Sei sulla pagina 1di 12

European Journal of Mechanics / B Fluids 80 (2020) 60–71

Contents lists available at ScienceDirect

European Journal of Mechanics / B Fluids


journal homepage: www.elsevier.com/locate/ejmflu

A revisit of Navier–Stokes equation


Wanan Sheng
SW MARE Marine Technology, Ireland

article info a b s t r a c t

Article history: An effort has been recently paid to derive and to better understand the Navier–Stokes (N–S) equation,
Received 15 April 2019 and it is found that, although the N–S equation has been proven to be correct by numerous examples,
Received in revised form 4 December 2019 some concepts and principles behind the equation may not be correct or consistent. For instance, from
Accepted 5 December 2019
an analysis of the simple classic Couette flow, the requirement of the symmetric stress tensor is in
Available online 10 December 2019
fact conflicting with the solution of the Couette flow.
To solve the inconsistencies identified in this research, a reformulation of the total tensor is
suggested for accommodating the fluid friction which bears a solid physics, and the new total tensor
could resolve all the inconsistencies and conflicts identified. The newly defined fluid friction tensor
is then used to derive N–S equation, and as expected, the same N–S equation as the original form of
N–S equation for incompressible flows is obtained. For compressible flows, to achieve the same N–S
equation as the original N–S equation, a slightly different assumption but yet in a very similar manner
as Stokes made in 1845 is needed.
It is the author’s intention that the N–S equation under the new defined total tensor has different,
but yet more physical background concepts and principles. It is hoped that the revisit of the N–S
equation could shed some light to better understand the dynamic flows and lead to establish new and
better approaches to solve the complicated flow problems in future.
© 2019 Elsevier Masson SAS. All rights reserved.

1. Introduction pipe flow and the Poiseuille’s capillary-tube flow (see [4]). Now
it is generally accepted that the establishment of the fluid dy-
Fluid dynamics is an ancient topic and people have been namics equation was finished with the work of Stokes in 1845,
trying to solve the fluid mechanics/dynamics problems since the and the fluid dynamic equation was later named as the Navier–
great Greek philosophers and scientists Aristotle (384–322 BC) Stokes equation, even though Navier and Stokes published their
and Archimedes (287–212 BC) [1]. The first partial differential equations independently in a gap of more than 20 years.
equation for fluid dynamics was much later formulated by Eu- The Navier–Stokes (N–S) equation is the fundamental equation
ler in 1752 when he considered only for inviscid fluids (that for governing fluid motion and dynamics, and so far numerous
is why the fluid dynamic equation for inviscid fluids is called examples have proven the correctness of the N–S equation for
Euler equation). After that, by adopting the Newton’s definition fluid dynamics. However, it has been well recognized that seeking
of friction due to the velocity gradient and fluid viscosity, Navier an analytical solution to the N–S equation has been proven too
and Stokes could independently include the viscous forces into difficult and analytical solutions can only be obtained for some
the equation, thus now it is called Navier–Stokes equation. Ac- simple laminar flows, therefore, turbulence is frequently referred
cording to Anderson [2], an interesting fact is that Navier did as the major unsolved problem of classical physics [5]. In 2000 the
get the correct equation in 1822, but his derivation has been Navier–Stokes equation was selected to be one of seven Millen-
nium Problems by the Clay Mathematics Institute of Cambridge,
greatly flawed. Stokes re-derived the fluid dynamic equation in
U.S. (http://www.claymath.org/millennium-problems), and a spe-
a more delicate manner in 1845 [3]. Darrigol [4] has indicated
cial award of $1 million is provided for the answer to each of
that in the period between Navier (in 1822) to Stokes (in 1845),
the 7 millennium questions. In 2008 the U.S. Defense Advanced
other three main scholars: Cauchy, Poisson and Saint-Venant, had
Research Projects Agency (DARPA) listed it as one of 23 DARPA
made their contributions to establish the fluid dynamic equation
Mathematical Challenges — ‘‘Mathematical Challenge Four: 21st
under different assumptions. Without particular reasons, the final
Century Fluids’’. The challenge statement is as following: al-
settlement of the fluid dynamics equation might be the successful
though classic fluid dynamics and the Navier–Stokes Equation
applications of Stokes’ pendulum theory in the Hagen’s Bessalian were extraordinarily successful in many practical problems, in-
cluding the understanding of shock waves and turbulent flows,
E-mail address: wanan_sheng@outlook.com. new methods (and understandings) are still needed to tackle

https://doi.org/10.1016/j.euromechflu.2019.12.005
0997-7546/© 2019 Elsevier Masson SAS. All rights reserved.
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 61

