Вы находитесь на странице: 1из 19

ELSEVIER

Three-Dimensional Vortex Dynamics in Bluff


Body Wakes
C. H. K. Williamson • Over the last few years, there has been a surge of activity concerning
Mechanical and Aerospace Engineering, wake flows, in particular regarding 3-D aspects of flows that are nominally
Cornell University, two-dimensional. In the laminar shedding regime, it has become evident
Ithaca, New York that end boundary conditions can control the three-dimensional wake
pattern over long spans. These end conditions can be manipulated to
produce oblique shedding, chevron patterns, parallel shedding, cellular
shedding, and vortex dislocations. Transient patterns known as phase
shocks and phase expansions can be induced by temporal control of end
conditions by using variable suction. These experimental patterns may
be modeled analytically using Guinzburg-Landau equations. The long-
standing controversy regarding discontinuities in the Strouhal-Reynolds
number relationship is caused by the phenomenon of oblique shedding, in
turn dependent on end conditions. The original scatter of around 20% in
the measurement of this relationship between different laboratories, up to
1988, has been reduced to 1% when parallel shedding is induced. In the
wake transition regime, the large discrepancy in reported critical Reynolds
numbers (Re = 140-190) over the last 40 years has been found to be due
to contamination from end conditions. Indeed, remarkably "clean" end
conditions can extend the laminar regime to Re = 194. The wake transition
is shown to involve two successive stages, each corresponding to a disconti-
nuity in both the character of wake formation and the S - R e relationship.
The first discontinuity (Re = 180-194) is associated with the inception of
vortex loops and is hysteretic. The second discontinuity (Re = 230-260)
corresponds to a change to a finer scale streamwise vortex structure. The
wake can exhibit enormous vortical structures, "vortex dislocations," which
explain the origin of the low-frequency velocity irregularities first observed
by Roshko in 1954. Such dislocations, which cause surprisingly large fluc-
tuation energy downstream, are conceivably an intrinsic feature of transi-
tion of shear flows in general. In the far wake, the honeycomb-like
three-dimensional pattern first observed by Cimbala et al. in 1988, is found
to be due to interactions between oblique shedding waves formed upstream
with growing 2D waves amplified in the far wake from free-stream distur-
bances. Previous observations may now be explained in terms of these wave
interactions. With these deductions as the starting point, we have gone on
to discover a new mechanism of "oblique wave resonance," whereby the
two wave systems above interact nonlinearly to produce a third "oblique
resonance wave," which is then preferentially amplified (by up to two
orders of magnitude over the 2-D waves). This single oblique wave reso-
nance can be visualized remarkably clearly and is distinctly different from
the models hitherto proposed for the far wake.

Keywords: vortex shedding, wake, cylinder wake, Guinzburg-Landau


equations, wake transition, streamwise vortices, vortex loop, vortex
dynamics, oblique wave resonance, phase shock, phase expansion,
vortex dislocation, oblique shedding, parallel shedding

Address correspondence to Professor C. H. K. Williamson, Mechanical and Aerospace Engineering, Upson Hall, Comell University, Ithaca, NY
14853.
Experimental Thermal and Fluid Science 1996; 12:150 168
© Elsevier Science Inc., 1996 0894-1777/96/$15.00
655 Avenue of the Americas, New York, NY 10010 SSDI 0894-1777(95)00085-2
Vortex Dynamics in Bluff Body Wakes 151

INTRODUCTION sional vortex structure. However, in the two-dimensional


view of the flow behind a bluff body, regions of concen-
Two-dimensional and three-dimensional vortical instabili- trated vorticity are "shed" into the downstream flow from
ties in wakes have been a subject of interest to engineers alternate sides of the body (and with alternate senses of
as well as to scientists for a great many years. Although rotation), giving the appearance of an upper row of nega-
the problem of an unseparated wake behind a splitter
tive vortices and a lower row of positive vortices, often
plate or symmetric airfoil is amenable to analytical studies
observed in an ordered arrangement as in Fig. 1. Such
(e.g., stability analysis), the wake of a nominally two-
formations of vortices were the origin of Strouhal's [1]
dimensional (2-D) or three-dimensional (3-D) bluff body
poses rather a greater analytical challenge. Approaches to classical measurements in 1878 of the sound frequency
this problem have been principally experimental, although produced by translating cylindrical rods through air, and
there has recently been vigorous activity in numerical for the Aeolian tones that are produced by the wind
simulation and analysis. In this paper, an overview is blowing over a wire or a string in an Aeolian harp. In a
presented of the vortex dynamics phenomena in the wake 1992 review, Rott [2] discusses the later contributions of
of a circular cylinder at low to medium Reynolds numbers. Lord Rayleigh in normalizing Strouhal's frequency data
The classical view of a "vortex street" in cross section is using the Strouhal number (fD/U, where f is frequency,
shown in Fig. 1. Interestingly, it will be shown later that D is diameter, and U is flow velocity) versus Reynolds
such cross-sectional photographs actually contain useful number. We later (Fig. 3) compare these original data
information regarding the distribution of three-dimen- with modern measurements, which have the advantage of

Re-'iS0

Figure 1. Visualization of laminar


and turbulent vortex streets. A
Karman vortex street consists of a
sequence of alternate-sign vortices
arranged in two rows, traveling
downstream in the wake of a body.
The vortices in the upper row are
rotating clockwise, while the lower
vortices rotate counter clockwise.
As the free stream flows from left
to right past the circular cylinder
(at the left-hand edge of the pic-
tures), the separating boundary lay-
ers from the sides of the body roll
up into discrete vortices. These
vortices are laminar in the case of
low Reynolds numbers, Re = 150,
and turbulent for higher Reynolds
numbers, Re = 300 and 700. This
aluminum flake visualization tech-
nique indicates white regions prin-
cipally where there is streamwise
vorticity and where the vortices are
turbulent rather than laminar.
(Photographs from Williamson
[34].)
152 C . H . K . Williamson

hot-wire anemometry, spectrum analyzers, and wind tun- dimensional origins. Our focus in this paper is on these
nels. Taking this into account, the agreement in the mea- new discoveries. An overview of the different vortex insta-
surement of vortex shedding frequency, after 100 years, is bilities that are found to appear as the Reynolds number
quite remarkable! is increased are shown schematically in Fig. 2. We shall
The first definition of flow regimes based on measure- discuss 3-D wake patterns in the laminar regime, fine-scale
ments of velocity fluctuation, spectra, and frequency was 3-D spanwise instabilities, and also the appearance of
given by Roshko [3] in 1954. He found a "stable" (peri- relatively enormous "vortex dislocations" in the wake
odic) laminar vortex-shedding regime for Re = 40-150, a transition regime, and finally we will demonstrate a new
transition regime in the range Re = 150-300, with an oblique wave resonance mechanism that has been discov-
"irregular" regime for Re = 300-10,000 + , where velocity ered in the far wake.
fluctuations showed distinct irregularities. Similar regimes
were confirmed by Bioor [3] in 1964. A surge of recent
THREE-DIMENSIONAL VORTEX DYNAMICS
work has shed further light on phenomena occurring in
these regimes and their precise Reynolds number ranges. 1N T H E L A M I N A R R E G I M E
In more recent experiments, significant steps forward in Oblique and Parallel Modes of Vortex Shedding
our understanding of wake vortex dynamics have come
from the many studies of three-dimensional phenomena, Prior to 1988, a number of (sporadic) observations and
which have led to some new explanations of long-standing measurements had been made within the laminar vortex
controversies that were hitherto assumed to have two- shedding regime that showed that vortices can shed at

Primary Karmsm
vortel 8heddJltj

Small scales I Large scales]

12trse-scaJe $-D distort|oun

Mo4e A
VOl~eZ |oops

J
~ortcx Dish

Figure 2. Instabilities involved in the de-


Mode B
velopment of turbulence in the wake. As
the Reynolds numbers increase (down voytel[ pairl
the page), the formation of wake turbu-
lence is characterized by the following
principal instabilities: (1) The formation
of primary Karman vortices; (2) the in-
ception of small-scale streamwise vortex
structures (modes A and B to be defined
later); (3) the formation of very large
scale three-dimensional structures, now
known as vortex dislocations; and (4) the
development of shear-layer instability Shear |lJyer iltsutbilities
vortices, which are themselves prone to st bigkar
Reyuoids l U i b e r l
three-dimensional instability of small
scale. For supercritical Reynolds num-
bers, the boundary layer on the body
itself becomes turbulent before separa-
tion.
Vortex Dynamics in Bluff Body Wakes 153

