Applied Mathematics and Mechanics

(English Edil~ion, Vol. 16, No. 3, Mar. 1995)

Published by SU,
Shanghai, China

ON THE C L O S U R E P R O B L E M OF T U R B U L E N C E M O D E L THEORY*
Tsai Shu-tang ( ~ . ~ )

Liu Yu-lu (~]~[~)

(Shanghai University; Shanghai Inst. of Appl. Math. and Mech.,Shanghai 200072)

(Received March 9, 1994)

Abstract
It is a wrong viewpoint that the turbulence, closure problem is due to the
non-lhwarity of N-S equation, because if we omit the non-linear terms in N - S
equation, many physical quantities can not be obtained other than the mean-values. In

Abstract
this paper, we proof that the closure
problem of turbulence be induced by lack of
statistical distribution in present turbulence theory. And the restriction of turbulence

The model
one-dimensional
of theof motion
of a rigid simulation
flying plateofunder
attack has
theory andproblem
shortcoming
direct numerical
N-S explosive
to solve the
an analytic solution only when the polytropic index of detonation products equals to three. In
tui'bulence have been pointed out.
general, a numerical analysis is required. In this paper, however, by utilizing the "weak" shock
behavior of theKey
reflection
in the closure
explosive
products,
and
applying
the small
parameter purwords shock
turbulence,
problem,
model
theory,
distribution
function
terbation method, an analytic, first-order approximate solution is obtained for the problem of flying
I. I ndriven
t r o d uby
c t various
ion
plate
high explosives with polytropic indices other than but nearly equal to three.
FinalAvelocities
of
flying
plate closure
obtained
agree very
well with
numerical
results by computers.
Thus
point, the turbulence
problem
stemmed
from
the non-linearity
of N-S, is wrong.
an
analytic
formula
with
two
parameters
of
high
explosive
(i.e.
detonation
velocity
and
polytropic
It is true that if the N-S were linear, the Reynolds stress would disappear in Reynolds
index) for estimation of the velocity of flying plate is established.
equation, so there exists no turbulence closure problem. It is obvious that from the Linear N-S
we can only get the averaged quantities such as U~ and P. But in engineering some valuesl
1. three-dimensional
Introduction
such a~s the turbulence intensity, two or
spectrum of turbulence etc. are very
important so that the theory of turbulence without nonlinear term is not complete and
Explosive driven flying-plate technique ffmds its important use in the study of behavior of
reasonable.
materials under intense impulsive loading, shock synthesis of diamonds, and explosive welding and
A complete
theory
must consist
of velocity
three necessary
parts,
that is,it are questions
cladding
of metals.turbulence
The method
of estimation
of flyor
and the way
of raising
1.
The
Reynolds
equation
(if
the
fluid
is
incompressible):
of common interest.
Under the assumptions of one-dimensional plane detonation and rigid flying plate, the normal
OU, " U
OU,
OP is t-vVW,
approach of solving the problem
of motion
of 1flyor
to solve the OU~U~
following system of equations
--0i - - I - ' - ~ - s =
pox,
~
dxj
governing the flow field of detonation products behind the flyor (Fig. I):
dU~
--0

2. Flunctuation equation:

--ff
ap +u_~_xp+
au

au

au

y1

{ Ou~ _ . , auj _ aS
au~ a s au~ =
---~--t-uj~-~j ~-uj-~-~j
a--T +uj axj
=o,
Ou#

Ox, = 0

=p(p, s),

(1.1)

=o,

=0,

i ap b vv~u, + ou~uj

pax,

(i.0

axj

(1.~.)

where3~p,The
p, S,
u are pressure,
density,
specific
entropy and particle velocity of detonation products
statistical
distribution
law
of turbulence.
respectively,
with
the
trajectory
R
of
reflected
shock of turbulence
detonation wave
D as
a boundary
The (3) was payed little attention in previous
studies
and
here we and
raisethethe
trajectory F of flyor as another boundary. Both are unknown; the position of R and the state parameters on it are governed by the flow field I of central rarefaction wave behind the detonation wave
*Project
supported
the National
Natural
Science
Foundation
China
D and
by initial
stage ofbymotion
of flyor
also; the
position
of F andofthe
state parameters of products
225
293

226

Tsai Shu-tang and Liu Yu-lu

problem again. On the other hand, we discuss the restriction of turbulence mooel theory and
shortcoming o f direct numerical simulation of N-S in this paper.
II.