the complex fluids such as foams, suspensions, gels, and liq- stress tensor for elastic materials). This Stokes’ symmetric stress
uid crystals (https://www.britannica.com/science/Navier-Stokes- tensor was so well accepted in the community that it has been
equation). taken as a law that all fluids must follow, in a similar manner
Due to the difficulties in obtaining analytical solutions to N–S as the Newton’s laws of motion as the universal law of motion.
equation, especially for those real complicated turbulent flows, However, as stated in [31], the symmetry in the stress tensor is
modern numerical methods, especially the Computation Fluid violated in an electric field on polarized fluid molecules, in which
Dynamics (CFD), have been developed and advanced with the in- antisymmetric stresses must be included in such circumstances.
creased computer power and the advanced numerical algorithms, A similar contradiction has been reported in [32]. The author is
and nowadays the computational fluid dynamics (CFD) have been wondered why the Stokes’ symmetric stress tensor is violated in
widely used to numerically solve the N–S equation, including some special flows?
some very complicated fluid dynamics problems, such as air- To find an answer to the question above, the author has
planes, air engines, ships, fire modelling, heat transfer, chemical recently made an effort to derive and better understand the
reaction and so on. To date, CFD have been very successful for N–S equation. It is found that, although the N–S equation is
studying various fluid dynamic problems, and great achievements proven to be correct for governing all various flows, including
have been attained. Good examples are the CFD applications in the compressible flows with the Stokes’ assumption [31] (page
development of Airbus 380 in Fig. 1 [6] and of Boeing 787 in 128), some concepts and principles behind the equation may not
Fig. 2 [7], from which we can already see how much computer be correct or consistent. For instance, an analysis to the classic
modelling work can be already carried out using CFD. With the Couette flow has shown that the requirement of the symmetric
increase of computer powers, the advancement of numerical stress tensor is conflicting with the solution of the Couette flow
algorithms and the understanding of the complicated flows (for when a real physics is applied in the analyses (more details on
instance, the better turbulence models), CFD will get more and these conflicts/inconsistencies can be found in Section 4).
more applications and become more and more capable in solving To understand the problems mentioned above, the forces act-
the problems in our daily and in the complicated fluid dynamics. ing on the fluids are re-examined which could show the incon-
Though CFD have been very successful in solving N–S equa- sistencies for the Stokes’ symmetric stress tensor and a refor-
tion, there are still many challenges in using CFD for complicated mulation has been made to the total tensor for accommodating
flows. This includes the examples mentioned above, where CFD the total surface force of more physical significance. Using the
is still limited to some specific problems and the explorations newly defined friction tensor, an N–S equation exactly same as
are still on going to use more CFD in the design of the whole the original form of N–S equation is obtained for incompress-
aircrafts and the off-design situations. In addition, the needs to ible flows. It is also possible to obtain a same N–S equation
use CFD to solve the multi-physical problems [6–10] and seek for compressible flows using a slightly different assumption as
global optima [7] are the current challenges. Stokes made. Therefore, under the newly defined total tensor,
The current approaches in CFD include the most computa- the N–S equation is maintained same, however, the concepts and
tional demanding approach: direct numerical simulation principles behind the equation are different but of more physical
(DNS, [11,12]); relatively less (compared to DNS) but still very significance.
computational demanding method: large-eddy simulation To present the research work, following arrangement is made
(LES, [13–17]) and the most practical method: the Reynolds- for the rest of the paper. Section 2 presents the simple introduc-
average N–S (RANS, [5,18–30]). The RANS method resolves the tion and derivation of the original form of N–S equation, with
mean flow numerically, and models the turbulent flow via tur- some discussions for the equation; in Section 3 viscous stress
bulence models, and it has the advantages of the moderate tensor and surface forces are illustrated, providing better under-
requirements in grid and temporal stepping (thus the gener- standings and physical significance to the viscous stress tensor
ally accepted computational burden), and of a good numerical and its components; Section 4 lays out 3 inconsistencies behind
convergence. The method has been proven to be very useful in the N–S equation, all based on the solid physical understanding
many practical flows, including the examples above. However, to the force analyses for the fluids; in Section 5, more details
it is also generally accepted the RANS method may not be very of fluid motions and forces are presented and the asymmetric
reliable in predicting the complicated flows since different turbu- friction tensor is proposed; in Section 6 Navier–Stokes equation
lence models may lead to different flow simulations. In the past is derived using the new total tensor for both incompressible and
decades, many researchers have tried to tune/modify/extend the compressible flows; and in Section 7, some more discussions are
turbulence models, but mostly for their specific problems, other made for the newly defined asymmetric stress tensor. In the last
section, conclusions are provided for the research work.
than for more general turbulence models. As a result of such
difficulties, so far there is hardly any universal turbulent model
2. The original form of N–S equation
for different flows, and we may even have difficulties in many
general flow problems, such as flow transition from laminar to
In this section, the original N–S equation is first derived briefly,
turbulent; the flows with adverse pressure gradients; and flow
to provide a base for further analyses to what problems may have
separations and re-attachments. The author thinks that it may be
behind the N–S equation.
a good idea to go back to the very fundamentals in fluid dynamics,
and the enhancement and better understanding of some concepts
2.1. Conservation of momentum
and principles behind the N–S equation may be needed, and
believes that a better physical understanding and significance to
The N–S equation can be easily derived from the transport
the fundamentals of fluid dynamic equation could pave a path
theorem and the conservation of momentum (note: the fluid
to construct better turbulence models for and thus the better
dynamic equation can also be derived in different ways, see [33,
solutions to the complicated flows. This is the main intention of
34]).
the research.
Based on the Newton second law of motion, the momentum
This research examines the very fundamentals of the N–S
of the fluid dynamics can be expressed as,
equation, for instance, the symmetric viscous stress tensor in
fluids introduced by Stoke in 1845 (named as the Stokes’ symmet- D y y {
⃗⃗ · ndS
ρ VdV = fdV + T (1)
ric stress tensor which originated from the Cauchy’s symmetric Dt V V S
62 W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71

Fig. 1. CFD applications in the design of Airbus A380 [6].

Fig. 2. Impact of CFD at Boeing. Green areas have strong CFD penetration; purple areas have some penetration; red areas present future opportunities [7]. (For
interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

In the equation, V and V represent different variables: the former where fi is the body force component and hereafter the sub-
is the fluid velocity (vector) and the latter the fluid volume scripts i, j = 1, 2, 3 mean the components along x-,y- and z-axes

⃗ is the total tensor (the double-sided arrow means a
(scalar); T respectively.
tensor) for surface stress (a force per unit surface area); f the body
2.2. Total and stress tensors in fluids
force (force per unit volume). For many applications, the simplest
body force is the gravitational force of the fluid, that is, f = ρ gk, Following the general convention, the total stress tensor has a
with g being the gravitational acceleration, k the unit vector along general form as,
z-axis.
T11 T12 T13
[ ]
By invoking the Gauss divergence theorem (see Appendix A.2) ⃗⃗ =
on the last term in the right-hand-side of Eq. (1), the momentum T T21 T22 T23 (5)
conservation can be expressed as, T31 T32 T33

D y y ( ) As such, the surface force (per unit surface area) acting on a


ρ VdV = f+∇ ·T
⃗⃗ dV (2) surface is calculated as
Dt V V F1 T11 T12 T13 n1
[ ] [ ][ ]
invoking the transport theorem, the conservation of momentum F= F2 = T⃗⃗ · n = T21 T22 T23 n2
(see Appendix A.3) leads to, F3 T31 T32 T33 n3
(6)
∂ (ρ V) T11 n1 + T12 n2 + T13 n3
[ ]
+ ∇ · (ρ VV) = f + ∇ · T⃗⃗ (3) = T21 n1 + T22 n2 + T23 n3
∂t
T31 n1 + T32 n3 + T33 n3
Using the Einstein summation convention, the conservation of
momentum can be simply expressed as where F1 , F2 , and F3 are three components along x-, y- and z-
axes; n is the normal vector of the surface, and n1 , n2 and n3 are
∂ (ρ ui ) ∂ ρ ui uj ∂ Tij
( )
the components of the normal vector along x1 -, x2 - and x3 -axes,
+ = fi + (4)
∂t ∂ xj ∂ xj respectively (see Fig. 3).
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 63

It should be noted that this relation is still controversial based on


the textbook ( [39], page 114).

2.3. Navier–Stokes equation

Based on the principles and assumptions above, the full


Navier–Stokes equation can be derived as

∂ (ρ u1 ) ∂ ρ u21 ∂ (ρ u1 u2 ) ∂ (ρ u1 u3 )
⎧ ( )

⎪ + + +
∂ ∂ ∂ x2 ( ∂ x3


⎪ t x1
∂ ∂ ∂ ∂ u2 ∂ u3
⎪ )
⎪ p 1 u1
+ µ∇ u1 + µ
2

= f1 − + +


Fig. 3. A stress force (vector) on a small surface, ∆S. ∂ x 3 ∂ x ∂ x ∂ x2 ∂ x3


⎪ 1 1 1
∂ (ρ u2 ) ∂ (ρ u1 u2 ) ∂ ρ u2

∂ (ρ u2 u3 )
⎪ ( 2)


⎪ + + +
∂t ∂ x1 ∂ x2 ( ∂ x3

(10)
∂p ∂ ∂ u1 ∂ u2 ∂ u3
)
In details, the definition of the total tensor was proposed by 1
= f2 − + µ∇ u2 + µ
2
+ +


