, the cylinder has a no-slip wall, and the top and bottom borders of
the computational domain are symmetric boundaries with zero normal velocity and zero shear.
Also, inflow and outflow conditions are applied at the left and right borders, respectively.

The momentum equations are discretized using a second-order upwind scheme, and a second-
order implicit formulation is chosen for temporal discretization. Moreover, the PISO (Pressure
Implicit with Splitting of Operators) scheme with second-order accuracy is employed for
pressure and velocity coupling. A fixed time step of 0.00125 ( is the rotation period of
the cylinder) is applied.

It should be highlighted that the current computational domain, time step and grid resolution, are
determined by independence tests on the thinnest elliptic cylinder, which are detailed in Lua et al.
(2018). At = 4.6, nevertheless, mode II single-sided vortex shedding may be sensitive to the
lateral far-field boundary. To examine this, the distance from the lateral boundary to the cylinder
is doubled for the case of = 0.875 and = 4.6 (i.e. from 20L to 40L), and the mean ̅ is found
to change from -20.06 to -20.07, whereas the mean ̅ varies from 0.08 to 0.07. Therefore, the
influence of the lateral boundary is negligible.

The present numerical method has also been validated against experimental and numerical
results reported in the literature. As an additional validation case at high values, the flow past a
circular cylinder at = 3.0, 4.0, 4.6 and 5.0 at = 200 is simulated, and the comparison of
mean force coefficients and Strouhal number ( / , is vortex shedding frequency) is
presented in Table 1. In general, the ̅ and values match well with the literature, whereas the
mean drag coefficient ̅ shows a larger discrepancy. Rao et al. (2013) hypothesized that a small
change in the position of a stagnation point in the flow can result in substantial changes in ̅ .
Considering that the absolute change in ̅ is small, the overall comparison in Table 1
demonstrates that the present results are reasonably accurate.

6
 
TABLE. 1. Comparison of mean force coefficients ( , ) and Strouhal number at = 3.0, 4.0, 4.6
and 5.0 at = 200.
Flow state Mittal and Padrino and Rao et al. Present
Kumar (2003) Joseph (2006) (2013)
3.0 ̅ Steady -10.366 -10.340 -10.334 -10.341
̅ 0.035 0.012 0.040 0.062
4.0 ̅ Steady -17.598 -17.582 -17.579 -17.534
̅ -0.055 -0.124 -0.041 -0.017
4.6 ̅ Mode II single- -23.280 -23.238
̅ sided shedding 0.138 0.197
0.021 0.021
5.0 ̅ Steady -27.055 -27.029 -27.114 -27.057
̅ 0.168 0.011 0.217 0.250

3. Results and Discussion

 
This section first discusses the characteristics of time-averaged lift, drag and moment coefficients,
and followed by a summary of wake topologies for different thickness ratios ( ) and velocity
ratios ( ). The final subsection provides a detail analysis on the time-dependent forces and flow
structures around the rotating cylinder, including the flow transition at the highest = 4.6 and
5.0.

3.1 Mean force and moment characteristics

 
To evaluate the overall performance of a rotating cylinder, mean values of the lift, drag and
moment coefficients ( ̅ , ̅ and ̅ ) are determined by time-averaging the data using no less
than 50 cycles of rotation after the flow has reached a fully developed state. The results are
displayed in Fig. 3. In all cases, is defined as /0.5 , and is positive for a
counterclockwise direction. Also, a flat plate of thickness ratio 0.0625 is simulated to ensure that
the thinnest case with = 0.0625 can reasonably approximate a flat plate (see the open symbols

7
 
in Fig. 3). The results in Fig 3 clearly show that variation in ̅ , ̅ and ̅ with respect to
thickness ratio are highly sensitive to velocity ratio . In particular, at the lowest = 3.0, the
magnitude of ̅ increases almost linearly with thickness ratio (see Fig. 3a). This trend is
analogous to < 2.0 cases reported in literatures but with lower producing lower values of ̅ .
At higher , Fig. 3 shows that there is a threshold thickness ratio , beyond which ̅ enters a
near plateau region between and  0.5, where it varies insignificantly. Beyond the plateau
with > 0.5, it resumes a near linear increase. Interestingly, the corresponding ̅ and ̅ (Fig.
3b and 3c) near the threshold thickness unexpectedly undergo a sudden and significant jump
in their respective coefficients for = 4.6 and 5.0. In the case of ̅ , its sign changes from
positive to negative, and in the case of ̅ , it doubles in magnitude. For example, at = 5.0, ̅
changes from 0.42 at = 0.3 to -1.19 at = 0.306, whilst ̅ jumps from -0.76 to -1.87. To the
best of the authors’ knowledge, this phenomenon has not been reported before. At the lower
value of = 4.0, although ̅ varies relatively smoothly around , it still transits from a positive
value before to near zero at and then into a negative value territory after .

