Documente Academic
Documente Profesional
Documente Cultură
l, February 1994
ISSN 0567-7718
SKEWNESS
FACTOR
VELOCITY
OF TURBULENT
DERIVATIVE*
theory of turbulence
I. I N T R O D U C T I O N
The skewness factor of turbulent velocity derivative is
( 0ul 3
0Ul 2 3/2
(1)
here Ul is the velocity component along the Xl direction, < . . . ) denotes the statistical average. Although the probability distribution function (PDF) of the turbulent velocity is
Gaussian, the P D F of the turbulent velocity derivative is not Gaussian, and the skewness
S is different from zero. Many efforts have been made to calculate S, but it is still a controversial problem what is the asymptotic behavior of the skewness of isotropic turbulence
when the Reynolds number R~ is very high. According to the LN (log-normal) model based
upon Kolmogorov's 1962 theory [1,21
ISl ~
> o
(2)
Here a is a positive constant, (2) implies that [S[ --+ oc as R~ -~ oc. Tennekes [3] proposed a
vortex t u b e model of small-scale structure and predicted t h a t S approaches a finite constant
when R~ is high. The fractal model of small-scale structure [4] gives a similar result as (2)
and also predicts that IS I --* oo as Rx --+ oc. Kraichnan used the mapping closure method
to study the Burgers turbulence, and then proposed a model equation of P D F of the velocity
derivative for the Navier-Stokes turbulence, by which it is concluded [5] t h a t S approaches a
finite constant when R~ is high. Recent DNS (direct numerical simulation) results [6] indicate
that S is about - 0 . 5 when R~ is high. The above-mentioned LN model, Tennekes' vortex
Received 17 November, 1992, revised 12 August, 1993
* The project supported by the National Basic Research Program "Non-linear Science"
Vol.lO, No.l
13
tube model, fractal model, and Kraichnan's model equation of P D F did not given numerical
value of S, which can be compared with the DNS results.
In this paper, we will study the asymptotic behavior of the skewness S from a totally
different point of view: it is unnecessary to use a particular model or make a special physical
assumption about the small-scale structure of turbulence. Similar to the other statistical
properties, the skewness S could be calculated from the basic equation of motion by the
method of statistical mechanics. In fact, by applying the nonequilibrium statistical mechanics closure method and some basic formulae of spectrum dynamics, we cannot only predict
whether ISI approaches a finite constant as R~ -~ oo, but also calculate the numerical value
of S and to make a comparison between the present theoretical result and the DNS results.
II. U P P E R
LIMIT OF ABSOLUTE
VALUE OF SKEWNESS
Suppose r is a random variable, then {((r3)r - (r2}r2) 2) _> O, which implies that
Oul
ISI ~ F 1/2
((0~11)4)/((~)2) 2
(3)
Here F is the flatness factor of the turbulent velocity derivative. From the Navier-Stokes
equation, by the nonequilibrium statistical mechanics closure method, it has been shown [71
that F approaches a constant (about 15) as the Reynolds number R~ -+ c~. Hence, according
to (3), ISI has a upper limit V~5 ~ 3.9 and cannot approach infinity as R~ --+ oo.
Actually we can make a better estimate of the upper limit of ISI than (3) which is
valid for any random variable and has not taken advantage of the characteristics of isotropic
Oul
ISl
(4)
here the asymptotic value'15 of the flatness F is used. The formulae (3) and (4) give only
an estimate of the upper limit of ISI. In the next section, we will apply the nonequilibrium
statistical mechanics closure method to calculate the value of S and study how S changes
with the Reynolds number.