Stokes [3] as ⎪

⎪ ∂ x2 3 ∂ x2 ∂ x1 ( ∂ x2) ∂ x3
∂ (ρ u3 ) ∂ (ρ u1 u3 ) ∂ (ρ u2 u3 ) ∂ ρ u3

−p 0 0 τ11 τ12 τ13 2
[ ] [ ] ⎪


⃗⃗ =
T 0 −p 0 + τ21 τ22 τ23
⎪ + + +
∂t ∂ x1 ∂ x2 ( ∂ x3



τ31 τ32 τ33 ∂p ∂ ∂ u1 ∂ u2 ∂ u3

0 0 −p
)
1


(7) + µ∇ 3 2
µ

⎩ = f3 −
⎪ u + + +
λ∇ · V 0 0 ∂ x3 3 ∂ x3 ∂ x1 ∂ x2 ∂ x3
[ ]
+ 0 λ∇ · V 0
0 0 λ∇ · V This is the widely accepted N–S equation for both compressible
and incompressible flows.
where λ is so called the secondary viscosity coefficient (see [33]),
In a vector form, the Navier–Stokes equation is expressed as
which is different from the physical viscosity coefficient µ of the
fluid. ∂ (ρ V) 1
Based on the conservation of angular momentum (from the + ∇ · (ρ VV) = fB − ∇ p + µ∇ 2 V + µ∇ (∇ · V) (11)
∂t 3
Cauchy’s second law of motion), the viscous stress tensor must
be symmetric, and the currently accepted stress components are or in the Einstein summation convention,
given as
∂ (ρ ui ) ∂ ρ ui uj ∂p 1 ∂ ∂ uj
( ) ( )
∂ u1 ∂ u2 ∂ u3 + µ∇ 2 ui + µ

+ = fi − (12)
⎪ τ11 = 2µ ; τ22 = 2µ ; τ33 = 2µ ∂t ∂ xj ∂ xi 3 ∂ xi ∂ xj
∂ ∂ ∂ x3

x x

1 2


∂ ∂
⎪ ( )
u u

1 2

⎨ τ12 = τ21 = µ ∂ x + ∂ x



2 1 3. Viscous stress tensor and surface forces
(8)
∂ u1 ∂ u3
( )
τ = τ = µ +

⎪ 13 31
∂x ∂ x1

Here the viscous stress tensor means the part of the stress

( 3


∂ ∂
⎪ )
u u3 tensor due to the fluid viscosity, written as

2

⎩ τ23 = τ32 = µ

⎪ +
∂ x3 ∂ x2 [
τ11 τ12 τ13
]
The first term in the right-hand-side of Eq. (7) is pressure ten- τ⃗⃗ = τ21 τ22 τ23 (13)
sor, and the diagonal tensor means the pressure could produce τ31 τ32 τ33
only normal forces on the surfaces (the negative sign means the
pressure force pointing into the body); the second term is the From the textbooks and research papers, it is frequently seen
stress tensor due to the fluid viscosity, which is very similar to that the stress tensor components are shown in Fig. 4 (note: on
pressure and has the same unit as the pressure. The difference surface S2 the normal points opposite to x2 -direction). However,
between the pressure force and the stresses is that the pressure to better understand the figure, some delicate explanations must
is a scalar at one point in the fluid, but the stress is a tensor, of be given to the illustration. Taking the small cube (enlarged
which the components have orientations when act on a surface.
for illustration) and considering the special orientations of the
For example, the stress component, τ21 , denotes a stress acting in
the y-direction on a surface of constant x; and τ12 , denotes a stress chosen plates, S1 , S2 and S3 , using Eq. (6), we can easily show that
acting in the x-direction on the surface of constant y. Similarly, there are only three simple surface force components on each of
τ11 denotes a normal stress acting on a surface of constant x those specific planes, see Eq. (14) and Fig. 5(a). Here the normal
(see [35,36]). vectors for these specific planes are still indicated by n1 , n2 and
The third term is a term of the fluid volume dilatation con- n3 for a clarification.
tributing to the total stress tensor, and Stokes [3] added for
τ11 τ12 τ13 τ11 n1

n1
[ ][ ] [ ]
accommodating the fluid compressibility. But it is still a con-
τ21 τ22 τ23 = τ21 n1

troversial term [37,38]. Obviously, this term disappears for in- Fs1 = 0



τ τ32 τ33 τ n

compressible flows. It is reported that it is be very important in ⎪ 0
[ 31 ] [ 31 1 ]


the analysis of the shock wave structure where the pressure and τ11 τ12 τ13 τ12 n2

0
⎪ ][

temperature could change dramatically in short time and short Fs2 = τ21 τ22 τ23 n2 = τ22 n2 (14)
distances [33]. Stokes [3] also proposed a relation between the ⎪ τ τ32 τ33 0 τ n
[ 31 ] [ 32 2 ]


physical and the secondary viscosity coefficients as τ11 τ12 τ13 0 τ13 n3

⎪ ][

τ21 τ22 τ23 = τ23 n3

2 F = 0

s3

λ=− µ (9)

τ31 τ32 τ33 τ33 n3

3 n3
64 W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71

(if the solid deformation is within the elastic deformation). In


a word, fluids can only have shear stress components when it
flows. This is a very important statement which will be used for
discussing the inconsistencies on the shear stress of fluids.

4.2. Inconsistency 1: symmetric shear stress in fluids

The Cauchy’s second law of motion is based on the assumption


of the conservation of angular momentum. This law was proposed
by Cauchy for elastic materials that the stress tensor must be
symmetric in the equilibrium state due to the force balance.
The Cauchy’s concept of the symmetric stress tensor was
adopted for fluid dynamics by Stokes in 1845 [3]. Analogous to
the Newton’s formula for fluid friction, Stokes formulated the
symmetric stress tensor, i.e., the viscous stress tensor in fluid
Fig. 4. Illustration of the stress tensor. dynamics, and the tensor components given in Eq. (8). To date,
this is well accepted definition, and hardly there are any doubts
for this definition since 1845, even though some research results
have shown that the viscous stress tensor may be asymmetric
For the specific faces S1 , S2 and S3 , their normal vectors are
for some special flows, such as compressible flows as seen in the
actually n1 = 1; n2 = −1; n3 = 1, therefore,
cases of rarefied gases, shock waves and gaseous flows through
τ11 τ12 τ13 τ11 n1 τ11

n1
[ ][ ] [ ] [ ]
micro fluidic channels [32,37,41,42], but these research results
Fs1 = τ21 τ22 τ23 = τ21 n1 = τ21

0


⎪ have been only taken as the exceptional cases in the fluid dynam-
τ τ32 τ33 τ n τ31

⎪ 0 ics community. For instance, Wilcox [24] has especially stated
[ 31 ] [ 31 1 ] [


τ11 τ12 τ13 τ12 n2

0 −τ12
][ ]
that the symmetric stress tensor is for simple viscous fluids,


Fs2 = τ21 τ22 τ23 n2 = τ22 n2 = −τ22 (15) but not for some anisotropic liquids (see page 39). However, no
⎪ τ τ32 τ33 0 τ n −τ32 reason is given for the statement.
[ 31 ] [ 32 2 ] [


τ11 τ12 τ13 0 τ13 n3 τ13
⎪ ][ ]
Another example for the inconsistent symmetrical stress ten-