In addition to the broad findings mentioned, some specific observations can be made. Firstly, the
plateau region of ̅ decreases as cylinder rotation increases from = 4.0 to 5.0 (see Fig. 3a).
Secondly, the threshold value of thickness ratio increases with , from 0.188 at = 4.0 to 0.3
at = 5.0. Thirdly, both the ̅ versus and ̅ versus curves exhibit a local minimum. The
minima are generally located at = 0.5, except for the case of = 3.0, where it is located at
around = 0.375 instead (see Fig. 3b). It is worth noting that in our previous study with 2.5,
the minimum drag also occurs at = 0.375 (see Lua et al. 2018). The fact that ̅ can dip to
negative value (i.e. below -2.0 at = 5.0) also indicates that a thrust force is generated by the
elliptic cylinder; the magnitude of which is significantly higher than that for a rotating circular
cylinder at = 4.0 (less than -0.1, see Table 1). Lastly, regardless of the value of , ̅ generally
increases with thickness ratio for > 0.3; more steeply for higher cases. However, for < 0.3,
the variation of ̅ with is relatively insensitive to changes in for the three higher cases. In
contrast, the variations of ̅ and ̅ with respect to thickness ratio are more complex. For
instance, ̅ is a non-monotonic function of for = 0.25 but decreases monotonically for =
0.5.

8
 
FIG. 3. T Time-averagged (a) lift, ((b)drag and (c) momentt coefficientss as a functioon of thickn
ness ratio
at differrent velocity ratios . Op pen symbolss denote resu
ults of a flat plate with thickness
t rattio 0.0625.

3.2 Sum
mmary of wake topoologies
 
The largge variationns in the mean forces annd momentss presented earlier indiicates the deependence
of the fuully developped flow strructure of a rotating cyllinder on thhe thickness ratio, especcially near
the threshold thicknness at hiigh velocityy ratios = 4.6 and 5.00. This behaavior can alsso be seen
in Fig. 4, which caategorizes tthe wake paatterns into four broadd flow regim
mes, i.e. steeady flow,
quasi-stteady flow, periodic floow and aperriodic flow. Incidentallly, previouss studies byy Lugt and
Ohring (1977), Luua et al. (20010) and Naik et al. (22017) show
w that for < 2.0, flow
w over an
elliptic cylinder is pperiodic; thhis is in line with the K
Kármán sheddding zone ffor a circulaar cylinder
(Mode I shedding, see Rao et aal. 2015).

To furtther examinne these floow regimess, representtative flow visualizatiion results based on
vorticityy contours aare providedd in Fig. 5. Note that vorticity
v andd time displaayed in the figure are
normalized by anngular frequuency and period off the cylindder rotationn, respectivvely. The

9
 
observaation one caan derive ffrom Fig. 4 is that thee steady floow with tim
me-indepenndent flow
structurees occurs ffor circular cylinder only. On thhe other haand, a slighht “flatteninng” of the
cylinderr from a cirrcular to ann elliptic shhape results in the genneration of ttime-dependdent flow.
Figure 55a displays the steady fflow structuures for all thhe three c
cases considdered here. It
I is worth
noting that
t at higheer , vorticity is confinned primarilyy around the cylinder, w
which correesponds to
closed streamlines
s rreported in past studiess such as Miittal and Kuumar (2003).

FIG
G. 4. Categorries of wake topologies oobserved in tthe present study.

For elliptic cylindeers, the asssociated flow

w behaviorr is quasi-stteady for most
m of the thickness
ratios annd velocity ratios conssidered heree (see Fig. 44). The term
m “quasi-steeady” is useed here to
mean thhat althoughh the flow field within a cycle of rootation is unnsteady, the cycle-averaaged mean
flow rem
mains unchhanged (see Fig. 5b). For
F aperiodiic flow, it iis typically generated through a
combinaation of low
w thickness ratio and high velociity ratio . For this caase, the flow
w structure
varies fr
from cycle tto cycle andd a regular pperiod cannoot be ascerttained. Thiss type of floow can be
seen in F
Fig. 5c wheere snapshotts for the thrree consecuttive cycles are
a presenteed.

10
 
FIG. 5. Typical
T flow
w structures visualized b
by normalizeed vorticity contours
c in different floow regimes.
dy flow; (b) quasi-stead
(a) stead dy flow; (c) aperiodic floow; (d) high
h-frequency periodic floow and (e)
low-freqquency perioodic flow.

As for a periodic flow, it is typically ggenerated unnder two ciircumstancees: 1) thickkness ratio
exceedss the threshoold value , and 2) largge thicknesss ratio at thee higher endd of , inclluding =
0.875 att 4.6 annd = 1.0 at
a = 4.6 (see
( Fig. 4).. Figure 5d and 5e resspectively shhow these
two circcumstances. The first sccenario has a more exteended wake region dow
wnstream compared to
the secoond scenarioo, and it is also associiated with a much higher sheddinng frequenccy. It is of
interest to note thhat Fig. 5dd exhibits two frequeency compponents; onne near thee cylinder
correspoonding to thhe rotational frequency, and the otther one in the
t far wakee with a vallue of 0.2.
The onee-sided vorttex sheddingg depicted iin Fig. 5e tyypically occcurs in a naarrow range 4.4 < <

11
 
4.8 for a circular cyylinder (see Mittal and Kumar, 20003). Howevver, the pressent simulation shows
that onee-sided shedding is sttill present even at higgher value of = 5.00 for = 0..875, thus
indicatinng a slight shift in its upper
u boundd value wheen a small amount
a of eellipticity iss imposed.
When is reduced further from
m 0.875 to 00.75, the flow
w changes tto quasi-steaady regime.

FIG. 6. Fundamental frequenccy of the liift forces ass a function

n of thickneess ratio. Daashed line
denotes the thresholld thickness ratio.

