Guidelines For Authors 1

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Guidelines for authors 1

This draft was prepared using the LaTeX style file belonging to the Journal of Fluid Mechanics 2

Zero crossing in turbulent planejet


Pramod Kumar1 †, O.N.Ramesh
1
Department of Aerospace Engineering, Indian Institute of Science, Bangalore

(Received xx; revised xx; accepted xx)

1. Introduction
The interval between two successive up and down crossing of a stationary random
process u(t) which has its own zero mean, called zero crossing, is a random variable. Let
us denote this variable by η. Much experimental work has been devoted to determining
the statistical properties of η for turbulent signals in several flows. The most detailed work
on zero crossings has occurred for velocity signals in the turbulent boundary layer. In
these papers, the temporal variable has been converted to a spatial variable via Taylor’s
frozen-flow hypothesis which assumes that turbulence convects with the mean velocity
without distortion. Let us denote the resulting zero-crossing variable by Λ. At all y within
the boundary layer,the mean value Λ of the zero-crossing variable η for the streamwise
velocity fluctuation u is approximately equal to the Taylor microscale, λ. The Taylor
microscale is defined by, following liepmann,
1/2
⟨u2 ⟩

λ=U (1.1)
⟨(∂u/∂t)2 ⟩
where the angular brackets indicate time averages, and U is the time average velocity
at the position under consideration; by definition, the fluctuation velocity u has zero
mean. The above result is analogous to the result by Rice (1945) which for stationary
stochastic signal y = y(x) gives frequency N0 of zero crossing as,

N02 = ẏ 2 /4π 2 y 2 (1.2)


if both y and ∂y/∂x have Gaussian distribution and statistical independence. Liepmann
proof of this result for turbulent velocity fluctuation gives independent indication of
viscous dissipation.

Λ = (2πN0 )−1 (1.3)


is defined as zero crossing (following liepmann) time scale where N0 is Zero crossing
frequency. Following Rice’s result, Λ = λ for any stationary Gaussian signal.
Since turbulence is almost always Non-Gaussian in nature, the Rice’s result to hold
true for turbulent signals can not be commented upon beforehand.

2. DNS of Planejet
This section describes the direct numerical simulations carried out in this work. The
simulations of Navier-Stokes equation is done using the fractional step method of Kim
† Email address for correspondence: jfm@damtp.cam.ac.uk
Guidelines for authors 3
5

y/h
0

-5
0 2 4 6 8 10 12 14 16 18 20

x/h

Figure 1: ωz in XY plane

4
z/h

0
0 2 4 6 8 10 12 14 16 18 20

x/h
Figure 2: ωz in XZ plane at Z centerline

and Moin (1985). The computational domain consists in a box with sizes Lx , Ly and
Lz along the streamwise (x), normal (y) and spanwise (z) directions, respectively. The
simulations were carried out with an incompressible Navier–Stokes solver developed by
Kim & Moin (1985). For temporal advancement, the algorithm uses an semi implicit RK3
method.
For boundary conditions, the code uses a convective outflow boundary condition in
the streamwise direction (x), while the spanwise (z) direction is treated as periodic. The
boundaries in the normal (y) direction were also treated as free boundary.

2.1. Physical and computational parameters

1.2, |y| ⩽ 0.5H



U (y) = (2.1)
0.2, |y| > 0.5H.
The DNS of planar turbulent jet was carried out on a domain with (Lx , Ly , Lz ) =
(40H, 30H, 8H), where Lx , Ly , Lz are the spatial extents of the computational domain
along the streamwise (x), normal (y) and spanwise (z) directions, respectively. The grid
is uniform in spatial directions x & z with ∆x = 0.0277H and ∆z = 0.033H. Grid is
stretched in y direction with denser refinement in the central region. ∆ymin = 0.024H.
The total number of grid points along these directions is N1 = 1440, N2 = 800 and
N3 = 240, respectively. The Reynolds number is ReH = UH/ν = 3000, where ν is the
kinematic viscosity, and U = U1 −U2 is the difference of maximum and minimum (co-flow)
jet velocities which were set to U1 = 1.2 and U2 = 0.2, respectively.
Figure 1 shows instantaneous contours of the spanwise vorticity, ωz , on an XY plane
through the domain at the centreline of Z plane. In the region 0 ⩽ x/h ⩽ 8.0 there is
clear indication of the presence of vortex roll up in both the upper and lower shear layers.
4 A. N. Other, H.-C. Smith and J. Q. Public