some oblique angle to the axis of the cylinder in what we 0.~'2

now term oblique shedding. Some of these early measure-


/
ments are discussed in the 1972 review paper of Berger S
and Wille [5], who show that typical oblique angles of
15-20 ° were found. An example of slantwise shedding was
photographed in 1964 by Berger [6], using smoke in a wind
0.18
tunnel, although contrary observations in 1957 by H a m a
[7], who used a towing tank, demonstrated only parallel
shedding. These contradictory reports remained unex-
plained.
A further phenomenon, which we shall show to be
directly related to the above, is that of discontinuities in 0.14
the relationship between Strouhal number and Reynolds
number in the laminar shedding regime, as first detected / ~ STIROUH(A1L87S1
very clearly in 1959 by Tritton [8] near Re = 75, and
subsequently the source of a great deal of debate over a
period of 30 years, beginning with a series of papers by
Tritton and Gaster (see, e.g., Gaster [9]). Some idea of the 0.1~
/
, Re

ensuing scatter among experimental measurements of a


Strouhal frequency over the period 1878-1978 is given in 0.20 , r F
Fig. 3a, in a plot adapted from Gerrard [10], showing - - Eisenlohr & Eckelmann (198g) WT 60
----~ Hammache & Gharib (1989) WT 265 ~ o
scatter on the order of 20% e v e n among the modern ....... S t ~
experiments. This scatter was present despite the fact, as O.IB
pointed out by Roshko [11], that "the quantities involved
(U, D, v, and f ) could be rather easily measured to better 0.16
than 1% accuracy." (v is kinematic viscosity.) Many expla-
S
nations were put forward over the years: Different modes
of vortex shedding have been attributed to different forms 0.14 /~le'* -'--- Norberg (1994) WT 2000
/ ~ w t a m , o n (1993) rr zoo
of 2-D wake instability [8], to shear in the oncoming free A/~" ° Williamson (1988a} WT 90
stream [9], to differences in free-stream turbulence [5], to ¢~ • WT 115
0.12 WT 250
changes in the influence of vorticity diffusion in the near o ~ 70
wake [10], and more recently to a scenario involving a XYTT 140

"route to chaos" in the cylinder wake [12]. A convincing 0.10 I h I I I I


40 60 80 100 120 140 160 180
case was offered in 1987 by Van Atta and Gharib [13] that
Re
flow-induced vibration could cause discontinuities in the
b
S - R e relation, and their meticulous work was subse-
quently supported by the computations of Karniadakis Figure 3. (a) Strouhal-Reynolds number relationship
and Triantafyllou [14]. However, it can be shown that in (1878-1978). Over a period of 100 years, beginning with the
the absence of certain effects, such as free-stream shear vortex frequency measurements of Strouhal [1], there has
and cylinder vibration, a discontinuity in the S - R e rela- existed on the order of 20% disparity among the many
tion can be caused by the unexplained phenomenon de- measurements of Strouhal number versus Reynolds number
scribed earlier, namely oblique shedding. In other words, in the laminar shedding regime. The data shown come from
we now know that the frequency measurements are influ- several of these past investigations (which were collected by
enced by three-dimensional flow phenomena, as described Gerrard [10], from which many of these curves may be
below. found). The curves presented here include the original data
In a new understanding of the above problem, it has from Strouhal. A concise and interesting discussion of some
recently been shown that the S - R e discontinuity originally of the history of the Strouhal number measurements can be
observed by Tritton [8], and subsequently in different found in Rott [2]. (b) Strouhal-Reynolds number relation-
ships. Agreement to the 1% level of the S-Re relationship
forms by many others, is caused by a changeover from one for laminar parallel shedding using different techniques (in
mode of oblique shedding to another oblique mode, as Re several different laboratories). WT indicates wind tunnel
is increased [15-19]. We may thereby link these two facility; XYTF, a water facility known as the XY Towing
hitherto unexplained phenomena. It is shown that the Tank. The numbers at the end of each reference indicate the
particular boundary conditions at the spanwise ends of the values of the length/diameter ratio (or aspect ratio) for the
cylinder dictate the angle of shedding over the whole cylinders.
span, even for a cylinder that is hundreds of diameters in
length, by what is termed "indirect" influence [16]. Gerich
and Eckelmann [20] had earlier shown that a region close dictated by the end conditions in that half. Below a critical
to the ends of a cylinder (about 10 diameters in length) Re for the Strouhal number discontinuity (Re = 64), a
can be influenced in a direct manner, causing a cell of quasi-periodic pattern may be found, caused by the pres-
lower frequency shedding to appear near the ends. ence of a further cell of higher frequency appearing in the
In the towing tank and wind tunnel experiments of central span region. The shedding assumes a cellular
Williamson [15, 16] in 1988-1989, it was found that the structure, with different frequencies found to coexist at
oblique vortices formed a periodic chevron pattern as different spanwise locations and with vortex dislocations
shown in Fig. 4a. Over each half-span, the oblique angle is [16] or vortex splitting [17] found to occur at the cell
154 C . H . K . Williamson

visualization was conducted near the start of motion. On


the other hand, Berger used a wind tunnel and would
have observed his patterns after a great many cycles of
shedding, allowing time for the oblique shedding to influ-
ence the complete span.
Judging from the above, one might question whether it
is possible to manipulate the end conditions to promote
parallel shedding. Indeed, it has now been found that
there are several means to achieve parallel shedding, as
described in a number of papers [15, 17, 22, 23]. The
mechanical devices used to achieve parallel shedding in-
clude angling the leading edge of endplates inwards, end-
ing the span with larger coaxial cylinders, or locating large
cylinders normal to and upstream of the test cylinder at
the ends. It appears that the techniques involve a slight
speeding up of the flow near the ends, which is evident in
II III III
the nonmechanical technique used by Miller and
Williamson [23]. In this case, the incident flow near the
ends is speeded up by suction tubes placed around 10
diameters downstream at each end of the test cylinder.
In the context of oblique waves, it is relevant to men-
tion a result that may be derived from Squire's transfor-
mation [24] of 1933. It can be shown (for a given wake
profile and Reynolds number and for a parallel flow) that
if the frequency and temporal growth rate of the most
unstable two-dimensional wave are respectively f0 and o-0,
then for an oblique wave at angle 0, the most unstable
frequency fo is given by fo = fo cos 0, and similarly the
growth rate tr0 is given by tr0 = tr 0 cos 0. From such
theoretical considerations, Garry Brown suggested (private
communication, 1988) that the cylinder wake frequencies
may follow the same trend. With parallel shedding, one
finds that the Strouhal-Reynolds number curve is com-
pletely continuous, as shown in Fig. 3b. It was also shown
[15] that one may define a "universal" Strouhal curve, in
b the sense that the experimental oblique-shedding data
( S o ) can be closely collapsed onto the parallel-shedding
Figure 4. Oblique and parallel vortex shedding. These pat- curve (S O) by the transformation
terns are controlled by end (spanwise) boundary conditions.
Flow is upward in both photographs. (a) The chevron pattern
S O = So~cos O. (1)
of oblique shedding [15]; (b) parallel vortex shedding [16].
The dependence of shedding frequency on the phe-
nomenon of oblique shedding would now explain much of
boundaries because vortices on each side move in and out the significant scatter found in the many measurements of
of phase with each other. A clear example of three coex- Strouhal number, illustrated earlier in Fig. 3a. Since 1988,
isting cells is shown in the smoke visualizations of Koenig a number of groups have measured the S - R e relationship
et al. [21]. in the laminar regime, and the agreement among the
The above experiments also reconcile the earlier dif- laboratories is remarkably close to the 1% level, as shown
ferences between the experiments of Berger [6] in 1964 in Fig. 3b.
and Hama [7] in 1957. It is found from towing tank Experimentally, the cosine "law" of Eq. (1) has been
experiments [16] that at the start of motion the vortices closely confirmed in several papers including those of
are shed parallel to the body (also observed by Gerrard Koenig et al. [17, 25] and Miller and Williamson [23]. In
[10]). This suggests that the most unstable mode is the the experiments of Miller and Williamson, a nonmechani-
parallel-shedding mode, although the shedding can be cal (suction) technique allowed continuous control of the
forced away from this mode by end boundary conditions oblique shedding angle between 0 ° and 23 °, resulting in
(much like a forced oscillator). Thereafter, an oblique very close agreement with the cosine formula. However, a
shock travels inwards from the end (under the indirect very revealing set of experiments by the Goettingen group
influence of spanwise phase coupling), bringing behind it a (Koenig et al. [25]) suggests that only discrete values of
region of oblique shedding, until ultimately the whole the shedding angles may exist naturally in the cylinder
span exhibits the chevron pattern. Surprisingly, one needs wake. The Strouhal curves for these different modes lie
to translate a 100-diameter-long body a distance of around along only four unique curves in the S - R e plane, all of
600 diameters to achieve the chevron pattern. In the which have very close agreement with the cosine formula.
experiments of Hama, he used a towing tank, and he It is perhaps a paradox that on the one hand it is possible
observed only parallel shedding due to the fact that his to induce a continuous variation of shedding angles (Miller
Vortex Dynamics in Bluff Body Wakes 155