Fundam
mental frequuencies of thhe lift forces (normalizzed by the aangular freqquency of thhe rotating
cylinderr) obtained ffrom fast Foourier transfform (FFT) are plotted in Fig. 6. F
For quasi-steeady flow,
∗
the funddamental freequency equuals to half oof the rotation cycle (i.ee. = 2.0), except for = 0.375
∗
at = 4.0 and 4.66, where = 1.0. Foor most peeriodic or aaperiodic floow, the funndamental
∗
frequenccy is much smaller andd is accompaanied by thee dominant ffrequency component = 2.0 in
∗
the forcces. For insttance, eqquals to 0.2 for = 0.313 at = 4..6; this is coonsistent with the far
∗
wake freequency in Fig. 5d. Alsso, an abruppt change inn is obserrved near thhe thresholdd thickness
ratio . This corressponds to thhe jump in ̅ and ̅ observed
o in F
Fig. 3.

The varrious types of wake strructures disscussed above demonsttrate the coomplexity off the flow
past a rapidly
r rotatting ellipticc cylinder. To better uunderstand the
t mechannisms that leead to the
variationn of forces, especially a sudden juump of meaan drag force near at = 4.6 andd 5.0, it is

12
 
worthwhile to compare the time history of forces for different thickness ratios and see how they
correlate with their corresponding flow structures.

3.3 Time history of forces and near-cylinder flow structure

This subsection is broadly divided into three parts based on different velocity ratio . Each part
is discussed with reference to transient lift and drag forces, instantaneous vorticity contours, etc.

3.3.1 = 3.0
 
For this particular , all the elliptic cylinders possess quasi-steady wake (see Fig. 4). Therefore,
it is not unexpected that the transient lift and drag coefficients have a fundamental period related
to the period of cylinder rotation. Since an ellipse has a symmetrical cross-section about its
major axis, the force signals in fact have a fundamental period equivalent to half of the cylinder
rotation. This is clearly shown in Fig. 7 which displays the plots of lift ( ) and drag ( )
coefficients against the dimensionless time in one cycle of rotation after the flow has attained a
quasi-steady state. Figure 7a clearly show that the overall value increases with the thickness
ratio and converges towards the steady-state value for = 1.0 (circular cylinder). In contrast,
∗
undergoes a sign change close to = 0.25, 0.5, and 0.75, when the major axis of the ellipse is
aligned with the horizontal axis, i.e. the angle of attack is 0° (see Fig. 7b). Also, the maximum
near the instant (i) typically has a smaller magnitude than that of the minimum near instant
(iii). For example, the maximum for = 0.375 (dashed line) is less than 3.0, whereas the
minimum is close to -4.0; this understandably leads to a negative mean drag or equivalently
positive mean thrust depicted in Fig. 3b.

13
 
FIG. 7. Transient (aa) lift and (b b) drag coeffficients for rotating cyllinders with different at = 3.0.
The resu ults are extrracted in one complete rotation
r cyccle after the flow has atttained fully developed
state. Laabels (i) to ((iv) in the fi
figure corresspond to dim mensionless time (i): ∗ = 0.125; (ii): ∗ = 0.25;
∗ ∗
(iii): = 0.375; (iv): = 0.5.

FIG. 8. Instantaneoous vorticity contours att = 3.0 forr = 0.0625, 0.375, 0.6255, 0.875 and d 1.0 (from
left to riight). Labelss (i) to (iv) coorrespond too the same in
nstants labeeled in Fig. 55.

14
 
Selectedd snapshotss of vorticiity contourss corresponnding to instants (i) to (iv) in F
Fig. 5 are
displayeed in Fig. 88. Here, as the cross-ssection of thhe cylinderr transits froom a circlee to a thin
ellipse, the shear laayers aroundd the cylindeer roll up into concentraated vorticees due flow sseparation.
In addittion, a vorteex with negaative vorticity (clockw
wise) stays abbove the ellliptic cylindder, and is
referredd to as a “hhovering vorrtex” (labelled as HV in
i Fig. 8), ffollowing L
Lua et al. (22010), Hu
(2013), Naik et all. (2017) aand Lua et al. (2018). The stabiility of thiss hovering vortex is
maintainned becausee the loss of vorticity ddue to one edge
e of the cylinder sw
weeping aw
way part of
the HV is compenssated by thee gain of neew vorticityy from the oother edge. With a low
w-pressure
center, the
t HV creaates a suctioon force thaat always reesult into ann upward lifft. Dependiing on the
relative position beetween the HV and thhe cylinder in the streaamwise direection, eitheer drag or
thrust ccan be geneerated. Speccifically, whhen the HV
V is upstreaam relative to the cyliinder (e.g.
forward, thuus resulting in thrust geenerated. Inn contrast,
instant i in Fig. 8),, it pulls thee cylinder fo
HV locaated downsttream of thee cylinder (innstant iii) w
would lead too drag generration.

FIG. 9. (a) Pressuree coefficient and (b) veloocity magnittude contours at instantt (iii) for = 3.0. Note
that insttant (iii) corrresponds to that labeled
d in Fig. 5 an
nd 6.