Figure 3: Regions of mixing transition

Figure 4: Regions of mixing transition in upper shear layer

Figure 5: λ2 = 0.05

In the upper shear layer, peaks in the magnitude of the spanwise vorticity are present at
x/h = 5.0 and x/h = 4.0.At x/h = 6.0 vortex merging is taking place in the upper shear
layer. A large engulfing of outer irrotational fluid is present downstream, and the same
process is observed in the middle of the domain, involving this time the potential core
of the jet immediately after the nozzle. This mechanism is responsible for the presence
of irrotational parcels of fluid inside the turbulent region, as shown by figure 2. The jet
undergoes a classical transition to turbulence prompted by the emergence of the primary
instabilities in the form of Kelvin–Helmholtz vortices discernible at x/H ≈ 3 which are
initially symmetric (in relation to the jet centreline plane) as predicted by the linear
stability theory for this case since H/θ = 30. Further downstream e.g. at x/H ≈ 5, the
Kelvin–Helmholtz rollers show small-scale eddies in their interior and streamwise eddies
start forming. These secondary structures quickly become fragmented and by x/H ≈ 9
vorticity structures exhibit a large range of scales and do not show the existence of any
Guidelines for authors 5

(a) X/H = 8.33 (b) X/H = 10.0

(c) X/H = 13.33 (d) X/H = 15.0

(e) X/H = 18.33 (f) X/H = 20.00

Figure 6: Zero crossing scale ratio to Taylor scale (Λ/λ), plotted against y/H:
• = ū′ ; · ·, ♦ = v̄ ′ , ▶ = w̄′ .

preferential direction, sign of a fully developed turbulent state. Figure 3 demonstrate