and Williamson), while on the other hand the flow chooses tinct phenomena, as illustrated in Fig. 5. If an oblique
only particular discrete angles (Koenig et al.). This para- angle is generated behind a front that is greater than the
dox remains to be explained, although it seems clear that angle ahead of the front, then a phase shock is produced.
it is a manifestation of particular end boundary condi- If the oblique angle behind the front is less than the angle
tions. ahead of the front, then a phase expansion is produced.
Some straightforward predictions of these transient
Three-Dimensional Phase Dynamics wake patterns have been made, Miller and Williamson
[23], from rather simple considerations as an extension of
Triggered by the recent surge of experimental discoveries some measurements made by Williamson [16] in 1989. It
of 3-D shedding patterns in the cylinder wake, there have was shown in the latter work that the normal vortex
been a number of fruitful analytical approaches to model- wavelength remained closely constant, at a given Reynolds
ing both the steady-state and transient 3-D shedding pat- number, irrespective of whether the vortices shed oblique
terns, involving oblique or parallel shedding, chevron pat- or parallel to the body. In Fig. 6a, it is shown that if one
terns, cellular shedding and vortex dislocations, phase starts off with a phase shock at the top of the sketch, then
shocks, and phase expansions (to be described in this invoking the normal-wavelength constant condition simply
section). Triantafyllou [26] in 1992 considered the prob- leads to a spanwise translation of the shock (to the right in
lems of "slow" transverse modulations. Noack et al. [27] in this example) at a constant speed. Measurements of the
1991 modeled the formation of cells of different shedding phase shock angle relative to the direction downstream
frequency in the wake of a slender cone using van der Pol [16] showed good agreement with predictions from this
oscillators with diffusive spanwise coupling. The group at model. If, on the other hand, one has a region of small-an-
Marseille (Albarede et al. [28]) in 1990 used diffusively gle shedding impulsively generated that is thereafter mov-
coupled Stuart-Landau oscillators along the span. The ing into a region of larger-angle shedding, then the model
resulting equation is a Guinzburg-Landau equation in yields a phase expansion. The initial shock fans out into a
time and spanwise coordinates, which was shown by A1- curved sequence of vortices, where the radius of curvature
barede and Monkewitz [29] in 1992 to be equivalent to the increases linearly with time, as shown schematically in Fig.
van der Pol model near the onset of Karman vortex 6b. In this example, the impulsively generated end condi-
shedding. Amplitude (A) in these equations represents a tion for parallel shedding is imposed at the right-hand
measure of wake velocity fluctuations in the cylinder wake. end, whose influence is then felt by the original oblique
The Guinzburg-Landau equation, involving a diffusive shedding region to the left.
spanwise (z) coupling term, may be written The present phase expansion and phase shock pat-
tern phenomena are actually predicted from the
Guinzburg-Landau equations [32], which take the form of
--Ot = ~rA - AIAJZA + /z oqz2 , (2) a Burger's equation for the spanwise wavenumber (q),

where the coefficients are evaluated experimentally. The 8q ,9 2q 8q


chevron pattern [15, 16] is well modeled by this approach. -"~ = /LLI~z2 -- Alq--~z , (3)
It is also of interest to note that if the initial conditions in
the model consist of small-amplitude random noise, then for which (phase) shocks and expansions are well-known
parallel waves emerge naturally from this noise because solutions. A major difference between these phenomena is
they are found to be the most unstable mode. Thereafter, that phase shocks (involving regions of straight vortices)
the boundary condition (A = 0 at the spanwise ends) translate at constant speed, crossing the complete span in
induces a region of oblique waves to travel inward from a finite time, whereas a phase expansion (involving curved
the ends behind a "phase shock" until the chevron pattern vortices) requires an infinite time (theoretically) to com-
appears fully across the span, as found in the experiments. plete its development.
Other complementary papers were published in 1992 by
Park and Redekopp [30] and Chiffaudel [31], who used a THREE-DIMENSIONAL VORTEX DYNAMICS
2-D Guinzburg-Landau equation (in streamwise and IN T H E TRANSITION REGIME
spanwise coordinates) to study the 3-D shedding patterns
and stability, respectively, and found phenomena such as The (wake) transition regime has had remarkably few
the phase shocks noted above. investigations in comparison to both the laminar regime
Transient patterns have recently been studied experi- and those regimes at higher Re involving shear layer and
mentally using a continuously variable suction technique boundary layer transitions. Possibly this is because of the
described earlier [23]. It was found that the chevron sensitivity of this regime to experimental conditions and to
pattern is in fact unstable in that any minutely small the difficulty in determining flow structure because of the
difference in the (magnitude of) oblique shedding angle intermittent nature of the flow. Nevertheless, recent work
will cause the apex of the chevron to translate spanwise shows that this regime is surprisingly rich in vortex dynam-
until there remains but one orientation of shedding across ics phenomena, both fine-scale and large-scale, which
the span. As discussed by Park and Redekopp [30] and have counterparts in free shear flow (mixing layer) transi-
Albarede and Monkewitz [29], the chevon is a "stationary" tion.
wavenumber shock, remaining stationary only under con- The transition to three-dimensionality in the wake can
ditions of exact spanwise symmetry. Transient experiments conveniently be described with reference to the measure-
were conducted that involved setting up one angle across ments of Strouhal-Reynolds number in Fig. 7, where it
the complete span and then impulsively changing the end may be observed that the transition, originally described
boundary conditions. These experiments led to two dis- by Roshko [3] in 1954, actually involves two discontinuous
156 C . H . K . Williamson

Figure 5. 3-D phase dynamics. (a)


"Phase shock." In this case, a re-
gion of small positive angle of
shedding (induced by the lower end
condition) is taken over by a region
of larger negative oblique shedding
angle that was impulsively trig-
gered by the top end boundary
condition. The result is a phase
shock, separating the two regions
of different shedding angle, that
translates down the span at a con-
stant speed. (b)"Phase expansion."
In this case, a large negative angle
of shedding (induced initially by
the top end condition) is pene-
trated by a smaller angle of shed-
ding impulsively generated by the
top end condition. The result is a
phase expansion that spreads out
gradually, yielding curved vortices
whose radius of curvature in-
creases in time. (From Miller and
Williamson [23].)

changes [33]. At the first discontinuity, the Strouhal fre- 3-D shedding modes corresponds to a spanwise instability
quency drops from the laminar curve to one correspond- in the wake; mode A involves the inception of vortex loops
ing to a " m o d e A " 3-D shedding at around Re = 180. and counterrotating streamwise vortex pairs; mode B in-
This discontinuity is hysteretic, and the exact critical Re volves a finer scale streamwise vortex structure.
depends on whether the flow speed is increased or de- The definition of such Strouhal discontinuities is clearly
creased, as we shall see below. As Re is increased up to possible only if one takes long time averages of the
the range Re = 230-260, there is a further discontinuity spectra of wake velocity fluctuations. Such spectra are
in Strouhal number, indicating a further " m o d e B." This shown in Fig. 8, where the first discontinuity in Fig. 8a
discontinuity may be contrasted with the first in that it is exhibits two possible spectra at the same Re due to the
not hysteretic and instead involves a gradual transfer of hysteresis effect. At the second discontinuity in Fig. 8b,
energy from mode A to mode B, as one increases Re. the spectra are twin-peaked. The lower peak correspond-
Interestingly, if one refers to the original S - R e data of ing to mode A gradually gives way to the peak at mode B
Roshko [3] of 1954, one finds that most of the scatter is as Re is increased. This is due to an intermittent swapping
centered around Reynolds numbers corresponding to these between modes, rather than the simultaneous existence of
two discontinuities. It will be seen later that each of these both modes [34]. Norberg's investigations (private commu-
Vortex Dynamics in Bluff Body Wakes 157

a ,\ ~ i bers for wake transition (Re = 140-190) in the literature.