Neverthheless, the ssuction from

m the HV is outweighhed by anotther effect; that is, a pprevailing
negativee pressure bbelow the ccylinder duee to cylindeer rotation. To illustratte this, insttantaneous
contours of pressurre coefficiennt as well ass velocity magnitude
m inn a laboratorry frame at iinstant (iii)
wn in Fig. 9.
are show 9 A negativve pressure region is coonsistently oobserved beelow all the cylinders,

15
 
which iss associatedd with a largge velocity magnitude,, indicating an acceleraated local fllow. Since
such neegative presssure regionn always loocates beloow the cylinnder regarddless of thee cylinder
rotationn, it has an opposite
o effeect to that off the HV, i.ee. a negativee lift and a ddrag/thrust depending
d
on the oorientation oof the cylindder. In addittion, as the nnegative preessure below
w the cylindder is more
profounnd than the low pressuree induced byy the hoveriing vortex aabove the cyylinder (see Fig. 9a,
= 0.06225), the signn of the ovverall lift annd drag/thruust is determ
mined primaarily by thee negative
pressuree region. Too examine thhis behaviorr more quanntitatively, force
f characcteristics on the upper
and low
wer surfaces of the cylinnder correspponding to thhe same insttant (iii) aree presented in
i Fig. 10,
bearing in mind that
t the uppper surfacee is dominaated by thee HV and the lower surface is
dominatted by the prevailing
p n
negative pressure regionn. As the cyylinder is “fflattened”, tthere is an
increasee in and on the uppper surfacee due to thee appearancce of the HV
V. The loweer surface,
on the oother hand, experiencess negative and off larger maggnitudes. Heence, the coompetition
betweenn the HV aand the neggative pressure regionn results inn a negativee total thhat varies
monotonnically withh , and a negative
n whose magnitude peaaks at an inntermediate thickness
ratio.

FIG. 10. Instantaneeous (a) lift and (b) draag coefficien

nts on the uupper, lower and total ssurfaces as
function hat labeled in Fig. 5, 6
ns of at insstant (iii), = 3.0. Note that instantt (iii) corressponds to th
and 7.

The aboove analysiss suggests tthat the vellocity differrence brougght about by
b a rotatingg cylinder
fferent flow behaviors
combineed with the geometric asymmetry from elliptiicity would lead to diff
above aand below the cylinder, which ttypically haave opposite effects oon the oveerall force

16
 
characteristics. This analysis also generally holds for higher . However, as will be discussed
later, other wake patterns also exist at higher , and this can lead to a more complex force and
flow evolutions.

3.3.2 = 4.0
 
As the velocity ratio increases to 4.0, periodic and aperiodic flows start to develop within a
narrow range of thickness ratio (see Fig. 4). To examine how these flow patterns are temporally
established, lift and drag coefficients averaged over each rotational cycle ( ̅ , and ̅ , )
are plotted in Fig. 11. As shown in Fig. 11a and 11d, the coefficients for = 0.0625 and 0.5
converge to a fixed value, corresponding to a quasi-steady wake. For = 0.156 and 0.188 in Fig.
11b and 11c, on the other hand, oscillatory signals are observed in ̅ , and ̅ , . In the
aperiodic case of = 0.156, a periodic waveform cannot be found even after 250 cycles of
simulation, which contrasts to the periodic pattern at the threshold thickness ratio = 0.188.
Moreover, the sign-switching trend of the mean drag near can be seen clearly. For example,
̅ , in Fig. 11b oscillates about a positive mean but converges to a negative value in Fig. 11d;
this is consistent with earlier discussions (see Fig. 2).
 

17
 
FIG. 11.. Mean lift aand drag cooefficients avveraged oveer every rotaational cyclee at = 4.0 for (a) =
0.0625; ((b) = 0.1566; (c) = 0.1188; (d) = 0.5.
0 Note thaat = 0.188 is the thresh hold thickneess ratio.

To furthher investigaate the signn-switching mean drag near , tim

me-dependennt forces (Fiig. 12) are
correlateed with insttantaneous vvorticity coontours (Fig. 13) at fourr instants laabeled as (vv) to (viii).
The neaar-cylinder fflow structuure shown inn Fig. 13 revveals that a hovering voortex does nnot exist at
this highh . Insteadd, the vorticces shed froom the edgees of the elliiptic cylindeer are forcedd to rotate
with thee rapidly spiinning cylinnder, and theey are termeed as rotatinng vortices (RVs) hereaafter. Take
the cylinnder with = 0.188 ass an examplle. At instannts (v) and ((vi), a vorteex is formedd and shed
from thee edge that is moving against the freestream (see Fig. 133v and vi, = 0.188). This new
RV, how
wever, doess not mergee with the previous
p onne to form a hovering vortex. Insttead, both
vorticess rotate withh the cylindeer (see Fig. 13vii and viiii, = 0.1888). The relaative strengtth of these
RVs at ddifferent thiickness ratioos affect thee ̅ cuurve in Fig. 3 significanntly. For insstance, the
cylinderr with = 0.156 has thee largest at instant (vv) amongst the other values (see Fig. 12b),
which ccontributes tto a positivve ̅ (see Fig. 3). Corrrespondinglyy, = 0.1566 appears too have the
weakestt RVs (see F
Fig. 13). Thhe associatedd suction efffect, therefoore, generattes the smalllest thrust

18
 
componnent on the uupper surfacce at instant (v). Hence,, the overalll drag is neggated to a least extent,
resultingg in the highhest peak near that iinstant.

FIG. 122. Comparisoons of transsient (a) liftt and (b) drrag coefficieents among different aat = 4.0.
Labels (v) to (viii)) correspon
nd to dimen me ∗ = 1000.625, 100.775, 100.875 and 101,
nsionless tim
respectivvely.

19
 
FIG. 133. Instantan neous vorticcity contourrs for = 00.0625 to 00.5, = 4.00. Labels (vv) to (viii)
correspoond to the saame instantss labeled in F
Fig. 10.