III. R E L A T I O N S H I P
BETWEEN
SKEWNESS
AND REYNOLDS
NUMBER
In order to use (1) to calculate the skewness S, it is necessary to calculate the moments
{(0Ul
]2~0Xl,/ and {(~11)3}. For isotropic turbulenceB]
O~Xl) ) = - ( 2 / 3 5 ) L
E ( k ) k 2 d k = e/(15v)
T(k)k2dk
(5)
(6)
14
1994
Here E(k) is the energy spectrum, e is the energy dissipation rate, v is the kinematic
viscosity, and T(k) is the energy transfer spectrum function. For a stationary turbulence,
T(k) = 2vk2E(k), by using (1) (5) and (6), we obtain
-(12v~-g/7)e-2/ak21~
E(k)k4dk
(7)
u 2 > 1 / 2 ,,~//,,
here
= [<
(8)
(9)
<
> = (2/3)
E(k)dk
(10)
From (5) (8) (9) and (10), after some manipulation, we obtain
E(k)dk
Ra = (2/3)v~(ka/e) 2/3
(11)
By using the nonequilibrium statistical mechanics closure method [9], we have derived
t h e following approximate formula for the energy spectrum in the high-wavenumber range
E(k) = e2/3k-5/3F(k/kd)
(12a)
(125)
Let k0 be the characteristic wavenumber of the energy range, and a simple approximate
formula of E(k), which is valid in both high-wavenumber and low-wavenumber ranges, is [7'10]
0.5
0.4
i0
l0 2
R~
Fig.1 Skewness S vs. Reynolds number Ra
10 3
(13)
Vol.10, No.l
15
IV. DISCUSSION
Fig.1 shows that the skewness S approaches a constant S ~ = -01515 as R~ ~ co.
This theoretical result agrees with (4), and is compatible with the DNS result S = - 0 . 5 of
Vincent and Meneguzzi [6]. Since the R~ of current DNS is of the order 102, it is doubtful
whether the skewness S obtained by the DNS can represent the asymptotic value S ~ . Fig.1
shows that S almost attains the value S ~ when R~ ~ 102~ and in fact S differs from S ~
by less than 1% when R~ > 60. Therefore, the theoretical result of this p a p e r confirms the
following conjecture made by Vincent and Meneguzzi[61: the skewness S obtained by the
DNS with R~ ~ 102 is nearly the same as the asymptotic value S ~ .
The issue, whether t h e skewness S and the flatness F approach a finite constant as
R~ --~ c~, is essentially related to the problem of universality of small-scale structure for
high-R~ turbulent flows. If the small-scale structure is universal, the statistical properties
of small-scale motion such as the skewness S and the flatness F will be independent of
the large-scale motion and the Reynolds number R~ so long as R~ is high enough. By the
L N model and the fractal model, IS I and F approach infinity as R~ --* c~, so they are
dependent of R~ no m a t t e r how high R~ is. According to the theoretical results of this
paper and Ref.[7], S and F become universal constants independent of R~ when R~ is high
enough, probably implying the universality of small-scale structure of high-R~ turbulent
flOWS.
As mentioned above, when k < < kd (13) becomes the low-wavenumber energy spect r u m used in Ref.[10], and E ( k ) ,,~ k as k/ko ~ O. Of course, it is reasonable to adopt other
forms of the low-wavenumber energy spectrum, for example, E ( k ) ~ k 4 as k/ko --~ O. We
have used different forms of low-wavenumber energy spectrum to calculate S, and it is found
that the forms of low-wavenumber energy spectrum only affect the skewness S for small R~.
The different forms of low-wavenumber energy spectrum give the same S when R~ is high,
in particular the same asymptotic value S ~ -- -0.515.
REFERENCES
[1] Monin AS, Yaglom AM. Statistical Fluid Mechanics of Turbulence. London, MIT, 1975
[2] Van Atta CW, Antonia RA. Reynolds number dependence of skewness and flatness factors of
turbulent velocity derivatives. Phys Fluids, 1980, 23(2): 252
[3] Tennekes H. Simple approximations to turbulent energy transfer in the universal equilibrium
range: Phys Fluids, 1968, 11(1): 669
[4] Frisch U, Sulem PL, Nelkin M. A simple dynamical model of intermittent fully developed turbulence. J Fluid Mech, 1978, 87(part 4): 719
[5] Kraichnan RH. Models of intermittency in hydrodynamic turbulence. Phys Rev Left, 1990,
65(5): 575
[6] Vincent A, Meneguzzi M. The spatial structure and statistical properties of homogeneous turbulence. J Fluid Mech, 1991, 225:1
[7] Qian J. A closure theory of intermittency of turbulence. Phys Fluids, 1986,. 29(7): 2165
[8] Betchov R. An inequality concerning the production of vorticity in isotropic turbulence. J Fluid
Meeh, !956, 1(5): 497
[9] Qian J. Universal equilibrium range of turbulence. Phys. Fluids, 1984, 27(9): 2229
[10] Leslie DC, Quarini GL. The application of turbulence theory to the formulation of subgrid
modelling procedures. J Fluid Mech, 1979, 91 (part 1): 65