Surprisingly, the origin of these differences has had al-
most no attention in the literature, except for the study by
Bloor [4] in 1964 and the recent studies by Hammache
'~ H i"
J 2 I and Gharib [19] and Miller and Williamson [23]. Bloor
verified Roshko's suggestion that wind tunnel turbulence
could be important in setting the wake transition Re, and
she found values extending from roughly Re = 140 to
Re = 190. However, one does not know how the end
conditions were affected by the different levels of turbu-
lence. It appears that in a number of facilities with good
flow quality (turbulence level 0.1%), investigators have
found a large disparity in critical Re; for example, Roshko
[3] and Tritton [8] found Re = 150, Eisenlohr and Eckel-
/ \\ ~.-_I v ..... mann [17] and Norberg [35] found Re = 165, and
Williamson [15, 16] found Re = 178.
/ "\~l Hammache and Gharib [19] showed that the critical Re
for wake transition was slightly increased if instead of
ending their test cylinder span on the sides of the wind
\ tunnel wall they introduced their "control cylinders" at
the ends of the span to induce parallel shedding. They
suggested that transition was triggered early at Re = 156
b )~Flow by the fact that the shedding was oblique rather than
parallel. However, other studies found a higher transition
Re on the order of 165-180 in the presence of the same
angles of oblique shedding, suggesting that early transition
is due to some other feature of the flow not yet accounted
for. It is perhaps not so much whether the shedding is
oblique or parallel that influences the transition Re, but
rather the end boundary conditions that may be signifi-
cant. Throwing some light on this question, it has recently
been found that nonmechanical end conditions (using
suction tubes downstream of the body) can yield rather
"clean" end conditions as shown visually in Fig. 9 [23],
without the large unsteady flow structures at the cylinder
spanwise ends that are normally present (see Fig. 4). With
the conditions in Fig. 9a, it was found that the laminar
Vortices regime for parallel shedding could be extended up to
Re = 194, as shown in Fig. 9b. A direct comparison may
be made between transition in flows with the same angle
of shedding: For parallel shedding with "contaminating"
I
end conditions in Fig. 4b, one has a transition at around
Figure 6. Patterns derived from a simple model using "nor- Re = 180; for parallel shedding in the absence of such
mal-wavelength constant." (a) A phase shock is induced by a end contamination, transition can be inhibited until Re =
large positive angle of shedding 01 (from the left-hand end 194.
condition) gradually taking over a region of smaller negative The above results suggest that transition is triggered
angle of shedding - 0 2. The shock translates at constant early due to end contamination, and in the absence of
speed to the right. (b) A phase expansion results (in this other effects it would appear that end contamination
example) when there is an impulsive change at the right-hand could account for the large scatter in the quoted critical
end. Initially this end induces a large negative angle -01, Reynolds numbers for transition reported over the last 40
which is then superseded by (for example) parallel shedding years. Physically, this contamination takes the form of
0 2 = 0. A phase expansion will result, from the right-hand regions of vortex dislocations moving across the span (this
end, whenever 02 > 0]. phenomenon is discussed later). It is not possible, given
nication, 1989; [35]) further confirm the existence of the the available evidence, to state that oblique shedding has
two modes A and B at very large aspect ratios (L/D> a naturally earlier transition than parallel shedding, be-
2000). cause for oblique shedding to occur the end conditions
necessarily contain a lower frequency end cell and thereby
End Contamination Effects: Cause of the Large also the presence of contaminating "dislocations."
Discrepancy in Critical Re (140-190)
Vortex Loops and Streamwise Vortices in Transition:
Although there is at least some small range of quoted "Small-Scale" Structures
values of the critical Reynolds numbers for the onset of
vortex shedding (Re from around 45 to 50), there is a Surprisingly few measurements have been made in the
relatively immense range of quoted critical Reynolds num- transition regime to date, and even fewer flow visualiza-
158 C . H . K . Williamson
i i i i i i , , , i , i
B
Ssr0o..'~o"wo'*%- I.~.- .- °
Figure 7. Strouhal-Reynotds num- 0.20
ber relationship: laminar and 3-D °o .~Oo°
oq,
transition regimes. The transition
regime is characterized by two dis-
tinct discontinuities in the mea-
sured wake parameters, as Re is j_...ca ~''- .~ o . . ~ - A
increased, and may be conveniently 0.18 ~'-~ "~"J" - I 2 nd D I S C O N T I N U I T Y
interpreted with reference to the 'A..LLEL S.EOO,N /..""" I
Strouhal-Reynolds number rela- //~.:"'" I st D I S C O N T I N U I T Y
tionship. At the first discontinuity,
the Strouhal number drops from x / _ j
the laminar shedding curve to a 0.16 •~° 0%
curve for mode A turbulent vortex
shedding. This first discontinuity is
hysteretic. At the second disconti-
nuity, there is a jump to a higher
Strouhal curve, corresponding to 0.14
mode B turbulent vortex shedding.
In this case, there is no hysteresis,
but instead a gradual transfer of
energy between modes A and B as
Re is increased. (From Williamson
[33].) 0.12 60 100 140 180 220 260

tion studies, although some new phenomena in this regime self-sustaining, due to Biot-Savart induction from one
have been discovered in the last few years. Based on his loop to the next, causing a whole string of loops at the
experimental velocity measurements, Roshko [3] sug- same spanwise position. Mode A and its lengthscale have
gested that transition to turbulence existed in the separat- similarity with the "in-phase" mode of vortex loop forma-
ing shear layers before the vortices were fully formed and tion in an unseparated wake studied by Meiburg and
shed from the cylinder. High-frequency oscillations were Lasheras [39].
later detected in the separating (and transitioning) shear At higher Reynolds numbers, when the Strouhal fre-
layers by Bloor [4]. Wei and Smith [36] observed in 1986 quencies lie on curve B in Fig. 7 (i.e., after the second
that "secondary" vortices were associated with these discontinuity), finer scale streamwise vortex pairs are
high-frequency oscillations and hypothesized that it is the formed, as shown in Fig. 10 (right half). In this case the
three-dimensional stretching of these secondary vortices primary vortex deformation is more spanwise-uniform than
that causes streamwise vortices, to appear in the wake. for mode A, and the streamwise vortex structure has a
However, although this scenario of shear layer transition markedly smaller spanwise wavelength of around one di-
is known to occur at Re > 1000, the brief but significant ameter, or one-fifth of a primary wavelength. Evidence of
visualizations of H a m a [7] showed that the instability in the vortex pair structure is shown clearly in the visualiza-
the wake transition regime takes the form of a three- tion in Fig. 10, using the laser-induced fluorescence tech-
dimensional waviness on the primary Karman vortices, nique. This photograph bears a striking resemblance to
and the formation of what Gerrard [10] later called "fingers the vortex pairs visualized in a mixing layer by Bernal and
of dye." Shear layer transition itself is a separate phe- Roshko [38]. A measure of the spanwise lengthscale of the
nomenon in a higher Re regime. It is now known that two instabilities, modes A and B, is given in Williamson
these dye "fingers" are associated with vortex loops and [34], where it is shown that there is a discontinuous
streamwise vortices [33, 34], in similarity with other free reduction in lengthscale from three-fifths to one-fifth of a
shear flows. primary wavelength as one passes through the second
Two different modes of 3-D shedding in wake transi- discontinuity (Re = 230-250). A central question pertains
tion, involving vortex loops and streamwise vortex pairs, to the origin of these three-dimensional instabilities, and
were demonstrated by Williamson [33, 34, 37], and these it seems from the jump in lengthscale that there exist two
have some analogy with the streamwise structure found in different (but related) 3-D instability phenomena, which
free shear layers (e.g., [38]) and also in the "unseparated" are discussed in [34].
wake that forms behind a splitter plate [39]. In mode A, At this point, we can return to the first visualizations in
corresponding to shedding frequencies along curve A in Fig. 1, where the cross-sectional views of the vortex shed-
Fig. 7, the primary vortices deform in a wavy fashion along ding illustrate the position of the streamwise vortices
their length during the shedding process, as visualized in shown, up to now, only in plan view. Due to the use of
Fig. 9 (left half). This results in the local spanwise forma- such an aluminum flake technique, and the orientation of
tion of vortex loops, which become stretched into stream- the light source in this case, the white filaments in the
wise vortex pairs. (That the two sides of a loop form a "braid" regions connecting the primary vortices indicate
vortex pair is proved in [34].) The spanwise length scale of streamwise vorticity. We may note that in Fig. 1 the top
these vortex loops is around three diameters, or three-fifths photograph, for the laminar shedding regime, shows no
of a primary wavelength. The process of loop generation is distinct white braid structure and thus no streamwise
Vortex Dynamics in Bluff Body Wakes 159

stretching and three-dimensional instability, as mentioned


,.:,,2, (a) above.
dB.
Spot-Like Vortex Dislocations in Transition:
"Large-Scale" Structure
-10
t° -30
-20 Small-scale structure in transition, as discussed above, is
perhaps to be expected. However, it is now found that
wake transition also involves structures that have a sur-
= 5 0 0 I-Iz. - - - - - - ~ - prisingly large structure, and this provides our focus for
this section. Large low-frequency irregularities in the wake
velocity fluctuations were first observed by Roshko [3] in
the transition regime. It was later suggested by Bloor [4],
on the basis of other work by Sato and Kuriki [40], that
B these low-frequency irregularities reflect the presence of
(b) three-dimensionalities in the flow that would render the
flow turbulent as it travels downstream. Almost no work
was undertaken to investigate the physical cause of this
phenomenon until it was discovered that these irregulari-
ties are due to the existence of vortex dislocations in wake
transition [37]. These structures are analogous with the
dS. crystal dislocations found in materials science. In the
present case, they are generated between spanwise cells of