tubular structures of concentrated vorticity field forming at the both shear layers in
laminar and transition regions which merges downstream giving away to fully developed
turbulent region.
6 A. N. Other, H.-C. Smith and J. Q. Public
3. Results:
3.1. Zero crossing and Taylor scale
Λη is defined as zero crossing time scale which corresponds to number of times the
fluctuating components of velocity signal crosses the 0 mean value. λη is defined as
Taylor microscale obtained from the relation η̇ 2 = η 2 /λ2η . Figure 5 shows the ratio Λ/λ
for turbulent planejet signal along transverse direction. The zero cross frequency N0η
where η being the three fluctuation quantities, is given by λη = 2πN0η .
Ylvisaker (1965) proved that Λ = λ for any continuous stationary Gaussian process
y with finite λ without invoking statistical independence between y and ẏ. Statistical
independence between signal and its derivative was not invoked in reaching the above
conclusion.Turbulence by its very nature, is always non-Gaussian. Rice’s (1945) result,
which in principal, holds Gaussian data under analysis cannot a priori be expected to
hold. However various measurements for zero crossing on turbulent data does have some
interesting results. The measurements of Liepmann, Laufer & Liepmann (1951) have
shown that Λ = λ (withinn 20 %) in isotropic turbulence. The implication is that the
zero-crossing scales effectively with the turbulence-energy dissipation. The post-facto
explanation for this interesting finding of Liepmann et al. has been that the one-point
probability density function (p.d.f.) of velocity in isotropic turbulence is not too far from
being a Gaussian. Intriguing are the similar findings (that Λ ≈ λ ) even when the one-
point p.d.f.s are strongly non-Gaussian, such as for the streamwise velocity fluctuation
u in a two-dimensional channel (Laufer 1950) and for wall shear-stress fluctuations in
a pipe (Wetzel & Killen 1972). On this basis, it could be concluded that Rice’s result
has a much more general applicability than it has been proved for. Measurements of
Wygnanski & Fiedler (1969), on the other hand found that the ratio Λ/λ for u’ as well
as the v’ varied generally between 1 and 1.5 along the centreline of an axisymmetric jet,
increasing to a value as high as about 2 for v towards the edges of the jet.
Badri Narayanan, Rajagopalan & Narasimha (1974, 1977) measurements of u in nearly
isotropic grid turbulence and in a turbulent boundary layer, the zero-crossing frequency
was counted visually from a trace of the signal u. These results suggested that, while
Λ/λ in grid turbulence nears 1, the ratio in the turbulent boundary layer varied between
about 3 and 5. This second result was somewhat off-mark from expected lines on the
basis of above result, was attributed to the stronger departures from Gaussianity of
turbulent fluctuations in the boundary layer on account of it being highly and Non-
isotropic. While this last result tends to suggest that Λ is not a direct measure of
dissipation in strongly non-Gaussian situations, further results of Badri Narayanan et al.
emphasized the role of the zero-crossing scale in another important dynamical context.
Their measurements suggested that the zero-crossing frequency N0 of a given signal was
nearly equal to the characteristic pulse frequency Np of that signal. Pulses are simply
patches of activity in a turbulent signal that has been passed through a narrow-band
filter set at a sufficiently high mid-band frequency. Their characteristic frequency Np was
measured in these experiments with the technique of Rao, Narasimha & Badri Narayanan
(1971). These high frequency pulses have a direct bearing on the fine structure and the
internal intermittency of turbulence. Two other properties of importance are attributed
to the pulse frequency. First Rao et al. implied a connection between these pulses and the
turbulent bursts observed by Kline et al. (1967) in the wall region of a turbulent boundary
layer, suggesting that these pulses in a boundary layer are a part of the signature of a
burst. Secondly Brown & Thomas (1977) showed that the pulses are coupled to the large-
scale motion. There are conflicting conclusions about the relation between the pulses,
bursts and non Gaussian turbulent signals for flows like wall boundary layer, Heated Jet
Guidelines for authors 7
and Wakes the ratio has been found to be closer to 1 as well, implying validity of Rice’s
result to the non Gaussian signals as well. In the current study involving Planejet, ratio
Λ/λ has been found to be closer to 1 at the central region of flow which remains almost
flat up to the region of inflexion points, then falling in the outer regions of flow. The
behaviour is found to be similar for all three perturbation quantities. These observations
agree very well with Liepmann’s interpretation of Rice’s result for Gaussian signal. The
deviation of turbulent signal from Gaussianity doesn’t seem to have much effect on Zero
crossing results.
While this last result tends to suggest that Λ is not a direct measure of dissipation
in strongly non-Gaussian situations, further results of Badri Narayanan et al. empha-
sized the role of the zero-crossing scale in another important dynamical context. Their
measurements suggested that the zero-crossing frequency N0 of a given signal was nearly
equal to the characteristic pulse frequency Np of that signal. Pulses are simply patches
of activity in a turbulent signal that has been passed through a narrow-band filter set a t
a sufficiently high mid-band frequency. Their characteristic frequency Np was measured
in these experiments with the technique of Rao, Narasimha & Badri Narayanan (1971).
These high-frequency pulses have a direct bearing on the fine-structure and the internal
intermittency of turbulence. Two other properties of importance are attributed to the
pulse frequency. First Rao et al. implied a connection between these pulses and the
turbulent bursts observed by Kline et al. (1967) in the wall region of a turbulent boundary
layer, suggesting that these pulses in a boundary layer are a part of the signature of a
burst. Brown & Thomas (1977) showed that the pulses are coupled to the large-scale
motion. There are conflicting conclusions about the relation between the pulses, bursts
and the large structure, but the connection between the high-frequency pulses and the
fine structure intermittency is almost tautological. In practice, the determination of the
characteristic pulse frequency Np is quite involved (e.g. Rao et al. 1971 ; Kuo & Corrsin
1971), and this in itself is therefore a sufficient reason for taking seriously the conclusion
of Badri Narayanan et al. that Np ≈ N0 in general ; their contention was that Np ≈ N0
is indeed the general result for turbulence which, in the relatively simple and degenerate
case of isotropic turbulence, simplifies to N0 = (2πΛ)−1 ≈ (2πλ)−1 .
Antonia, Danh & Prabhu (1976) made zero-crossing measurements. in a variety of
shear flows including the atmospheric surface layer, and the temperature fluctuation θ
in a slightly heated turbulent boundary layer. They found that the ratio Np ≈ N0
for all signals (except in the atmospheric surface layer). However, in contrast with the
conclusions of Badri Narayanan et at., Antonia et al. found that Λ/λ ≈ 1 for u even
in a turbulent boundary layer.In addition to this major difference, the data of Badri
Narayanan et al. and Antonia et al., both obtained at comparable Reynolds numbers,
differ also in the value of the non-dimensional pulse frequency by a factor of 3 or 4.