Vs may also be crucial in the formaation of an aperiodic

The RV a waake. This is shown in F
Fig. 13. At
instant ((v), a newlyy-formed R
RV from thee upper edgge of the cyylinder withh = 0.156 is able to
interact with a previously form
med RV; suuch interacttion might cause variaations in thee vorticity
distributtion in everry half-cyclee, thus causiing the flow
w to becomee aperiodic. For higherr thickness
ratios, nno such interraction occuurs, due to laarger distannce of separaation betweeen the two R
RVs. This
could exxplain why a periodic oor quasi-steaady flow cann be establisshed.

For thicckness ratio above = 00.25, the ellliptic cylindder has a quaasi-steady w
wake (see Fiig. 4), and
the force signals haave a fundam
mental periood that is eqqual to half of a rotatinng cycle (seee Fig. 14).
The oveerall trends of transientt ad arre analogouus to those aat = 3.0 (ccompare Figg. 14 with
Fig. 7): systemaatically incrreases in m
magnitude aas increasees whereas the maxim
mum is
consisteently smalleer in magnitude than thhat of the minimum.
m T underlyying mechaanisms are
The
similar, i.e. suctionn from the vvortex abovve the cylindder and neggative presssure region below the
cylinderr. Since the RVs rotatee with the cyylinder, the suction cann apply to bboth upper and lower

20
 
surfacess, and the ovverall effectt on may be weaker compared tto that brougght about byy the HVs
at lowerr . This coould explainn for existeence of the plateau reggion mentiooned in the mean lift
(approxximately witthin 0.2 < < 0.5, see F
Fig. 3a, m Fig. 13, it can also bee seen that
= 4.0). From
the nearr-cylinder fllow topologgy is similarr in the plateeau region (compare
( = 0.188, 0.225 and 0.5
in Fig. 113). As increases to ooutside the pplateau regioon, no RVs are observed (see = 00.75 in Fig.
13), andd the mean llift resumes an approxim
mately lineaar change w
with the thickkness ratio ((Fig. 3a,
= 4.0).

FIG. 14.. Transient ((a) lift and (b)

( drag coeffficients for rotating cyllinders with different at = 4.0.
The resuults are extrracted in one complete rotation
r cyccle after the flow has atttained fully developed
Labels (v) to (viii) corresp
state. L pond to the same instan
nts labeled inn Figs. 10 an
nd 11.

To furthher show hoow the flow transitions ooccur in thee plateau reggion, a phasse plot of versus
wn in Fig. 15. As
is show incrreases from
m 0.188 to 0.5,
0 the variiation rangee in both and is
clearly reduced, buut whereas the mean
oscillaates aroundd a relatively consistennt mean, w
graduallly decreasess. As channges to 0.75 (beyond tthe plateau region), thhe mean m
magnitude
becomes much largger due to thhe absence oof RVs.

21
 
FIG. 155. Transientt lift coefficcient as a fu
unction of drag coefficcient for rootating cylin
nders with
different at = 4.0.

3.3.3. = 4.6 and 5.0

5
 
As menntioned in suubsection 3.1, one distinnctive featuure associateed with the ttwo highestt values of
is thee sudden juump of ̅ riight after thhe thresholdd thickness ratio (seee Fig. 3). The sign-
switchinng of ̅ is also accom
mpanied witth a changee in the flow
w regime, from quasi--steady to
periodicc (see Fig. 44). To find out
o the phyysical reasonn for these pphenomena,, analysis of the fully
developped flow maay not be adequate;
a a more carefu
ful examinattion into the initial traansience is
necessarry. Figure 16 shows tthis changee at = 5.00 in the foorm of ̅ , and ̅ , . For
instancee, ̅ , foor = 0.3 coonverges too a fixed positive valuee (Fig. 16b), indicating that ̅ is
positivee, and the flow is of a quasi-steady
q y type. In ccontrast, as tthe thicknesss ratio incrreases just
slightly to = 0.306, ̅ , turns negative and experiiences largge oscillatiions until
∗
approxim
mately = 140 (Figg. 16c). Annother impoortant detail is that thhere is a ssubstantial
oscillatiion in ̅ , before ∗
= 140 (witthin the dashed circle in Fig. 16c),, causing it to decline
sharply from positivve to negatiive. Such oscillation
o inn the force ssignals may be the critical reason
behind tthe sudden jjump in ̅ iin Fig. 3.

Selectedd snapshots of instantaaneous vortiicity contouurs for = 0.306

0 corresponding too the time
when oscillation
o oof ̅ , ooccurs are displayed
d inn Fig. 17. At ∗
= 1002.375, the flow has
reachedd a state wheere the RVss are rotatingg with the eelliptic cylinnder and shedding upw
wards. Due
to combbined effects of cylindder rotationn and freestream, the shed vorticces gather aabove the

22
 
cylinder and form a large-scale shear layer of negative sign (termed as SL hereafter). However,
∗
this SL is unstable, and its strength is oscillating. For example, a weaker SL is observed at =
∗
105.375 and 111.375, whilst a stronger one is recorded at = 108.375 and 116.375, respectively
(see Fig. 17). One possible explanation for the unstable SL is hypothesized as follows. The
change of overall vorticity in the SL may come from three parts: 1) feeding of vorticity from the
shed RVs to the SL in the downstream of the cylinder; 2) draining of vorticity from the SL to the
newly-formed RVs in the upstream; and 3) loss of vorticity via viscous diffusion. Therefore, any
imbalance from these three components would result in a net change of vorticity in the SL as
well as in the RV system. After a few oscillations, the SL can no longer maintain its position
∗
relative to the cylinder, and it start to roll up and move closer towards the cylinder ( = 118.375
and 119.375). In the meantime, the RVs start to shed downstream, resulting in a much weaker
∗
vorticity feeding to the SL. Eventually, the whole SL is drained away by newly-formed RVs ( =
∗
120.375), and a new stable state with extended wake is established ( = 130.375).