t -10
different frequencies when the primary vortices move out
0
-20 of phase with each other. It is interesting that dislocations
or "defects" have also been found to be a fundamental
-30 aspect of free shear layer transition (Browand et al.,
1980-1993, see [41, 42]), and such structures could be
important to transition in general.
In the transition regime, measured profiles of velocity
fluctuation in the wake are distinctly different from those
associated with the laminar regime, as illustrated in Fig.
l l a . For example, the turbulent profile (Re = 183) shows
a large symmetrical peak in the centerplane of the wake,
as distinct from the two side peaks for the laminar regime
(Re = 152) usually associated with the two rows of lami-
nar vortices traveling downstream. The fluctuation energy
can be two orders of magnitude greater in a transitional
2 5 0 HZ. - - - . - - ~ , - wake than in a laminar vortex street wake, as evidenced by
the development of fluctuating velocity with downstream
Figure 8. Velocity spectra from the near wake in the transi- distance in Fig. l l b . It is found that the cause of such
tion regime. (a) Spectra at the first discontinuity, showing large fluctuation energy in the transition wake is related
hysteresis. Re = 172.8; L / D = 200. (b) Spectra at the sec- to the intermittent low-frequency irregularities first ob-
ond discontinuity, showing gradual transfer of energy from served by Roshko and shown from his 1967 paper [43] in
mode A to mode B, as Re increases. Re = 207-318; L / D = Fig. l lc. The traces shown below (from [37]) demonstrate
200. (From Williamson [33].) that such irregularities grow, as the wake travels down-
stream, into large low-frequency oscillations, in distinct
contrast with the much smaller regular oscillations re-
maining in the exponentially decaying laminar wake. It
vortex instabilities. For the higher Re = 300 and 700, one would seem that there must be structures associated with
can clearly observe the streamwise vortices in the braid such large oscillations.
regions between the primary Karman vortices. In the Evidence of large structure in transition can be ob-
separated wake behind a cylinder, the streamwise vorticity served from a number of visualizations in the literature,
originates mainly (but not wholly) from the deformation of including Cimbala et al. [44], H a m a [7], and Gerrard [10].
primary "core" vorticity, which is subsequently stretched By conducting close-up visualization simultaneous with
in the braid regions between the primary vortices. This is velocity measurements [37], it is possible to prove that
rather different from the origin of streamwise vortices in the irregular glitches in the velocity traces correspond
an unseparated wake, as described by Meiburg and with the passage of vortex dislocations past the velocity-
Lasheras [39], where the streamwise vortices originate in measuring device (hot wire). This suggests that the irregu-
the first instance from vorticity that has remained in the larities observed originally by Roshko [3] indeed corre-
braid region as a result of the primary roll-up process. spond with the formation of large-scale structure in the
Nevertheless, in both the separated and unseparated wake, form of vortex dislocations.
once significant spanwise vorticity is found in the braid To study their evolution in detail, dislocations have
region, it is then subject to strong streamwise vortex been (passively) forced to occur at a local spanwise posi-
160 C . H . K . Williamson

Figure 9. Extension of the laminar


shedding regime to Re > 194 by
minimization of end boundary con-
dition contamination. The results
shown suggest that end contamina-
tion is responsible for the immense
disparity among reported critical
Re values for wake transition (Re 0.21 I 1 t I I I I

= 140-190) measured over the last


40 years. By minimizing the con- S t e a d y (5 m i n or m o r e ) (b)
tamination of the flow from the o Transient
x
~xx x x X X x X x x
large unsteady fluctuations nor- 0.20 Williamson 1988 xx
x xxx
mally present at the ends of a
cylinder, it has proven possible to
r x
inhibit the spread of turbulent re-
gions along the span and thereby 0.19 •
x
t5 i
~
i
i x x
x x -x"
to retain laminar shedding at higher
Reynolds numbers. An example of S x x xx
x ~ Ie ~ • tNI~ Kxxxx
minimal e n d contamination result- xx x
ing from the use of an end suction 0.18 x
x

tube control technique (from Miller x


x

and Williamson [23]) is shown in


(a) for Re = 200. The "clean" end
conditions can be compared with 0.17
those in Fig. 4. The prolongation of
the laminar regime, and a substan-
tial hysteresis region, are indicated I a I ~ I , t , 1 , I , I
by the extended curve of laminar 0.16
shedding frequencies in the S - R e 120 140 160 180 200 220 240 260 280
plot in (b).

tion with the use of a small ring disturbance, creating grow to immense proportions on the order of 50 diameters
low-frequency periodic structures [37]. Forcing techniques (10 primary wavelengths) in width and 150 diameters (30
have been common for the study of boundary layer spots wavelengths) in streamwise extent. Evidence of vortex
and were used by Browand and Prost-Domasky [41] to dislocations in natural wake transition are shown typically
study defects in the mixing layer. In the wake, it was found in Fig. 12b, and it has been found that there are close
that "two-sided" dislocations are stable in a symmetric similarities with the forced dislocations shown, for exam-
in-phase configuration, as shown in Fig. 12a. As the wake ple, in Fig. 12a.
travels downstream, the energy at the low dislocation Vortex dislocations have been discovered in many other
frequency decays only slowly, in contrast to the rapid flows and can indeed be expected in a large variety of
decay of other frequencies, leaving the downstream wake flows, such as shear flows past cylinders; flows past a
dominated by large dislocation structures, as is found for nonuniform cylindrical geometry, such as a cone or stepped
the natural wake transition. It is found that across the cylinder; free shear flows (temporal, or spatial as for a
boundaries of the out-of-phase region there is a rather tilting tank); convection patterns; flows over sand dunes
contorted web of vortex linking. Dislocation structures can and ripples; and cloud patterns. Of particular significance
Vortex Dynamics in Bluff Body Wakes 161

Figure 10. Visualization showing vortex deformation process (mode A), and mushroom-shaped streamwise vortex pairs (mode
B). On the left, self-sustaining vortex loops forming at a local spanwise position behind a cylinder. The evolution of a typical
vortex loop can be seen by following the vortex marked with a blob, in photographs (a)-(e). Flow is from right to left, and the
spanwise extent of the vertical cylinder at the right edge is around three diameters. Dye visualization is used, at Re = 180. In
the right-hand photograph, we can see clearly the smaller-scale structures of mode B vortex shedding. Laser-induced
fluorescence of a thin sheet of light illuminates particularly these streamwise mushroom-shaped vortex pairs. The flow is left to
right, with a vertical cylinder at the left edge. Re = 250. (Photographs from Williamson [33, 34].)

in this discussion is that they have been shown to be lent wakes demonstrated the decay of the original Karman
fundamental to the growth of free shear flows by the work street wake and the growth in the far wake of a larger
by Browand and coworkers [41, 42] and recently by Ny- scale secondary vortex street that he suggested, arises out
gaard and Glezer [45]. It would seem conceivable that of a hydrodynamic instability based on the local mean
vortex dislocations could be a generic feature of transition velocity profile. In contrast, Matsui and Okude [47] sug-
in all shear flows. gested that such a secondary street is generated by vortex
amalgamations or pairing of the original Karman vortices
W A V E I N T E R A C T I O N S IN T H E FAR W A K E into larger vortical structures. In a computational study,
Meiburg [48] showed that vortex pairing would be able to
Although there have been a large number of investiga- account for the growth of a secondary vortex street.
tions concerned with the near wake behind a body, there An extensive experimental study in the laminar regime
are relatively few papers whose focus has been to study by Cimbala in 1984 [49] and by Cimbala et al. in 1988 [44]
the structure of far wakes. One of the central questions is suggested that the far-wake structure does not depend
whether there is a connection between the flow and vortex directly on the precise scale or frequency of Karman
dynamics right behind a body (in the near wake) and the vortices shed from the cylinder. The fact that frequencies
structure that is found far downstream of a body (in the unrelated to the Karman shedding frequency (fK) were
far wake). In other words, one might question whether the amplified in the far wake (and have good agreement with
far wake retains a "signature" of the details of the near- those predicted from linear stability theory) strongly sug-
wake dynamics. In both the laminar and turbulent wake gests that hydrodynamic instability is the principal mecha-
regimes, the width of a nominally two-dimensional far nism for secondary street growth. A particularly interest-
wake grows as x 1/2, for large values of downstream dis- ing result from the study of Cimbala et al. was found from
tance (x), and we thus expect the size of the large wake spanwise visualizations, where they discovered a honey-
structures to increase, while the passage frequency of comb-like cellular 3-D pattern in the far wake. They
these structures should decrease. Some original observa- suggested that this 3-D pattern could be caused by a
tions by Taneda in 1959 [46] for both laminar and turbu- secondary 3-D parametric instability of the subharmonic
162 C . H . K . Williamson

(a) ~ Re= 183

u~=, 0.08
U~ ~ ~.~ Re= 152 R O S H K O (1967)

x/D = 4 0
0 ,
- -4 0 4

0.25
y/D (c)
(b)
0.20

/'trms 0.15 ~.,,..~'~ Re= 248 x/D = 5

U~
o.Io ~ ~
0.05 / -,,~
Re= 183 Re= 152 Re= 183

Re= 152
LAMINAR WAKE TURBULENT WAKE
0 10 20 30 40 50
Wl L L I A M S O N (1992)
x/D
Figure 11. Velocity measurements in the transitional wake. (a) Velocity fluctuation profile measured in both the laminar
regime (Re = 152) and the wake transition regime (Re = 183). The twin peaks for Re = 152 are indicative of the two rows of
vortices traveling downstream past the hot-wire probe. The profile for Re = 183 is distinctly different, with a much larger
central peak. In this case, normalized rms velocity is plotted versus normalized transverse distance y / D ( x / D = 10). (b)
Downstream decay of normalized rms velocity fluctuations, showing the distinctly different rates of decay in the laminar regime
(Re = 152) versus the transition regime (Re = 183, 248). Rather than the exponential decay in the laminar case, the transition
regime exhibits a plateau-like region of much slower decay, at least down to x / D = 50. (Data for (a) and (b) from Williamson
[37].) (c) Large low-frequency intermittent velocity fluctuations in the transition regime explain the large fluctuation energy
measured in the profile and downstream decay plot shown in (a) and (b). Such large fluctuation irregularities were first
observed by Roshko in 1954, and a typical velocity trace (black) from his 1967 paper [43] is included here. The contrast between
laminar (on the left) and turbulent velocity traces (on the right) is particularly noticeable further downstream at x / D = 40, in
the white traces (Williamson [37]).