The Closure Problem of Turbulence

In 1895, the Reynolds equation was obtained by 0. Reynolds according to the theory of
molecular motion. But he overlooked the distinctions between the turbulence vortex motion
and the molecular motion. In fact, there exist many differences and the most important ones
are:
1. The delay time: the collision time between the molecules is I0 -it s but the delay time
for turbulence vortex is order of 10'~103s. Hence the properties of memory can be omitted in
dealing with the molecular motion and considered in turbulence:
2. The statistical distribution law: in gas dynamics, there is a famous statistical law of
Maxwell, which is controlled_by Boltzman differential-integral equation. With the distribution,
the various moments can be got by the integration,
Abstract so that there is no closure problem. In the
turbulence, neither the statistical law has been found, nor its controlling-equation. On the other
problem of the
motion
of anormal
rigid flying
plate underand
explosive
attack has of
hand, The
theone-dimensional
turbulence distribution
is far
from
distribution
the skewness
an
analytic
solution
only
when
the
polytropic
index
of
detonation
products
equals
to
homogeneous turbulence is not equal to zero from analysing the experiment data three.
which Incan
general, a numerical analysis is required. In this paper, however, by utilizing the "weak" shock
not be unified for different situations, so that the relationships among the turbulence quantities
behavior of the reflection shock in the explosive products, and applying the small parameter purcan not be found, from which the turbulence closure problem is deduced.
terbation method, an analytic, first-order approximate solution is obtained for the problem of flying
the by
other
handhigh
the explosives
averaged with
free polytropic
path and velocity
o f thermal
motion
gas todynamics
plateOn
driven
various
indices other
than but
nearly inequal
three.
are
functions
of pressure
and temperature
only,with
butnumerical
the turbulence
length
and
Finalthevelocities
of flying
plate obtained
agree very well
results bymixing
computers.
Thus
an analytic formula
two parameters
of high
explosive
(i.e. detonation
velocity and
flUnctuation
velocitywith
of turbulence
are the
functions
of pressure,
temperature,
timepolytropic
and space,
index) only
for estimation
the velocity
of flyingprocess.
plate is established.
which
cause theofdifficulty
in solving

III.

The R e l a t i o n s B e t w e e n t h e S o l u t i o n and S t a t i s t i c a l L a w
1. Introduction
Discussing the problem in three steps:

1.Explosive
With the driven
Reynolds
equationtechnique
only, theffmds
turbulence
can not
directly
becauseofthe
flying-plate
its important
usebeinsolved
the study
of behavior
materials (1,1)
underisintense
impulsive
loading,
shockstress;
synthesis of diamonds, and explosive welding and
equation
not.closed
for the
Reynolds
cladding
of metals.
Theand
method
estimation
flyor velocity
way too.
of raising
are questions
2. With
the (1.1)
(1.2)oftogether,
theofequations
are and
not the
closed
If theit (1.1)
added to
of
common
interest.
the fl.2), we obtain a soluble equation (N-S), whoes solution is momentary values of U~ and
Under the assumptions of one-dimensional plane detonation and rigid flying plate, the normal
P. The mean-values of U~ and P, which are very useful in engineering, must be seperated from
approach of solving the problem of motion of flyor is to solve the following system of equations
Ut
and P with
the field
statistical
law;
governing
the flow
of detonation
products behind the flyor (Fig. I):
3. The complete turbulence theory includes three important parts: the Reynolds equation
(1.1), the flunctuation function(1.2) and the statistical distribution law. With the statistical law
--ff
=o,
ap +u_~_xp
the various kinds o f moments in turbulence
can+ be au
obtained according to the method proposed
by O. Reynolds (1895). According au
to the au
different
law, the averaging method is
y1 statistical
=0,
different; then the mean-value of velocity, pressure and their moments are different
too.
(i.0
Hence, with unaccurated statistical aSlaw the
a s relations among the averaged quantities and
a--T
=o,
moments can not be found, from which the turbulence closure problem is deduced.

IV.

=p(p, s),

R e s t r i c t i o n s o f T u r b u l e n c e Model T h e o r y

where
p, S, u arethepressure,
density,
specific
entropy
and on
particle
velocity
detonation
In p,
Substance,
turbulence
model
theory
is based
the (1.1)
andof(I.2)
adding products
the same
respectively,
with
the
trajectory
R
of
reflected
shock
of
detonation
wave
D
as
a
boundary
and the
relations or equations to the Reynolds stress. It is very difficult to solve the (1.2) directly so
trajectory F of flyor as another boundary. Both are unknown; the position of R and the state parausually we use the equations o f u~'s moments, for example,
meters on it are governed by the flow field I of central rarefaction wave behind the detonation wave
D and by initial stage
of motion of flyor also; the position
a
OUt ~ofUFk and the state
O parameters of products
293