τ τ22 τ23 = τ23 n3 = τ23

F = 0


⎪ s3 21 sor is given in the textbook of Kundu etc. ([31], page 126), with
τ31 τ32 τ33 τ33 n3 τ33

n3 a statement as: the stress tensor symmetry is violated in the
The transformed result is shown in Fig. 5(b), which is actu- electric field on polarized fluid molecules, where antisymmetric
ally the same as those in Fig. 4. This explains why the surface stresses must be included in the analysis.
forces can be specified using the stress tensor components (on A question would be if the Cauchy’s law of angular moment
S2 , the negative signs mean the opposite directions of the force conservation is a universal law for fluids, which requires the fluid
components). viscous shear stress must be symmetric. Then why there are so
many exceptions? After all, a universal law must be satisfied for
4. Inconsistencies on shear stress for N–S equation all cases as we do not see any exceptions for Newton’s laws of
motion. Obviously, there is an inconsistency in some practical
4.1. Fluid definition and shear stress in fluids examples when the fluid viscous stress tensor is required to be
symmetric. Hence the inconsistency is termed as Inconsistency
Let us go back to the very fundamental concept for fluid: 1 here.
what is fluid? Here a stricter technical or physical definition
of fluid is referred: based on Britannica Encyclopaedia (https:// 4.3. Inconsistency 2
www.britannica.com/science/fluid-physics), Fluid, including any
liquid or gas or generally any other material, cannot sustain a According to the friction definition of a Newtonian fluid, if a
tangential/shear force when at rest and could undergo a contin- fluid velocity gradient is between adjacent fluid particles, a fluid
uous change in shape under such a stress. In other word, under friction occurs due to the fluid viscosity, as
a shear stress (regardless how small it could be), a continuous ∂u
and irrecoverable change of position of the material forms a flow, τ12 (τxy ) = µ (16)
∂y
which is a very basic property of fluids. In contrast, the shear
stresses can be maintained within a deformed elastic solid, and This is actually a formula for defining the fluid viscosity coeffi-
the deformed solid could spring back to its original shape when cient, and an illustration can be seen in Fig. 6(a).
the stresses are removed. From this example in Fig. 6(a), it can be seen obviously that the
While in Bertin and Smith [33], it states that the technical shear stress or friction given by Eq. (16), the shear stress tensor
distinction between a fluid and a solid lies with their reaction component, τ12 , has a direction along with the velocity increment,
to an applied shear/tangential stress acting to them: a solid can ∆u, that is, τ12 = µ ∂∂ uy has a direction on x-axis. Similarly, in
resist a shear stress by a static deflection; a fluid cannot. Any Fig. 6(b), the normal stress, τ11 , has a direction with the velocity
shear stress applied to a fluid, no matter how small, will result increment, ∆u, hence the normal stress τ11 = µ ∂∂ ux has a direction
in a flow of that fluid, a continuous deformation of the shape. A on x-axis. Hence, τ11 and τ12 both have a direction in x-axis,
very similar statement can be found in White [40]. but acting on different surfaces of constant x and constant y,
In both definitions, a significant distinction between a fluid respectively, see Fig. 5.
and a solid is that a fluid cannot resist a shear stress without a It will be seen later in Eq. (27) that the fluid friction direc-
permanent deformation (i.e., flowing) regardless how small of the tion has always the same direction with the velocity increment,
shear stress could be, while a solid could spring back to original regardless of whether the velocity increment is on x-, y- or
form after the removal of the shear stress acting on the solid z-directions.
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 65

Fig. 5. Surface forces and stress tensor components.

Fig. 6. Shear stress due to the velocity gradient.

Next the symmetric shear stress is examined. Taking τ12 as the flow between the plates can be regarded as a laminar flow,
an example, it is a stress acting in the x-direction (subscript see Fig. 7(a).
‘1’) upon a surface of constant y (subscript ‘2’) (see [33,35] and The dynamic problem can be simplified mathematically as a
Fig. 5(b)). Based on the Stokes’ symmetric stress tensor, the tensor 2D flow as shown in Fig. 7(b). From the plot, we have: u = u(y),
component is defined as, and v = w = 0. As such, the N–S equation (zero pressure gradient
here) is degraded to,
∂ u ∂v ∂u ∂v
( )
τ12 =µ + =µ +µ (17) ∂ 2u
∂y ∂x ∂y ∂x =0 (18)
∂ y2
This tensor component consists of two terms in the right-hand-
side in Eq. (17). In physics as shown above, the first shear stress Applying the boundary conditions: u = 0 at y = 0; u = u0 at
term has a direction along x-axis due to the increment ∆u while y = h, the solution of the Eq. (18) is
the second shear stress term has a direction along y-axis due y
u = u0 (19)
to ∆v. These two orthogonal stress terms (along x- and y-axes h
respectively) together make a single stress tensor component τ12 . This is the analytical solution of the Couette flow, and the flow
In addition, following the definition of stress tensor definition velocity distribution is shown in Fig. 7(b).
in [35], the tensor component, τ12 , is acting in x-axis on a plane Based on the definition of the symmetric stress tensor compo-
of constant y, while its symmetric pair τ21 (of a same definition) nents in Eq. (8), we have
is acting along the y-direction on a plane of constant x. From
∂u ∂v u0
the physical standpoint, the symmetric shear stress component τ12 = µ +µ =µ (20)
is ambiguously defined, which has no clear physical significance. ∂y ∂x h
Therefore, a question will be: how a shear stress acting in one Following the definition of the orientation of the stress tensor
direction could physically hold two terms of different orientations component (see Fig. 5), it is a stress in x-axis, which is caused
of frictions? Here this conflict is termed as Inconsistency 2. due to the different velocities of u in y-direction.
From the symmetrical stress tensor, it requires
4.4. Inconsistency 3 u0
τ21 (= τ12 ) = µ (21)
h
Here we examine the classic Couette flow. The Couette flow It must be noted that this is a stress in y-direction.
is confined between two large parallel plates: the lower plate is As shown in Fig. 8, for a small rectangular fluid element, on
fixed and the upper plate moves at a constant speed (u0 ). The its four sides, there are 4 stresses: τ12 & τ21 , τ12

& τ21

making
speed of the upper plate is taken to be relatively small, so that two pairs of stresses due to the symmetric stress tensor. For the
66 W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71

Fig. 7. Couette flow: Fluid flow between two plates.

Fig. 9. The fluid particle A and its neighbouring point B.