type acting on the far-wake initially 2-D waves. Further spanwise waviness on the downstream parallel waves. Some
investigations that involve forcing 3-D structure in far careful measurements demonstrate that the spanwise
wakes were undertaken by Lasheras and Meiburg [50] and wavelength of the downstream waviness is equal to the
by Corke et al. [51], both of whom consider the interac- spanwise wavelength of the oblique shedding vortices. The
tions of subharmonic pairs of oblique waves with a pri- results of Williamson [53] and Williamson and Prasad [54]
mary 2-D wave. However, it has remained unclear, until demonstrate that a honeycomb pattern can be formed as
recently, whether such a secondary instability actually the direct result of an interaction between oblique shed-
occurs in "naturally" observed wakes. Although it is rea- ding vortices and 2-D large-scale waves that grow in the
sonably apparent from simple observation of their smoke far wake. With parallel (2-D) shedding upstream, no hon-
visualization that such a pattern is caused by parallel and eycomb pattern is found downstream, which proves
oblique waves interacting, one might ask, From where do straightforwardly that the far-wake 3-D patterns observed
these waves originate? We shall see that there is a direct to date are dependent on the presence of oblique shed-
link between the near-wake phenomenon of oblique shed- ding "waves" in the wake. It clearly follows, from this
ding and the far-wake honeycomb pattern of Cimbala et particular type of o b l i q u e - - 2 - D wave interaction, that the
al. spanwise wavelengths upstream and downstream must be
In agreement with the conclusions of Cimbala et ai. the same, which is consistent with the meticulous mea-
[44], Hammache and Gharib [52] stated that no direct surements of Hammache and Gharib. There would seem
relationship should be expected between the frequencies to be little question, from the further available evidence
in the primary and secondary regions. They found that the demonstrated in [55], that the honeycomb pattern of Cim-
secondary vortices are parallel (2-D) irrespective of bala et al. is also caused by their oblique shedding waves
whether there is oblique shedding or parallel shedding interacting with growing 2-D waves in the far wake.
upstream, although with oblique shedding there exists a Further research now indicates that in "natural" far
Vortex Dynamics in Bluff Body Wakes 163

a b
Figure 12. "Vortex dislocations": origin of low-frequency fluctuations. Recent research has shown that the origin of the large
low-frequency irregularities in wake transition, first observed by Roshko [3], is the growth of large vortical structures in the
wake. (a) Forced two-sided vortex dislocation. An enormous vortex dislocation in this case is induced to occur at a local
spanwise position in the wake by placing a small "ring" disturbance around the cylinder. This has the effect of causing a slightly
lower frequency of vortex shedding behind the ring relative to either spanwise side, and thus a sequence of periodic large-scale
dislocations ensue. The streamwise extent of the symmetric dislocation is around 20 primary wavelengths and is therefore very
much larger than that of the small-scale wake instabilities [37]. Flow is upward. (b) Natural vortex dislocations. Visualization in
plan view shows that the large fluctuations in natural wake transition are due to large structures known as vortex dislocations
[37]. Re = 210. Flow is upward. The development of clusters of these structures in transition has strong similarities with vortex
defects in free shear layers studied by Fred Browand and coworkers over the last decade (see, e.g., [42]).

wakes, not only the scale but also the frequency are appears that one needs only a remarkably small free-
precisely dependent on the near wake (Williamson and stream spectral peak (of order U'rms/U = 0.00005) to trig-
Prasad [56]). Surprisingly, the characteristic that actually ger an amplification in the far wake. The disturbance in
forges a "connection" between the near and far wakes is the free stream of a wind tunnel facility are generally of
the sensitivity to free-stream disturbances. Indeed, the far an acoustic origin and are thus generally 2-D waves, with
wake, being a convectively unstable flow, will tend to amplitudes typically of this order.
amplify extremely small frequency peaks in the free Measurements of frequency under conditions of far-
stream. With an understanding of the exquisite sensitivity wake resonance indicate that the predominant far-wake
of the far wake to free-stream disturbances, a new mecha- frequency is a combination frequency (f~¢ - fT), where fK
nism of "oblique wave resonance" in the far wake has is the Karman frequency upstream and fT is the fie-
recently been discovered [56, 57] that is distinct from the quency of the 2-D wave. That this frequency corresponds
parametric subharmonic instabilities discussed above. A to oblique resonance waves is apparent if one interprets
nonlinear (quadratic) interaction between the oblique the phenomenon as an example of the type of resonance
shedding waves generated from upstream and the 2-D in fluid mechanics that has been described in the book by
waves amplified in the far wake triggers the growth of an Craik [58]. For such resonance, it is required that the
oblique resonance wave, as shown in Fig. 13a, where the following two waves travel at the same phase speed:
smoke was introduced into the wake at x / D = 50 in
order to highlight the three wave systems. The photograph alc ei(,~xx+13Kz--,oKt), (4)
in Fig. 13b shows that if the smoke wire is located further
downstream at x / D = 100, then what one observes is
almost wholly the strong oblique resonance waves. It
aT ei(aTx +13Tz COT/), (5)
164 C . H . K . Williamson

Figure 13. A mechanism of oblique


wave resonance in the far wake.
This phenomenon is due to the
nonlinear interaction between
oblique shedding waves from up-
stream and the large-scale 2-D
waves that evolve from instability
in the far wake. Flow is left to
right. With the smoke wire located
at x/D = 50 in (a), one can ob-
serve at the left the oblique shed-
ding waves. These waves interact
with the 2-D waves in the middle
of the picture to produce the
large-angle oblique resonance
w,aves to the right. The lower pho-
tograph in (b) shows that if the
smoke is introduced further down-
stream at x/D = 100, then what
one observes is almost wholly the
strong oblique resonance waves.
(Note that the orientation of the
waves is different due to the dif-
ferent orientation of the oblique
shedding waves in these two exper-
iments.) In essence, two waves in-
teract nonfinearly to produce a
third wave, which can grow to
"swamp" the first two waves. (From
Williamson and Prasad [54, 56].)

where x and z are the distances downstream and span- what is found experimentally, namely, the resonance waves
wise, respectively. Subscript K refers to the K a r m a n shed- passing through the nodes formed by the intersection
ding waves, and subscript T to the 2-D waves. With such a pattern of the other two waves. A n investigation into the
two-wave interaction, one may expect quadratic terms on typical downstream development of velocity spectra in Fig.
the o r d e r of a 2 to a p p e a r (e.g., aKa T, aKa~). Let us 14 demonstrates that although the 2-D waves (frequency
denote a third wave corresponding to aKa ~ as fT) are involved in triggering the oblique wave resonance
a3ei(a3 x+ ~3 z- t°3t), (6) at around x / D = 100, they are almost lost in the back-
ground noise at x / d = 300, whereas the oblique waves
where the following relationships may then be written: (fK - - f v ) are still relatively energetic, decaying only very
slowly as the wake travels further downstream. A clear
60 3 = (..OK - - O ) T , (7) schematic illustration of the oblique wave resonance phe-
nomenon, summarizing in a simple manner the fundamen-
f13 = flK -- fiT' (8)
tal concept of the two-wave interaction, is provided in
Ot 3 : OtK - - O / T . (9) Fig. 15.
Since these resonances are caused by extremely small
A resonance exists if such a relationship between three
waves holds true, and this is precisely what is found to acoustic disturbances present in the free stream, the non-
exist in this far-wake flow. Relation (7) shows that the linear oblique wave resonance has been induced by the
resonance wave has a frequency fK -- fT, and relations (8) use of 2-D wave acoustic forcing, Williamson and Prasad
and (9) give the geometry of the oblique waves as precisely [57]. A whole set of oblique resonance modes have been
Vortex Dynamics in Bluff Body Wakes 165

responsible for the regular 3-D patterns and for the phe-
nomenon of oblique wave resonance that evolves in the
oblique waves
far wake. This sensitivity to disturbances makes it difficult
l 2-D waves Karmanshedding to properly define a "natural" wake, as its structure is
exquisitely dependent on the environment in which it is
studied. In analogy with the mixing layer: The natural far
wake is by nature an excited flow.