23
 
FIG. 16. Mean lift and
a drag cooefficients avveraged oveer every rotaational cyclee at = 5.0 for (a) =
0.25; (b)) = 0.3; (c)) = 0.306; (d) = 0.5 and (e) 0.755. Note that = 0.3 is th he threshold
d thickness
ratio.

24
 
FIG. 17. Flow evoluution for thee case of = 0.306 and = 5.0 show
wing the varying shear layer and
ment of a neew stable staate.
developm

The aboove-mentionned shear laayer of flucttuating strenngth before the flow is fully develloped is in
fact obsserved consiistently for a large rangge of thicknness ratio 0..306 0.875 at = 5.0 (see
To see this, an additionnal comparison of flow
the dashhed circle inn Fig. 16d and 16e). T w structure
betweenn = 0.3 annd 0.5 is proovided in F
Fig. 18. For = 0.3, a stable
s SL caan be seen above the
cylinderr (Fig. 18a), probably bbecause the dynamic gain
g and loss of its vortticity happeen to be in
an equillibrium. Nevvertheless, ssuch equilibbrium seemss to be achieevable only for a narrow
w range of
beforee . For thee thicker ellliptic cylindder with = 0.5, the SL
L experiencces similar oscillation
o
on top of the cylinnder (see F
Fig. 18b). Inn the meanttime, positiive vorticityy is seen too build up
betweenn the SL andd the cylindder. Eventuaally, a large vortex of positive vorticity and the unstable
SL arouund it shed into the waake. Concurrrently, the RVs start tto shed dow
wnstream, foorming an
extended wake in thhe new stabble state.

25
 
 FIIG. 18. Flow
w evolutions ffor (a) = 00.3 and (b) = 0.5 at = 5.0.

FIG. 19. Mean lift and

a drag cooefficients avveraged oveer every rotaational cyclee at = 4.6 for (a) =
0.266 an
nd (b) = 0.2274. Note th
hat = 0.2666 is the threshold thickneess ratio.

26
 
At = 44.6, the suddden jump off ̅ after = 0.266 is also linkedd to the SL ooscillation. As shown
in Fig. 119, a slight iincrease in tthe cylinderr thickness ffrom = 0.2266 to 0.2744 leads to a change of
sign in ̅ _ froom positivee and negaative after the
t flow iss fully deveeloped. Addditionally,
substanttial oscillattions in botth ̅ _ aand ̅ _ are recordded before ∗
= 70 (w
within the
dashed ellipse in F
Fig. 19b). Frrom vorticity contourss visualized in Fig. 20bb, it is founnd that the
overall flow evoluttion is analoogous to thee case of = 0.5 at = 5.0 (Fig. 18b), includiing the SL
oscillatiion, merginng of SL annd RVs, devvelopment oof extendedd wake, etc. On the otther hand,
althoughh oscillatioons are alsoo found inn ̅_ annd ̅ _ for = 0.266, the ooscillation
magnituudes decreasse in time. Eventually,
E a stable SL
L is maintaiined above tthe cylinderr (see Fig.
20a). Thhe overall fllow behavioors near aare hence in line with thhose at = 55.0.

FIG.. 20. Flow evvolutions forr (a) = 0.2666 and (b) = 0.274 at = 4.6.

3.4 Thrree-dimenssionality

The thrree-dimensionality of cylinder w

wake is a ccrucial featture that m
may affect the force
characteeristics signnificantly (seee for exam
mple, Mittal 2004, Pralits 2010, 20013, Thomppson et al.
2014b, Navrose 20015). Accorrding to paast studies ((Rao et al. 2013, 20155), three-dim
mensional
modes appear as the rotatioon rate off a circularr cylinder is increased. Althouugh three-
dimensiionality is nnot the focuus of the prresent studyy, a few 3D
D simulations are carriied out to
examinee how threee-dimensionnality affectts forces actting on the elliptic cyllinder and aassociated

27
 
flow structures. Here, two thickness ratios, i.e. = 0.125 and 0.5 are selected for the 3D
simulations, and the aspect ratio (AR) of the cylinders is set to 10 (i.e. the span of the cylinder is
10 times of its characteristic length L). This value is consistent with our previous experimental
work (Lua et al. 2010). In another numerical study by Navrose et al. (2015), an AR of 12 was
employed. The 3D mesh consists of 5.5 million tetrahedral cells clustered around the cylinder
surface, with 500 nodes uniformly distributed along the span. According to Rao et al. (2015),
mode E instability prevails in the current range of 3.0 5.0 at = 200 for a circular
cylinder, with a spanwise wavelength larger than unity. Therefore, assuming a wavelength of one,
the present spanwise resolution leads to at least 50 points within a wavelength.