The corresponding velocity components at B can be simply put


Fig. 8. Stresses in the simple Couette flow based on Eqs. (20) and (21). as
u1 = u0 + ∆ u
{
v1 = v0 + ∆v (22)
Couette flow, the horizontal viscous stresses τ12 and τ12 on the ′
w1 = w0 + ∆w
upper and lower sides can be easily understood, and they have
an obvious physical foundation, since the stress in x-direction The fluid friction (from the friction definition of a Newtonian
can cause a flow of the fluid (a constant deformation) in the fluid) is caused due to the velocity difference between the two
same direction. However, from the requirement of the symmetric neighbouring points, A and B, i.e., the relative velocity compo-
stress tensor the vertical viscous stresses τ21 and τ21 ′
must exist nents (B to A) can be calculated as
as shown in Fig. 8. The question here is what causes the vertical
stresses in the Couette flow? Or if there exist vertical stresses τ21 ∂u ∂u ∂u
⎡ ⎤
∆x + ∆y + ∆z
and τ21′
, why they do not cause a flow along y-direction? [
∆u
] ⎢ ∂x ∂y ∂z ⎥
As it is already shown in the fluid definition, a shear stress in a
⎢ ∂v ∂v ∂v ⎥
∆v =⎢ ∆x + ∆y + ∆z
⎢ ⎥
∂x ∂y ∂z

fluid can only exist when it flows. But for the Couette flow, there ∆w ∂w ∂w ∂w
⎢ ⎥
is only a flow in x-direction, but no flow in y-direction, then how
⎣ ⎦
∆x + ∆y + ∆z
the vertical shear stresses are caused? ∂x ∂y ∂z
(23)
∂u ∂u ∂u
⎡ ⎤
The similar conflict can be seen from the solution of the classic
flow in a horizontal pipe (see examples in [34]). ⎢ ∂x ∂y ∂ z ⎥ [ ∆x ]
The shear stress acting on a fluid causes a flow of the fluid,
⎢ ∂v ∂v ∂v ⎥
=⎢ ⎥ ∆y
⎢ ⎥
and as a result of the flow, the flow induced fluid friction could ∂x ∂y ∂ z ⎥ ∆z
∂w ∂w ∂w ⎦

balance the shear stress. This is very different from the solids in ⎣
which a force balance must be required on the conservation of the ∂x ∂y ∂z
angular moment, that is, the symmetric stress tensor in solids.
From Eq. (23), it can be seen that the directions of fluid friction
Therefore, this is a conflict, termed as Inconsistency 3 in this
research. is always in the same direction with the increment of velocity
component, that is, ∆u in x-direction, ∆v in y-direction and ∆w
in z-direction.
5. New understanding of fluid motion and force

5.1. Relative velocity between two points in fluid 5.2. Fluid friction tensor

At a certain time t, the fluid point A is taken as a reference


point, given by a position vector rA = (x0 , y0 , z0 )T , and the The relative velocity components between the joint fluid
corresponding velocity components as (u0 , v0 , w0 )T . At the same points are the requisite condition for inducing viscous forces
time, one of its neighbouring point B is positioned by a vector, (frictions) in fluid. Based on this physical understanding and the
rB = (x0 + ∆x, y0 + ∆y, z0 + ∆z)T . So the vector from A to B is definition of fluid friction (note: friction is used here to avoid
calculated as dr = (∆x, ∆y, ∆z)T . (See Fig. 9.) using the word ‘stress’), it is reasonable to define a friction tensor
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 67

σ⃗⃗ as with the symmetric part as following,


∂u ∂u ∂u ∂u ∂v ∂w
⎡ ⎤ ⎧
⎢ µ ∂x µ µ
⎪ S11 = ; S22 = ; S33 = ;
∂ ∂ ∂z

∂y ∂z x y
⎥ ⎪

⎢ ∂v ∂v ∂v

1 ∂u ∂v
⎥ ⎪
⎪ ( )
σ⃗⃗ = ⎢ µ µ µ ⎥ = µa⃗⃗

(24)
⎢ ⎥
⎨ S12 = S21 = 2 ∂ y + ∂ x

⎢ ∂x ∂y ∂z


⎣ ∂w ∂w ∂w

⎦ (31)
µ µ µ 1 ∂u ∂w
( )
∂x ∂y ∂z ⎪ S13 = S31 =

⎪ +


⎪ 2 ∂z ∂x
with the friction tensor components being given as ∂v ∂w
⎪ ( )
1



⎩ S32 = S23 =
⎪ +
∂u ∂u ∂u 2 ∂z ∂y

⎪ σ11 = µ ; σ12 = µ ; σ13 = µ
∂ ∂ ∂z

x y and the anti-symmetric part with the tensor components defined


∂v ∂v ∂v


σ21 = µ ; σ22 = µ ; σ23 = µ (25) as
∂ x ∂ y ∂z 1 ∂w ∂v 1 ∂u ∂w
( ) ( )
∂w ∂w ∂w

ω1 = ; ω2 =

⎪ − − ;
⎩ σ31 = µ ; σ32 = µ ; σ33 = µ ∂y ∂z 2 ∂z ∂x

⎪ 2
∂x ∂y ∂z (32)
1 ∂v ∂u
( )
Obviously, the friction tensor is an asymmetric tensor, which is ω3 = −
2 ∂x ∂y
different from the symmetric stress tensor, shown in Eq. (8).
here ω1 , ω2 , ω3 are the components of the angular velocity vector
And the fluid velocity gradient tensor (an asymmetric tensor) or vorticity vector, ω, following the definition given in [34],
defined as
1

∂u ∂u ∂u
⎤ ω= ∇ ×V (33)
2
⎢ ∂x ∂y ∂z ⎥ [
a11 a12 a13
] From Eq. (30), it can be seen that the velocity gradient tensor

⃗⃗ = ⎢ ∂v ∂v ∂v ⎥
can be written into two parts: the symmetric part and the anti-
a ⎥ = a21 a22 a23 (26)

∂x ∂y ∂z

a31 a32 a33 symmetric part. The symmetric strain-rate has been adopted in
∂w ∂w ∂w
⎢ ⎥
Stokes stress tensor for deriving the N–S equation.
⎣ ⎦
∂x ∂y ∂z The basic difference between the symmetric stress tensor and
the asymmetric friction tensor is the last pure rotation term in
The corresponding friction force on a unit surface is given as Eq. (30). If we consider a special case of an irrotational flow, we

∂u ∂u ∂u
⎤ have
⎢ µ ∂x µ ∂y µ
∂z ⎥[ ∇ ×V=0 (34)
Fx ⎥ n1
[ ] ]
⎢ ∂v ∂v ∂v
F= Fy = σ⃗⃗ · n = ⎢ µ µ µ ⎥ n2 Under such a flow condition, the last term in Eq. (30) disappears.
⎢ ⎥
Fz ∂ x ∂y ∂z ⎥ n Therefore, these two tensors: the symmetric stress tensor and the
⎣ ∂w ∂w ∂w

⎦ 3
µ µ µ asymmetric friction tensor; are exactly same.
∂x ∂y ∂z For the Couette flow, we can show that it is rotational, since
(27)
∂u ∂u ∂u
⎡ ⎤
µ n1 + µ n2 + µ n3 ⎥ 1 u0
∂x ∂y ∂z ω3 = − ̸= 0 (35)
2 h