C O N C L U D I N G REMARKS

I I /i \
Despite the fact that the wake of a bluff body does not
easily admit analytical approaches, it is exceedingly rich in
flow phenomena. Over the last six years, there has been a
wave of experimental discoveries concerning several as-
pects of bluff-body wakes, particularly three-dimensional
aspects. Cellular shedding, vortex dislocations, oblique
shedding, phase shocks and expansions, vortex loops, and
streamwise vortices are all three-dimensional vortex dy-
100 namics phenomena that are becoming understood. As a
part of understanding these phenomena, there has been a
I new appreciation of the effects of end boundary condi-
tions influencing the flow over long cylinder spans. From
all of this, one might question what it is that we are trying
to understand: Is it the flow over an infinitely long body,
or the flow over a body with finite span and therefore
under the influence of end boundary conditions? There
are scientific and engineering reasons to consider both of
these as goals.
In considering the wake vortex dynamics in this paper,
one may define two kinds of three-dimensionality, as
discussed in the 1992 ONR Wakes Workshop and further
in Roshko [59] and Williamson [60]. One may be called
extrinsic, for example, effects due to end conditions; the
Other may be called intrinsic, for example, 3-D motions
arising from natural instabilities. One may perhaps con-
Figure 14. Development of downstream velocity fluctuations sider that extrinsic end effects are extraneous to the flow.
for oblique wave resonance. The velocity spectra demonstrate On the other hand, one may alternatively view oblique
that as the wake travels downstream, the energy of the shedding and parallel shedding as equally intrinsic phe-
oblique Karman shedding frequency fK dies out (exponen- nomena; they are both solutions of the flow over a body
tially) but nevertheless interacts nonlinearly with the 2-D but are simply the result of different boundary conditions.
waves (frequency fT) sufficiently to trigger the oblique reso- Following the results emerging from studies on the ring
nance wave (frequency f K - fx)- This resonance wave then cylinder wake with no end effects, studied by the group at
grows to become by far the dominant wave further down- Marseille (Provansal, Leweke, and Boyer), the matter is
stream at x / D = 300. (From Williamson and Prasad [54, not so clear-cut, since stable wake configurations comprise
56].) both helical oblique shedding modes and vortex ring
(parallel shedding) modes. In this case, the oblique helical
modes are as intrinsic as the parallel vortex ring modes.
found that correspond to frequencies fK -- rifT, where n Perhaps we should view all of the three-dimensional vor-
is an integer. Surprisingly large angles for the oblique tex dynamics phenomena pertaining to uniform long cylin-
resonance waves (Ores up to at least 75 °) have been found ders in uniform flows as indeed intrinsic phenomena.
to be preferentially amplified over the 2-D waves that Finally, one might consider some of the many outstand-
trigger the resonance. Paradoxically, the effect of an in- ing questions remaining on the problems of bluff-body
crease in amplitude (A) of the 2-D wave forcing is to flows. A central question is, To what extent do the three-
further amplify the oblique resonance waves in preference dimensional vortex dynamics phenomena recently found
to the 2-D waves. for low Reynolds numbers carry over to high Reynolds
It seems that although the general approach to under- number flows? How do the 3-D phenomena influence the
standing the far wake has hitherto followed classical para- steady and, more particularly the unsteady, fluid forces on
metric subharmonic studies, there are distinct differences a cylinder? There is the question of whether we can use
between these theories and the actual wave interactions our improved knowledge regarding 3-D flow phenomena
found in wake experiments to date, as described above. It for uniform cylinders in uniform flows in the cases of
also seems clear, and is indeed to be expected, that the far other more complex flows such as flows with active/pas-
wake has an extreme sensitivity to free-stream distur- sive control, nonuniform problems such as cone wakes
bances. It is these disturbances that not only lead to a and cylinders in shear flows, bodies in oscillatory and wave
large scatter in previous wake measurements but also are flows, and bodies of very small aspect ratio, which have
166 C . H . K . Williamson

Body produces
fx
Decaying

7' I
I
Free-streamproduces fr

'1
l
l I
2D waves
I I I | I
Growing

The interactionproduces fx-fr


Figure 15. Schematic diagram demonstrating
fundamental concept of two-wave interaction Rapidly growing
(fK and fT) yielding an oblique resonance wave

recently drawn the attention of Norberg [35] and Szepessy annual workshops), of which the present work at Cornell is a part
[611. (ONR contract No. N00014-90-J-1686).
Modern techniques are certainly being brought to bear
in the experiments, particularly the use of standard and
digital PIV methods to determine velocity and thereby NOMENCLATURE
vorticity fields. Although further development is taking a amplitude of wave,
place with these methods, a main challenge at present is A amplitude of wake oscillator,
to develop a three-dimensional D P I V technique. In analy-
D diameter of cylinder,
sis, there continues to be research in stability theory and
in the use of amplitude equations to characterize the f shedding frequency,
three-dimensional wake patterns in the near wake. Analyt- q spanwise wavenumber, dimensionless
ical efforts are also under way to explain the phenomenon Re Reynolds number, dimensionless
of oblique wave resonance in the far wake. In computa- S Strouhal number, dimensionless
tions, there has been a great deal of progress in develop- t time, s
ing numerical techniques such as vortex methods, finite U free-stream velocity
difference schemes, and spectral methods in two dimen- x downstream coordinate, dimensionless
sions. One of the technical challenges at present appears
to be in pushing the three-dimensional computations to z spanwise coordinate, dimensionless
higher Reynolds numbers, larger aspect ratios, and for Greek Symbols
longer times. a downstream wavenumber, dimensionless
There is clearly much work still to be done to discover /3 spanwise wavenumber, dimensionless
and understand three-dimensional vortex dynamics in w radial frequency, r a d / s
wake flows, over and above the development of the new A coefficient, dimensionless
techniques in experimentation, analysis, and simulation.
/z coefficient, dimensionless
v coefficient of viscosity
I wish to thank Greg Miller, Anil Prasad, and Chantal Champagne, ~r growth rate
Ph.D., for superb assistance with this paper. The continued support 0 oblique shedding angle
from the Ocean Engineering Division of the Office of Naval Re-
search, monitored by Dr. Tom Swean, is gratefully acknowledged.
Subscripts
Much of the recent surge of research in this area has directly K Karman shedding wave
resulted from an ONR "Bluff Body Wake Vortex Dynamics and T two-dimensional wave
Instabilities" Accelerated Research Initiative (with corresponding 0 oblique
V o r t e x D y n a m i c s in Bluff Body W a k e s 167