Figure 21 compares the cycle-averaged lift and drag coefficients obtained from 2D and 3D (with
or without endplates at the two ends of the cylinder) simulations. At = 3.0, it is observed that
the 3D results without endplates deviate from the 2D results, thus indicating the existence of
flow three-dimensionality. This is further shown in Fig. 22, whereby the vorticity distribution at
/ = -2.5 differs from that at the mid-span position / = 0. However, adding endplates to the
3D cylinder can effectively reduce flow three-dimensionality, as is evidenced by the matching of
forces with the 2D results (see Fig. 21, = 3.0) and the appearance of consistent flow structures
along the span (see Fig. 22, = 3.0 with endplate). At high of 4.0, the endplates are still
effective for = 0.5 but not so for the case of = 0.125. At even higher of 5.0, the endplates
fail to suppress three-dimensionality for both thickness ratios. Multiple components of spanwise
wavelengths seem to exist (see Fig. 22b, = 5.0), suggesting that there may be more than one
instability mode, and the three-dimensionality of a rotating elliptic cylinder may be more
complex than that of a rotating circular cylinder. Also, there is a phase shift in the temopral force
distribution between 2D and 3D calculations. This phase shift may cause the threshold values
in 3D to differ from those in 2D. A more detailed study on the effect of flow three-dimensionality
is certainly required. In this context, and similar to many 2D simulations on a rotating circular
cylinder reported in the literature (e.g. Chew et al. 1995, Mittal and Kumar 2003, Rao et al. 2013,
2015, Thompson et al. 2014a), the present study provides useful two-dimensional base flow data
for future stability analysis, such as that performed by Thompson et al. (2014a) on a stationary
elliptic cylinder.

28
 
FIG. 21. Comparisoon of cycle-averaged forcces between 2D and 3D, for = 0.125, 0.5 and = 3.0, 4.0,
5.0.

29
 
FIG. 22. (a) Contou urs of spanwwise vorticityy at three spaanwise positions for = 0.125 and = 3.0, 4.0,
5.0; (b) iiso-surfaces oof spanwise vvorticity for = 3.0 and 55.0, with end
dplates.

4. Con
nclusions
In this paper, a nuumerical invvestigation has been cconducted too study thee flow past a rapidly
rotating 2D ellipticc cylinder aat = 2000. Two paraameters aree varied, naamely the normalized
n
rotating velocity = 3.0, 4.0, 4.6 and 5.0
5 and thicckness ratioo ranging from 0.06225 to 1.0.
Results show thatt these twoo parameterrs play a ssignificant role in thee aerodynam
mic force
characteeristics of thhe rotating cylinder, annd contribuute to the geeneration vaarious wakee patterns,
includinng steady, quuasi-steady,, aperiodic aand periodicc flows. At the lowest = 3.0, thee mean lift

30
 
magnitude increases almost linearly with cylinder thickness (similar to < 2.0 cases reported in
literatures), and a mean thrust is generated when 0.75 and peaked at = 0.375. Our flow
visualizations show that the force behavior for this particular velocity ratio is dictated by a
“hovering vortex” (HV) above the cylinder and a negative pressure region below the cylinder.
For above 3.0, there exists an threshold thickness ratio , above which the mean lift enters a
plateau until  0.5, and the mean drag suffers an abrupt change of sign from positive to
negative. Due to the rapid rotation, the HV for these cases gives way to multiple rotating vortices
(RVs) that are forced to move with the cylinder. The RVs may play a role in the existence of
aperiodicity in the wake at the lower end of thickness ratios. As to exactly how these RV interact
with each other and with the cylinder to cause the observed changes in forces around and sign-
switching of drag after is not clear and requires further investigation.

Several 3D simulations have also been conducted to examine the three-dimensionality of a

rotating elliptic cylinder. It is found that adding endplates to the 3D cylinder can suppress three-
dimensionality at = 3.0, but it is ineffective at = 5.0. A linear stability analysis may be
necessary in future work.

References

Alexander, D. E. (2004). Nature’s flyers: birds, insects, and the biomechanics of flight. The Johns Hopkins
University Press.
Badr, H. M., and Kocabiyik, S. (1997). Symmetrically oscillating viscous flow over an elliptic. J. Fluids Struct.
11(7), 745-766. doi: https://doi.org/10.1006/jfls.1997.0100
Bénard, H. (1908). Formation de centres de giration à l’àrriere d’un obstacle en mouvement. C. R. Acad. Sci. Paris
147, 839–842.
Chandna, A. (1997). Flow past an elliptic cylinder. J. Comput. Appl. Math. 85(2), 203-214. doi:
https://doi.org/10.1016/S0377-0427(97)00035-6
Chew, Y. T., Cheng, M., and Luo, S. C. (1995). A numerical study of flow past a rotating circular cylinder using a
hybrid vortex scheme. J. Fluid Mech. 299, 35-71. doi: 10.1017/S0022112095003417
Hu, R. (2015). Three-dimensional flow past rotating wing at low Reynolds number: a computational study. Fluid
Dyn. Res. 47(4), 045503.
Iversen, J. D. (2006). Autorotating flat-plate wings: the effect of the moment of inertia, geometry and Reynolds
number. J. Fluid Mech. 92(2), 327-348. doi: 10.1017/S0022112079000641
Kang, S., Choi, H., and Lee, S. (1999). Laminar flow past a rotating circular cylinder. Phys. Fluids 11(11), 3312-
3321. doi: 10.1063/1.870190
Kármán, Th. v. (1911). Über den Mechanismus des Widerstandes, den ein bewegter Körper in einer Flüsseigkeit
erfährt. Gött. Nachr. 509–511.
Kumar, S., Cantu, C., and Gonzalez, B. (2011). Flow past a rotating cylinder at low and high rotation rates. J. Fluids
Eng. 133(4), 041201-041201-041209. doi: 10.1115/1.4003984
Lua, K. B., Lim, T. T., and Yeo, K. S. (2010). A rotating elliptic airfoil in fluid at rest and in a parallel freestream.
Exp. Fluids 49(5), 1065-1084. doi: 10.1007/s00348-010-0847-7
Lua, K. B., Lu, H., and Lim, T. T. (2018). Rotating elliptic cylinders in a uniform cross flow. J. Fluids Struct. 78, 36-