⎢ ∂v ∂v ∂v ⎥
=⎢ µ n1 + µ n2 + µ n3 ⎥ meaning,
⎢ ⎥
∂x ∂y ∂z
∂w ∂w ∂w
⎢ ⎥
⎣ ⎦ ∇ × V ̸= 0 (36)
µ n1 + µ n2 + µ n3
∂x ∂y ∂z From the practical point of view, the rotational flow condition
holds in most flows within the viscous boundary layers, in which
Now we come back to the Couette flow, the friction tensor com-
the perfect pairs of the viscous stress tensor components do not
ponent τ12

give the friction along x-axis as exist.
∂u u0
σ12 = µ =µ (28) 6. New concepts for Navier–Stokes equation
∂y h
6.1. Surface force for fluid dynamics
but the corresponding vertical shear stress component is
∂v For fluid dynamics, the forces acting on the fluid include both
σ21 = µ =0 (29) the body and surface forces, with the surface force being calcu-
∂x ⃗
⃗′ , as shown in Eq. (1). However,
lated based on the total tensor, T
Obviously, there is no friction on y-direction. Therefore in this based on the discussions above, a change in the total tensor is
way the physical conflict of the stresses in the Couette flow is proposed to make it more physical.
resolved. ⃗⃗′ is defined as
Now the new total tensor T
−p 0 0 λ∇ · V 0 0
[ ] [ ]
⃗⃗′ =
T 0 −p 0 + 0 λ∇ · V 0
5.3. Analysis of velocity gradient tensor 0 0 −p 0 0 λ∇ · V
∂u ∂u ∂u
⎡ ⎤

The velocity gradient tensor can be rewritten as


⎢ µ ∂x µ
∂y
µ
∂z ⎥ (37)
⎢ ∂v ∂v ∂v ⎥
+⎢ µ µ µ
⎢ ⎥
S11 S12 S13 0 −ω3 ω2 ⎢ ∂x ∂y ∂z
[ ] [ ] ⎥
⎣ ∂w ∂w ∂w

⃗⃗ =
a S21 S22 S23 + ω3 0 −ω1 (30) ⎦
µ µ µ
S31 S32 S33 −ω2 ω1 0 ∂x ∂y ∂z
68 W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71

In this new definition of the total tensor, the symmetric stress


tensor is replaced by the friction tensor, such a new definition
(the friction from the stress tensor) could avoid the conflicts
behind the original N–S equation, including:

- In fluids, shear stresses may only exist when the fluids


are flowing. This is very different from the solids. Hence,
the requirement of the symmetric stress tensor developed
for solids may not be correct for fluids, because we have
already had exceptions (see [24,32,37,41,42]). In [31] (see
page 126), it states that the stress tensor symmetry is vio-
lated in the electric field on polarized fluid molecules, where Fig. 10. A shear stress τ23 is applied on a solid element.
antisymmetric stresses must be included in the analysis.
In fact, this problem can be solved using the asymmetric
friction tensor proposed in the research, Eq. (30), because This is exactly same N–S equation for incompressible flows as that
both the symmetric and anti-symmetric stresses have been of original form of N–S equation.
automatically included. As such, Inconsistency 1 is resolved.
This means that the N–S equation is exactly same as the orig-
- All components of the fluid friction tensor have a correctly- inal form as Stokes established in 1845, however, some concepts
defined physical property, including the consistent and and principles are different and of more physics.
unique orientations of the frictions, hence Inconsistency 2
is resolved.
6.3. Compressible fluids
- Using the new friction tensor in Eq. (24), rather than the
symmetric viscous stress tensor defined in Eq. (8), Inconsis-
For compressible flows, the Navier–Stokes equation can be
tency 3 for the Couette flow is removed, shown by Eqs. (28)
derived as
and (29).
∂V
[ ]
ρ + (V · ∇) V = fB − ∇ p + µ∇ 2 V + λ∇ (∇ · V) (44)
6.2. Incompressible fluids ∂t
To make the equation exactly same as the original Navier–Stokes
For an incompressible fluid, its density is constant. Together equation for compressible flows as Eq. (12), it is assumed as
with∇ · V = 0, Eq. (3) can be rewritten as
1
∂V 1( ⃗⃗′
) λ= µ (45)
+ (V · ∇) V = fB + ∇ · T (38) 3
∂t ρ
which is slightly different from the proposal that Stokes made in
with 1845, see Eq. (9). This difference is caused because of the different
∂ p ∂σ11 ∂σ12 ∂σ13
( )
symmetric stress tensor and the asymmetric friction tensor.
∇ · T⃗⃗′ = − + + + i The choice of Eq. (45) seems reasonable since the rate of
∂x ∂ x1 ∂ x2 ∂ x3
volume dilatation appears 3 times in the Navier–Stokes equation
∂ p ∂σ21 ∂σ22 ∂σ23
( )
+ − + + + j (39) (in each component in the Cartesian coordinate system), so the
∂y ∂ x1 ∂ x2 ∂ x3 coefficient of 1/3 is taken. However, more work must be done on
∂ p ∂σ31 ∂σ32 ∂σ33
( )
how the second viscosity coefficient can be decided.
+ − + + + k
∂z ∂ x1 ∂ x2 ∂ x3
7. Discussions
Using the new friction tensor of Eq. (25),
∂σ11 ∂σ12 ∂σ13 ∂ 2 u1 ∂ 2 u1 ∂ 2 u1 7.1. A shear stress applied on a solid element and a fluid element
+ + = µ 2 + µ 2 + µ 2 = µ∇ 2 u1 (40)
∂ x1 ∂ x2 ∂ x3 ∂ x1 ∂ x2 ∂ x3
Here we will see what will happen if a shear stress is applied
Similarly, we have
on a solid element and a fluid element, and will explain why the
∂σ21 ∂σ22 ∂σ23

asymmetrical stress tensor will not cause a fluid spinning in the
⎪ + + = µ∇ 2 u2
∂ x1 ∂ x2 ∂ x3 fluid.

∂σ31 ∂σ32 ∂σ33 (41)
If the element is an elastic material (e.g., a solid), under the
⎪ + + = µ∇ 2 u3
∂ x1 ∂ x2 ∂ x3 shear stress τxz , there will be a deformation of the element.

However, due to the elasticity, the element will produce a shear
Putting these together, we have the Navier–Stokes equation for
stress τzx to balance the applied shear stress and avoid an infinite
the incompressible fluid as
rotation of the element as seen in Fig. 10. Also, τzx = τxz (this is
∂ u1 ∂ u1 ∂ u1 ∂ u1 1 1 ∂p required for solids by the Cauchy symmetric stress tensor).

⎪ + u1 + u2 + u3 = f1 − + ν∇ 2 u1 If the element is a fluid, the applied shear stress, τxz , will sim-
∂ ∂ ∂ ∂ ρ ρ ∂ x1

t x x x

1 2 3

⎨ ∂u

∂ u2 ∂ u2 ∂ u2 1 1 ∂p ply cause a flow of the element (see based on the fluid definition).
2
+ u1 + u2 + u3 = f2 − + ν∇ 2 u2 Due to the flow motion, a fluid friction is induced which would
∂ t ∂ x 1 ∂ x2 ∂ x3 ρ ρ ∂ x2
∂ u3 ∂ u3 ∂ u3 ∂ u3 1 ∂p


⎪ 1 balance the applied shear stress. Therefore, this shows that there
+ ν∇ 2 u3

⎪ + u1 + u2 + u3 = f3 − will be no shear stress pair to balance the applied shear stress in
∂t ∂ x1 ∂ x2 ∂ x3 ρ ρ ∂ x3

fluids, but a fluid friction to balance the applied shear stress (see
(42) Fig. 11), calculated as,
In a vector form, it is ∂u
τxz = µ (46)
∂V 1 1 ∂z
+ (V · ∇) V = fB − ∇ p + ν∇ 2 V (43)
∂t ρ ρ
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 69

Fig. 11. A shear stress τ23 is applied on a fluid element.