3 third wave 24. Squire, H. B., On the Stability for Three-Dimensional Distur-
1 r e f e r s to o t h e r c o e f f i c i e n t s bances of Viscous Fluid Flow Between Parallel Walls, Proc. Roy.
0 two-dimensional Soc. Lond. A142, 621, 1933.
25. Koenig, M., Noack, B. R., and Eckelmann, H., Discrete Shedding
Modes in the yon Karman Vortex Street, Phys. Fluids A5, 1846,
REFERENCES 1993.
1. Strouhal, V., Uber eine besondere Art der Tonerregung, Ann. 26. Triantafyllou, G. S., Three-Dimensional Flow Patterns in Two-
Phys. Chem. (Liepzig), Neue Folge, 5(10), 216, 1878. Dimensional Wakes, J. Fluids Eng., Trans. A S M E 114, 356, 1992.
2. Rott, N., Lord Rayleigh and Hydrodynamic Similarity, Phys. 27. Noack, B. N., Ohle, F., and Eckelmann, H., On Cell Formation in
Fluids 4, 2595, 1993. Vortex Streets, J. Fluid Mech. 227, 293, 1991.
3. Roshko, A., On the Development of Turbulent Wakes from 28. Albarede, P., Provansal, M., and Boyer, L., Modelisation par
Vortex Streets, NACA Rep. 1191, 1954. l'equation de Guinzburg-Landau du sillage tri-dimansionel d'un
4. Bloor, M. S., The Transition to Turbulence in the Wake of a obstacle allonge, Compt, Rend. Acad. Sci. Paris 310 (Ser II), 459,
Circular Cylinder, J. Fluid Mech. 19, 290, 1964. 1990.
5. Berger, E., and Wille, R., Periodic Flow Phenomena, Ann. Rev. 29. Albarede, P., and Monkewitz, P., A Model for the Formation of
Fluid Mech. 4, 313, 1972. Oblique Shedding Patterns and "Chevrons" in Cylinder Wakes,
6. Berger, E., Transition of the Laminar Vortex Flow to the Turbu- Phys. Fluids A4, 744-756, 1992.
lent State of the Karman Vortex Street Behind an Oscillating 30. Park, D. S., and Redekopp, L. G., A Model for Pattern Selection
Cylinder at Low Reynolds Number (in German), Jahrb. Wiss. in Wake Flows, Phys. Fluids A4, 1697, 1991.
Gess. L. R. 1964, 164, 1964. 31. Chiffaudel, A., Nonlinear Stability Analysis of Two-Dimensional
7. Hama, F. R., Three-Dimensional Vortex Pattern Behind a Circu- Patterns in the Wake of a Circular Cylinder, Europhys. Lett. 18,
lar Cylinder, J. Aeronaut. Sci. 24, 156, 1957. 589, 1992.
8. Tritton, D. J., Experiments on the Flow Past a Circular Cylinder 32. Monkewitz, P. A., Williamson, C. H. K., and Miller, G. D.,
at Low Reynolds Numbers, J. Fluid Mech. 6, 547, 1959. Three-Dimensional Phase Dynamics in a Wake, Phys. Fluids 8, 1,
9. Gaster, M., Vortex Shedding from Circular Cylinders at Low 1996.
Reynolds Numbers, J. Fluid Mech. 46, 749, 1971. 33. Williamson, C. H. K., The Existence of Two Stages in the
10. Gerrard, J. H., The Wakes of Cylindrical Bluff Bodies at Low Transition to Three-Dimensionality of a Cylinder Wake, Phys.
Reynolds Number, Phil. Trans. Roy. Soc. A288, 351, 1978. Fluids 31, 3165, 1988.
11. Roshko, A., Modelling, Past and Future, Comment paper for 34. Williamson, C. H. K., The Transition to Three-Dimensionality ih
Whither Turbulence Workshop, Cornell Univ., Ithaca, NY, 1989. the Wake of Cylinders, J. Fluid Mech., submitted.
12. Sreenivasan, K. R., Transition and Turbulence in Fluid Flows 35. Norberg, C., An Experimental Investigation of the Flow Around
and Low-Dimensional Chaos, in Frontiers in Fluid Mechanics, a Circular Cylinder: Influence of Aspect Ratio, J. Fluid Mech.
S. H. Davis and J. L. Lumley, Eds., pp. 41-66, Springer, New 258, 287, 1994.
York, 1985. 36. Wei, T., and Smith, C. R., Secondary Vortices in the Wake of
13. Van Atta, C., and Gharib, M., Ordered and Chaotic Vortex Circular Cylinders, J. Fluid Mech. 169, 513, 1986.
Streets Behind Circular Cylinders at Low Reynolds Numbers, J. 37. Williamson, C. H. K., The Natural and Forced Formation of
Fluid Mech. 174, 113, 1987. Spot-Like "Vortex Dislocations" in the Transition of a Wake, J.
14. Karniadakis, G. E., and Triantafyllou, G. S., Frequency Selection Fluid Mech. 243, 393, 1992.
and Asymptotic States in Laminar Wakes, J. Fluid Mech. 199, 38. Bernal, L. P., and Roshko, A., Streamwise Vortex Structure in
441, 1989. Plane Mixing Layers, J. Fluid Mech. 170, 499, 1986.
15. Williamson, C. H. K., Defining a Universal and Continuous 39. Meiburg, E., and Lasheras, J. C., Experimental and Numerical
Strouhal-Reynolds Number Relationship for the Laminar Vor- Investigation of the Three-Dimensional Transition in Plane
tex Shedding of a Circular Cylinder, Phys. Fluids 31, 2742, 1988. Wakes, J. Fluid Mech. 190, 1, 1988.
16. Williamson, C. H. K., Oblique and Parallel Modes of Vortex 40. Sato, H., and Kuriki, K., The Mechanism of Transition in the
Shedding in the Wake of a Circular Cylinder at Low Reynolds Wake of a Thin Flat Plate Placed Parallel to a Uniform Flow, J.
Numbers, J. Fluid Mech. 206, 579, 1989. Fluid Mech. 11, 321, 1961.
17. Eisenlohr, H., and Eckelmann, H., Vortex Splitting and Its Con- 41. Browand, F. K., and Prost-Domasky, S. A., Experiment on Pat-
sequences in the Vortex Street Wake of Cylinders at Low tern Evolution in the 2-D Mixing Layer, in New Trends in
Reynolds Number, Phys. Fluids A1, 189, 1989. Nonlinear Dynamics and Pattern Forming Phenomena, P. Coullet
18. Koenig, M., Eisenlohr, H., and Eckelmann, H., The Fine Struc- and P. Huerre, Eds., NATO ASI Ser. 8:159. Plenum, New York.
ture in the S-Re Relationship of the Laminar Wake of a 42. Browand, F. K., and Troutt, T. R., A Note on the Spanwise
Circular Cylinder, Phys. Fluids A2, 1607, 1990. Structure in the Two-Dimensional Mixing Layer, J. Fluid Mech.
19. Hammache, M., and Gharib, M., A Novel Method to Promote 97, 771, 1980.
Parallel Shedding in the Wake of Circular Cylinders, Phys. Fluids 43. Roshko, A., Transition in Incompressible Near-Wakes, Phys.
A1, 1611, 1989. Fluids (Suppl.), S181, 1967.
20. Gerich, D., and Eckelmann, H., Influence of End Plates and Free 44. Cimbala, J. M., Nagib, H. M., and Roshko, A., Large Structure in
Ends on the Shedding Frequency of Circular Cylinders, J. Fluid the Far Wakes of Two-Dimensional Bluff Bodies, J. Fluid Mech.
Mech. 122, 109, 1982. 190, 265, 1988.
21. Koenig, M., Eisenlohr, H., and Eckelmann, H., Visualization of 45. Nygaard, K. J., and Glezer, A., Core Instability of the Spanwise
the Spanwise Cellular Structure of the Laminar Wake of Wall- Vortices in a Plane Mixing Layer, Phys. Fluids A2, 461, 1990.
Bounded Circular Cylinders, Phys. Fluids A4, 869, 1992. 46. Taneda, S., Downstream Development of Wakes Behind Cylin-
22. Hammache, M., and Gharib, M., An Experimental Study of the ders, J. Phys. Soc. Jpn. 14, 843, 1959.
Parallel and Oblique Vortex Shedding from Circular Cylinders, 47. Matsui, T., and Okude, M., Vortex Pairing in a Karman Vortex
J. Fluid Mech. 232, 567, 1991. Street, Presented at Seventh Biennial Symp. Turbulence, Rolla,
23. Miller, G. D., and Williamson, C. H. K., Control of Three- MO, 1981.
Dimensional Phase Dynamics in a Cylinder Wake, Exp. FluMs 48. Meiburg, E., On the Role of Subharmonic Perturbations in the
18, 26, 1994. Far Wake, J. Fluid Mech. 177, 83, 1987.
168 C.H.K. Williamson

49. Cimbala, J. M., Large Structure in the Far Wakes of Two- 55. Williamson, C. H. K., and Prasad, A., Oblique Wave Interactions
Dimensional Bluff Bodies, Ph.D. Thesis, Graduate Aeronaut. in the Far Wake, Phys. Fluids AS, 1854, 1993.
Lab., California Institute of Technology, Pasadena, CA, 1984. 56. Williamson, C. H. K., and Prasad, A., A New Mechanism for
50. Lasheras, J. C., and Meiburg, E., Three-Dimensional Vorticity Oblique Wave Resonance in the "Natural" Far Wake, J. FluM
Modes in the Wake of a Flat Plate, Phys. Fluids A2, 371, 1990. Mech. 256, 269, 1993.
51. Corke, T., Krull, J. D., and Ghassemi, M., Three-Dimensional 57. Williamson, C. H. K., and Prasad, A., Acoustic Forcing of Oblique
Mode Resonance in Far Wakes, J. Fluid Mech. 239, 99, 1992. Wave Resonance in the Far Wake, J. Fluid Mech. 256, 315, 1993.
52. Hammache, M., and Gharib, M., On the Evolution of Three- 58. Craik, A. D. D., Wave Interactions in Fluids, Cambridge Univ.
Dimensionalities in Laminar Bluff Body Wakes, in Proceedings of Press, New York, 1982.
the IUTAM Conference on Bluff Body Wake Instabilities, H. Eckel- 59. Roshko, A., Perspectives on Bluff Body Aerodynamics, Presented
mann, J. M. R. Graham, P. Huerre, and P. A. Monkewitz, Eds., at 2nd Int. Colloq. on Bluff Body Aerodyn. Appl., Melbourne,
pp. 349-357, Springer-Verlag, Berlin, 1992. Australia, 7-10 Dec., 1992.
53. Williamson, C. H. K., Wave Interactions in the Far Wake, in 60. Williamson, C. H. K., Vortex Dynamics in the Wake of a Cylin-
Proceedings of the IUTAM Conference on Bluff Body Wake Instabil- der, in Fluid Vortices, S. I. Green, Ed., Kluwer. Chapter 5.
ities, H. Eckelmann, J. M. R. Graham, P. Huerre, and P. A. 61. Szepessy, S., On the Control of Circular Cylinder Flow by End
Monkewitz, Eds., pp. 333-337, Springer-Verlag, Berlin, 1992. Plates, Eur. J. Mech. B, Fluids 12, 217, 1993.
54. Wiiliamson, C. H. K., and Prasad, A., A Mechanism for Oblique
Wave Resonance in the Far Wake, Presented at IUTAM Symp.
Nonlinear Instability of Nonparallel Flows, Clarkson College,
Potsdam, NY, 1993. Accepted August 24, 1995

Вам также может понравиться