31
 
 51. doi: 10.1016/j.fluidstructs.2017.12.023
Lugt, H. J., and Ohring, S. (2006). Rotating elliptic cylinders in a viscous fluid at rest or in a parallel stream. J.
Fluid Mech. 79(1), 127-156. doi: 10.1017/S002211207700007X
McCutchen, C. W. (1977). The spinning rotation of ash and tulip tree samaras. Science 197(4304), 691.
Mittal, S., and Kumar, B. (2003). Flow past a rotating cylinder. J. Fluid Mech. 476, 303-334. doi:
10.1017/S0022112002002938
Mittal, S. (2004). Three-dimensional instabilities in flow past a rotating cylinder. J. Appl. Mech. 71, 89-95. doi:
10.1115/1.1631032
Modi, V. J., and Dikshii, A. K. (1975). Near-wakes of elliptic cylinders in subcritical flow. AIAA J. 13(4), 490-497.
doi: 10.2514/3.49736
Naik, S. N., Vengadesan, S., and Prakash, K. A. (2017). Numerical study of fluid flow past a rotating elliptic
cylinder. J. Fluids Struct. 68, 15-31. doi: 10.1016/j.jfluidstructs.2016.09.011
Naik, S. N., Vengadesan, S., and Prakash, K. A. (2018). Linear shear flow past a rotating elliptic cylinder. J. Fluid
Eng. 140, 121202. doi: 10.1115/1.4040365
Navrose, Meena, J., and Mittal, S. (2015). Three-dimensional flow past a rotating cylinder. J. Fluid Mech. 766, 28-
53. doi: 10.1017/jfm.2015.6
Padrino, J. C., and Joseph, D. D. (2006). Numerical study of the steady-state uniform flow past a rotating cylinder. J.
Fluid Mech. 557, 191-223. doi: 10.1017/S0022112006009682
Pralits, J. O., Brandt, L. and Giannetti, F. (2010). Instability and sensitivity of the flow around a rotating circular
cylinder. J. Fluid Mech. 650, 513-536. doi: 10.1017/S0022112009993764
Pralits, J. O., Giannetti, F. and Brandt, L. (2013). Three-dimensional instability of the flow around a rotating circular
cylinder. J. Fluid Mech. 730, 5-18. doi: 10.1017/jfm.2013.334
Rao, A., Leontini, J. S., Thompson, M. C., and Hourigan, K. (2013). Three-dimensionality in the wake of a rapidly
rotating cylinder in uniform flow. J. Fluid Mech. 730, 379-391. doi: 10.1017/jfm.2013.362
Rao, A., Radi, A., Leontini, J. S., Thompson, M. C., Sheridan, J., and Hourigan, K. (2015). A review of rotating
cylinder wake transitions. J. Fluids Struct. 53, 2-14. doi: https://doi.org/10.1016/j.jfluidstructs.2014.03.010
Sreenivasan, K. R., Strykowski, P. J., and Olinger, D. J. (1987). Hopf bifurcation, Landau equation, and vortex
shedding behind circular cylinders. In American Society of Mechanical Engineers, Fluids Engineering
Division (Publication) FED (Vol. 52, pp. 1-13). ASME.
Stojković, D., Breuer, M., and Durst, F. (2002). Effect of high rotation rates on the laminar flow around a circular
cylinder. Phys. Fluids 14(9), 3160-3178. doi: 10.1063/1.1492811
Thompson, M. C., Radi, A., Rao, A., Sheridan, J., and Hourigan, K. (2014a). Low-Reynolds-number wakes of
elliptical cylinders: from the circular cylinder to the normal flat plate. J. Fluid Mech. 751, 570-600. doi:
10.1017/jfm.2014.314
Thompson, M. C., Radi, A., Rao, A., Sheridan, J., and Hourigan, K. (2014b). The existence of multiple solutions for
rotating cylinder flows. 19th Australasian Fluid Mechanics Conference, Melbourne, Australia.
Tokumaru, P. T., and Dimotakis, P. E. (1991). Rotary oscillation control of a cylinder wake. J. Fluid Mech. 224, 77-
90. doi: 10.1017/S0022112091001659
Williamson, C. H. K. (1996). Vortex Dynamics in the Cylinder Wake. Annu. Rev. of Fluid Mech. 28(1), 477-539. doi:
10.1146/annurev.fl.28.010196.002401
Zdravkovich (1997). Flow Around Circular Cylinders: Volume I: Fundamentals. Oxford Science Publications.

32
 
 The flow past an elliptic cylinder rotating with a high velocity ratio from 3.0 to 5.0 is
numerically investigated.
 There exists a threshold thickness ratio, above which the mean lift enters into a plateau,
and the mean drag experiences a sudden jump.
 After the threshold thickness ratio, a large-scale, unstable shear layer is observed to
merge with multiple vortices that rotate with the cylinder, resulting in a new stable wake.
 Various wake topologies are revealed, including steady, quasi-steady, periodic and
aperiodic flows.