Fig. 13. An experiment setup for proving that the vertical shear stresses is zero.

Fig. 12. An opening on the upper plate, is there flow coming out due to the
vertical shear stress τyx ? To resolve these conflicts behind N–S equation, it is proposed
that the fluid friction is regarded as the special surface force, not
a part of the Stokes’ symmetric stress. Under this new concept,
the friction tensor is independent of the stress tensor, and more
7.2. A proof for no vertical shear stress in the Couette flow
importantly, the friction tensor is not required to be symmetric if
the real physics of the fluid frictions are referred to.
It may be argued that there may be a non-zero vertical shear
From the study and the understanding of the concepts and
stress τyx in the Couette flow, but the upper plate stops the
principles behind N–S equation in this research work, the follow-
vertical flow. Imagine if we make an open on the upper plate,
there will be a flow coming out from the opening because of the ing conclusions can be drawn:
vertical shear stress (see Fig. 12). Is this true?
- Identify the conflicts of the concepts and principles in de-
However, the answer is NO. Because in reality, whether there
riving the N–S equation.
will be a flow flowing out from the opening totally depending on
- Illustrate a conflict on the requirement of the symmetric
the pressure of the fluid between the plates. If the flow speed is
stress tensor using the well-known Couette flow.
large enough, there may be a suction of flow in the opening.
- Reformulate the total tensor by replacing the symmetric
Another proof is the experiment setup similar to the
Bernoulli’s principle. At the bottom of the water tank, two dif- stress tensor with the fluid friction tensor, and the friction
ferent size pipes are connected. Different sizes of the pipes tensor is more based on the real physics of the friction acting
would allow different flow velocities, as V1 and V2 . Based on the on the fluids.
Bernoulli’s principle, we could have different water heights in the - The formulation of new friction tensor could resolve all
small tubes, as h1 and h2 . identified inconsistencies, including the shear stress in the
If the non-zero vertical stress, τyx , exists, it surely induce Couette flow and the special problem as in the electric
flows in the vertical small tubes (based on the fluid definition). field on polarized fluid molecules, since the anti-symmetric
Therefore, there will be outflows in the small tubes regardless tensor is automatically included in the new total tensor
of the height of the tubes because of the non-zero vertical shear formulation.
stress. Obviously, this is not true. The heights of the fluid in the - The exactly same N–S equation as the original N–S equation
small tubes, h1 and h2 , totally depend on the pressures of the flow can be derived using the newly defined total tensor for
in the pipes. (See Fig. 13.) the incompressible flows. However, to obtain the same N–S
equation for compressible flows, a slight different yet very
8. Conclusions similar assumption from the Stokes’ assumption should be
taken.
The research has examined some fundamental issues behind
the Navier–Stokes equation, and it is identified that there exist The new friction tensor does lead to the exactly same N–S equa-
some conflicts on the concepts and principles in deriving the N–S tion, hence it is hoped that this revisit of N–S equation with the
equation. To solve these conflicts, the forces acting on the fluids newly proposed friction tensor could provide a better physics
have been re-analysed based on a solid physics. By considering in understanding the N–S equation. With the better physics as
the fundamental properties of fluids, it can simply show that the proposed in this research, it may pave a way to understand the
symmetric stress tensor is violated in the classic Couette flow. fluid dynamics better, especially for those complicated flows.
70 W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71