3.2. Result: Zero crossing and Production scale scale


Zero crossing time of production signal −u′v ′ ∗ (∂Ui /∂y) is calculated. High frequency
component of production signal is filtered out
that the differentiated signal can strengthen the contrast between quiescent and active
periods. However, as figures 1 (a, b) show, it is the use of filters that brings out most
clearly the existence of intermittent periods of considerable activity, which we may,
tentatively and for convenience, call ‘bursts’. we have used a narrow band pass filter.
It is clear that if the spectrum of the turbulent signal is falling rapidly at the cut-off
frequency (as it must apparently be for ‘bursts’ to become noticeable), there will not be
much difference between the output signals from the high-pass filter and from a wave
analyzer with pass band near the cut-off frequency.
8 A. N. Other, H.-C. Smith and J. Q. Public

(a) X/H = 8.33 (b) X/H = 10.0 (c) X/H = 11.66

(d) X/H = 13.33 (e) X/H = 15.0 (f) X/H = 16.66

(g) X/H = 18.33 (h) X/H = 20.00 (i) X/H = 21.66

Figure 7: Zero crossing scale ratio to Taylor scale (Λ/λ), plotted against y/H:
• = ū′ ; · ·, ♦ = v̄ ′ , ▶ = w̄′ .

Figure 8: PDF of the interval tz between successive zero-crossings in turbulent region.


Guidelines for authors 9
1.2

p(tz − t̄z )/t′z


0.8

0.6

0.4

0.2

0
-1 0 1 2 3 4
(tz − t̄z )/t′z

Figure 9: Effect of filtering on number of production peaks.

Figure 10: Unfiltered,Differentiated and filtered u data.

Figure 11: Production series and filtered u data.

Even from these filtered signals, however, it is clear that the duration or frequency of
bursts cannot always be estimated with accuracy, and that further careful processing, of
the kind found necessary in intermittency measurements (e.g. Piedler & Head 1966), may
be required before completely reliable values can be obtained. However, certain general
trends can be established beyond reasonable doubt, as we shall see below.
10 A. N. Other, H.-C. Smith and J. Q. Public

(a) X/H = 8.33 (b) X/H = 10.0 (c) X/H = 11.66

(d) X/H = 13.33 (e) X/H = 15.00 (f) X/H = 16.66

(g) X/H = 18.33 (h) X/H = 20.00 (i) X/H = 21.66

Figure 12: Production scale to Zero scale ratio (Zp /Λ), plotted against y/H:
• = Zp /λ; · ·, ♦ = P rod./Diss.

Although there is some difficulty in clearly identifying ’bursts’ in the data traces, a
reasonable procedure is based on using a narrow band pass filter that leads to quite
reliable values of the time interval between bursts. The dependence of the mean burst
rate so found on the pass-band centre frequency is interesting: at low frequencies it
is approximately linear, as it is in white noise; at higher frequencies, it settles down
to a nearly constant value. This is consistent with the view that the smaller eddies
have a characteristic structure of their own (Batchelor 1953, ch. 8). As it is known that
convection velocities are of the order of Ū , the fact that T̄ is of order δ/U over a wide
frequency range suggests that the ‘ bursts correspond to regions of concentrated vorticity,
which are rich in spectral content and are separated on an average by distances L ∼
UT̄ of the order of several boundary layer thicknesses. Preliminary observations made in
turbulence behind a grid and in wakes show similar results, that the bursts are related to
the well-known phenomenon of intermittency in the small-scale structure of turbulence
(observed by Batchelor & Townsend (1949).
From the scaling of T̄ and the distribution of T, it therefore appears that the bursts
observed in a turbulent boundary layer are related to the small scale intermittency of
Batchelor & Townsend, and the spottiness of Kolmogorov & Landau. In other words,
such bursts may well be a fairly general feature of all turbulent flows.
Guidelines for authors 11

(a) Bef ore T1 (b) At T1 (c) Af ter T1

(d) Bef ore T2 (e) At T2 (f) Af ter T2

(g) Bef ore T3 (h) At T3 (i) Af ter T3

′2 ′2 ′2
Figure 13: (u +v2 +w )
, Before,At and After Production peaks at 3 different time instants
at X = 10.00:
12 A. N. Other, H.-C. Smith and J. Q. Public

(a) T1 (b) T2

(c) T3 (d) T4

Figure 14: Production time series with instantaneous U velocity profiles Before, At and
After the peak, at X = 10.00:
Guidelines for authors 13

(a) (b)

(c) (d)

Figure 15: Production time series with instantaneous U velocity profiles Before, At and
After the peak, at X = 13.33:
14 A. N. Other, H.-C. Smith and J. Q. Public

(a)

(b)

′2 ′2 ′2
Figure 16: (u +v2 +w )
, Before,At and After Production peaks at 3 different time instants
at X = 15.00:
Guidelines for authors 15

(a)

(b)

′2 ′2 ′2
Figure 17: (u +v2 +w )
, Before,At and After Production peaks at 3 different time instants
at X = 16.66:

You might also like