On the Closure Problem of Turbulence Model Theory

1 (. a p uk
= ----~,-~

227

--

+ 0-~'u, ) + vV 2 U~Uk

Ou ~

Out

Ox~

Ox~

(4.1)
The physical information involving in (1.4) which is not closed too is much more than
that in (1.2). In order to close the (1.1), (I.2) or (4.1), some special relations or equations must.
be added, which are developed into the turbulence model theory. As we knew, the hypothese
or assumptions introduced in adding those relations or equations have not the solid basement
of physics and experimental data, so that the restrictions for the turbulence model theory are
very obvious. On the other hand, the mean-velocity and pressure are not obtained directly
Abstract
from (I.I) with (1.2), sothe error is very large.
V. TThe
h e one-dimensional
P o s s i b i l i t y o f problem
D i r e c t of
N uthe
m emotion
r i c a l Sofi ma urigid
l a t i oflying
n o fplate
N-S under
f o r Turbulence
explosive attack has
an analytic solution only when the polytropic index of detonation products equals to three. In
As shown in Sec. II, even if we obtained the momentary value of velocity and pressure,
general, a numerical analysis is required. In this paper, however, by utilizing the "weak" shock
we would not separate the mean-value of/7~ and P from Ur and _19without the statistical law.
behavior
of the reflection shock in the explosive products, and applying the small parameter purIn fact the
quantities
needed first-order
in engineering
can besolution
obtained
from theforexperimental
terbation
method,
an analytic,
approximate
is obtained
the problem ofdata,
flyingbut
we can
not by
dovarious
every high
experiment
each
engineering
problem,
so but
the nearly
numerical
method
plate
driven
explosivesforwith
polytropic
indices
other than
equal to
three. is
Final
velocities
flying plate
obtained
agree very
numerical toresults
Thus
commonly
usedof instead
of the
experiment.
Butwell
it iswith
impossible
use by
the computers.
direct numerical
an
analytic
formula
with
two
parameters
of
high
explosive
(i.e.
detonation
velocity
and
polytropic
simulation of N-S equation to solve the turbulence problem (DNS). Firstly, by the DNS
index)
estimation
of the
velocity ofvalue
flying be
plate
is established.
methodforonly
can the
momentary
got.
Secondly, the boundary conditions must be
supplied in the D N M , which satisfy some statistical law. Hence the limited numerical
1. Introduction
experiments can not express the practical
turbulence. A very interesting example can show the
point: one flow picture among the three hundred numerical results of flow field can be found
Explosive driven flying-plate technique ffmds its important use in the study of behavior of
to be similar to one flow photo picture in two hundred practical flow fields, from which we
materials under intense impulsive loading, shock synthesis of diamonds, and explosive welding and
conclude that the DNS has only the reliability of 1/60000.
cladding
of metals. The method of estimation of flyor velocity and the way of raising it are questions
A current
method used in solving the turbulence is large vortex simulation (LVS), which is
of common
interest.
stemmed
from
D N M , soofthere
exists the same
Under
the the
assumptions
one-dimensional
planedifficulty.
detonation and rigid flying plate, the normal
approach of solving the problem of motion of flyor is to solve the following system of equations
VI. Conclusion
governing the flow field of detonation products behind the flyor (Fig. I):
From Sec. I - V , we can conclude that the turbulence closure problem comes from the
absence of accurate statistical distribution
law. The
--ff
=o,non-linearity of N-S only adds the
ap turbulence
+u_~_xp+ problem.
au
difficulty in the process of solving the

References

au

au

y1

=0,
(i.0

Tsai Shu-tang, et ah, Prospect

theory in previous half decade, Advances in
aS of turbulence
as
a--T
=o,
Mech., 10, 1 (1980), 16--36.
p =p(p,
[ 2 ] Tsai Shu-iang, Lin Duo-min,
A news),expression for turbulence field, Appl. Math. and
Mech., 12, 1 (1991), 101-- 105.
where
p, p, S, u are pressure, density, specific entropy and particle velocity of detonation products
[ 3 ] Launder, B. E. et al., Mathematical Models of Turbulence, Academic Press, London
respectively, with the trajectory R of reflected shock of detonation wave D as a boundary and the
and New York (1970).
trajectory F of flyor as another boundary. Both are unknown; the position of R and the state para[meters
4 ] Launder,
B. E.et al.,
Turbulence
and Their
Applications,
Editions
Enrolles wave
(1984).
on it are governed
by the
flow fieldModels
I of central
rarefaction
wave behind
the detonation
[ 5and
] by
Lestie,
C., of
Developments
in the
of Turbulence,
Oxford
Universityof Press,
New
D
initialD.stage
motion of flyor
also;Theoo'
the position
of F and the
state parameters
products
York (1983).
293
[ 1]