For instance, currently, solving the N–S equation uses turbulence References
models in most practical applications, and the turbulence models
have been frequently constructed using the symmetric stress [1] J. Anderson, A History of Aerodynamics, and Its Impact on Flight Machines,
tensor following the requirement of the symmetric strain-rate in: Cambridge Aerospace Series, vol. 8, Cambridge University Press, 1997.
[2] J. Anderson, Brief history of the early development of theoretical and
tensor, see [21,43]. Under the new physical significance presented experimental fluid dynamics, in: Encyclopedia of Aerospace Engineering,
in the research work, the asymmetric friction tensor could allow John Wiley & Sons, Ltd., 2010.
different views on the complex flows and it may shed some light [3] G.G. Stokes, On the theories of the internal friction of fluids in motions,
for reconstructing different turbulence models so to help to solve and the equilibrium and motion of elastic solids, Trans. Camb. Philos. Soc.
the fluid dynamic problems better. 8 (1845) 75–129.
[4] O. Darrigol, Between hydrodynamics and elasticity theory: The first five
births of the Navier–Stokes equation, Arch. Hist. Exact Sci. 56 (2) (2002)
Declaration of competing interest 95–150.
[5] C.G. Speziale, Analytic methods for the development of Reynolds stress
The authors declare that they have no known competing finan- closures in turbulence, Annu. Rev. Fluid Mech. 23 (1991) 107–157, http:
cial interests or personal relationships that could have appeared //dx.doi.org/10.1146/annurev.fl.23.010191.000543.
[6] F.D. Witherden, A. Jameson, Future directions of computational fluid
to influence the work reported in this paper.
dynamics, in: 23rd AIAA Computational Fluid Dynamics Conference, AIAA
2017-3791. Denver, Colorado, US, 2017.
Appendix. Useful mathematical tools [7] P.R. Spalart, Venkatakrishnan, On the role and challenges of CFD in the
aerospace industry, Aeronaut. J. 120 (1223) (2016) 209–232, http://dx.doi.
org/10.1017/aer.2015.10.
A.1. Material derivative [8] D.R. McDaniel, R.H. Nichols, S.A. Morton, Capabilities and challenges in
CFD: A perspective from the DoD HPCMP CREATETM-AV kestrel develop-
ment team, in: 23rd AIAA Computational Fluid Dynamics Conference, 5–9
In the Euler expression, a given physical function, f, in the June 2017, Denver, Coorado, USA, 2017.
fluid motion can be expressed as the function of the independent [9] L. Cozzi, et al., Facing the challenges in CFD modelling of multistage axial
variables: the coordinates (x, y, z) and time, t, as compressors, in: Proceedings of ASME Turbo Expo 2017: Turbomachinery
Technical Conference and Exposition, June 26–30, 2017, Charlotte, NC, USA,
f = f (x, y, z ; t) (A.1) 2017.
[10] D.J. Garmann, M.R. Visbal, Challenges & perspectives on CFD and experi-
Note: the arbitrary function can be either a scalar, such as the mental interactions for complex unsteady flows (Invited), in: AIAA Aviation
fluid density, ρ ; pressure, p, etc.) or a vector, such as, the fluid Forum, June 25–29, 2018, Atlanta, Georgia, US, 2018.
velocity, V, or the force, F and so on (hereafter the bold letters [11] P. Moin, K. Mahesh, Direct numerical simulation: A tool in turbulence
denote vectors). research, Annu. Rev. Fluid Mech. 30 (1998) 539–578.
[12] T.K. Sengupta, S. Bhaumik, DNS of Wall-Bounded Turbulent Flows: A First
The material derivative of the function can be easily obtained Principle Approach, Springer, 2019.
using the chain rule of derivative as [13] N.J. Geogiadis, D.P. Rizzetta, C. Fureby, Large-eddy simulation: Current
Df ∂f ∂f ∂f ∂f capabilities, recommended practices, and future research, in: Proceedings
= +u +v +w (A.2) of 47th Aerospace Sciences, Jan. 5–8, 2009, Orlando, Florida, US, 2009.
Dt ∂t ∂x ∂y ∂z [14] C. Fureby, Large eddy simulation: A useful tool for engineering fluid dy-
namics, in: Proceedings of 18th Australasian Fluid Mechanics Conference,
In a more compact manner, it is
3–7 Dec. 2012. Launceston, Australia, 2012.
Df ∂f [15] S.B. Pope, Ten questions concerning the large-eddy simulation of turbulent
= + (V · ∇) f (A.3) flows, New J. Phys. 6 (2004) 1–24, http://dx.doi.org/10.1088/1367-2630/6/
Dt ∂t 1/035.
The flow acceleration is the material derivative of the flow ve- [16] R. Rogallo, P. Moin, Numerical simulation of turbulent flows, Annu.
locity vector with regard to time and it can be derived in similar Rev. Fluid Mech. 16 (1984) 99–137, http://dx.doi.org/10.1146/annurev.fl.
16.010184.000531.
manner as above, so
[17] Z. Yang, Large-eddy simulation: Past, present and the future, Chin. J.
DV ∂V Aeronaut. 28 (1) (2015) 11–24, http://dx.doi.org/10.1016/j.cja.2014.12.007.
= + (V · ∇) V (A.4) [18] K. Hanjalic, B.E. Launder, Modelling Turbulence in Engineering and the
Dt ∂t Environment: Second-Moment Routes to Closure, Cambridge University
The first term of the right-hand-side of Eq. (A.4) is the local Press, UK, 2011.
acceleration (the independent coordinate variables, (x, y, z), are all [19] B.E. Launder, D.B. Spalding, The numerical computation of turbulence
kept constant), and the second term is the convective acceleration flows, Comput. Methods Appl. Mech. Engrg. 3 (1974) 269–289.
[20] B.E. Launder, G.J. Reece, W. Rodi, Progress in the development of a
due to the different velocities of the fluid particles over different
Reynolds-stress turbulence closure, J. Fluid Mech. 68 (1975) 537–566.
positions. [21] P.R. Spalart, S.R. Allmaras, A one-equation turbulence model for aerody-
namic flows, in: 30th Aerpspcae Science Meeting & Exhibit, Jan. 6–9, 1992,
A.2. Gauss convergence theorem Reno, Nevada, USA, 1992.
[22] C.G. Speziale, S. Sarkar, T.B. Gatski, Modelling the pressure–strain correla-
tion of turbulence: an invariant dynamical systems approach, J. Fluid Mech.
Gauss divergence theorem is
227 (1991) 245–272, http://dx.doi.org/10.1017/S0022112091000101.
{
⃗⃗ · ndS = y ∇ · TdV
T
⃗⃗ (A.5)
[23] D.C. Wilcox, Reattachment of the scale-determining equation for advanced
turbulence models, AIAA J. 26 (1) (1988) 1299–1310, http://dx.doi.org/10.
S V 2514/3.10041.
[24] D.C. Wilcox, Turbulence Modelling for CFD, third ed., DCW Industries, Inc.,
A.3. Transport theorem 2006.
[25] D.C. Wilcox, Formulation of the k-w turbulence model revisited, AIAA J.
The transport theorem is a very important tool to derive the 46 (11) (2008) 2823–2838, http://dx.doi.org/10.2514/1.36541.
[26] F.R. Menter, Two-equation eddy-viscosity turbulence models for engineer-
equation for fluid motion. For a given physical function, f, in the
ing applications, AIAA J. 32 (8) (1994) 1598–1605, http://dx.doi.org/10.
fluid domain, V, the transport theorem gives 2514/3.12149.
D y y ∂f [27] F.R. Menter, M. Kuntz, R. Langtry, Ten years of industrial experience with
[ ]
fdV = + ∇ · (f V) dV (A.6) the SST turbulence model, in: K. Hanjalic, Y. Nagano, M. Tummers (Eds.),
Dt V V
∂t Turbulence, Heat and Mass Transfer, 2003.
[28] F.R. Menter, Review of the shear-stress transport turbulence model ex-
The details of the derivation of the transport theorem can be perience from an industrial perspective, Int. J. Comput. Fluid Dyn. 23 (4)
found in [35, pages 57–59]. (2009) 305–316, http://dx.doi.org/10.1080/10618560902773387.
W. Sheng / European Journal of Mechanics / B Fluids 80 (2020) 60–71 71

[29] F.R. Menter, Turbulence Modeling for Engineering Flows: A Technical [36] J.J. Bertin, R.M. Cummings, Aerodynamics for Engineerings, sixth ed.,
Paper, ANSYS INC., 2011, Cited at. Pearson, 2013.
[30] P.A. Durbin, Some recent development in turbulence closure model- [37] H. Brenner, Navier–Stokes revisited, Phyisca A 394 (2005) 60–132, http:
ing, Annu. Rev. Fluid Mech. 50 (2018) 77–103, http://dx.doi.org/10.1146/ //dx.doi.org/10.1016/j.physa.2004.10.034.
annurev-fluid-122316-045020. [38] H. Brenner, Beyond Navier–Stokes, Internat. J. Engrg. Sci. 54 (2012) http:
[31] P.K. Kundu, I.M. Cohen, D. Rowling, Fluid Mechanics, sixth ed., Elsevier, //dx.doi.org/10.1016/j.ijengsci.2012.01.006.
Inc., 2016. [39] E.L. Houghton, et al., Aerodynamics for Engineering Students, sixth ed.,
[32] J. Rigelesaiyin, et al., Asymmetry of the atomic-level stress tensor in Elsevier, Ltd., 2013.
homogeneous and inhomogeneous materials, Proc. R. Soc. A 474 (2018) [40] F.M. White, Fluid Mechanics, eighth ed., McGraw-Hill Education, New York,
20180155, http://dx.doi.org/10.1098/rspa.2018.0155. US., 2016.
[33] J.J. Bertin, M.L. Smith, Aerodynamics for Engineers, second ed., [41] E. Zhang, et al., Asymmetric tensor analysis for flow visualization, IEEE
Prentice-Hall, Inc., US, 1989. Trans. Vis. Compu. Graph. 15 (1) (2009) 106–122, http://dx.doi.org/10.
[34] H. Schlichting, K. Gersten, Boundary-Layer Theory, ninth ed., Springer- 1109/TVCG-2007-12-0188.
Verlag, Berlin Heidelberg, 2017. [42] M.J. Vedan, S.M. Panakkal, Vorticity and stress tensor, J. Inform. Math. Sci.
[35] J.N. Newman, Marine Hydrodynamics, The MIT Press, Cambridge, 10 (2018) 351–357.
Massachusetts, USA, 1977. [43] D.C. Wilcox, Formulation of the k-w turbulence model revisited, in: 45th
AIAA Aerospace Sciences Meeting and Exhibit, Jan. 8-11, 2007, Reno,
Nevada, USA, 2007.

Potrebbero piacerti anche