Vous êtes sur la page 1sur 16

2 Elementary Viscous Flow

2.1 Introduction
We find that there is a very thin boundary layer in inviscid flow past a fixed aerofoil.
In this boundary layer, inviscid theory fails and viscous effects become important while they are negligible
in the main part of the flow.
Consider a simple shear flow such as
u = (u(y), 0, 0).
It is known that the shear stress τ is proportional to the velocity gradient dy .
du

i.e.
du
τ =µ (2.1)
dy
where µ is the coefficient of viscosity.
We define the kinematic viscosity as
µ
ν= (2.2)
ρ

Note.
1. The viscous effect is important in a thin boundary layer since the velocity gradients in a boundary
layer are much greater than in the main part of the flow due to a substantial changes in velocity across
a very thin layer.
In other words, for a small value of µ in a main stream, the viscous effect is negligible, while τ cannot
be neglected because of the large factor du
dy across the thin boundary layer.

2. The boundary layer can separated from the surface (boundary) in certain circumstances.
For example, consider the flow of a low-viscosity fluid past a circular cylinder.
In reality, the boundary layer separation occurs which causes a large vorticity-filled wake (b) instead
of irrotational flow (a).
This can be explained as follow.
At the point A and C, we have the local maximum pressure so that

pA = p C

And at the point B, we have the local minimum pressure.


This implies that between B and C, there is a substantial increase in pressure along the boundary
in the direction of flow.
i.e. severe adverse pressure gradient along the boundary, which causes the boundary layer to separate.
3. The behaviour of a fluid of small viscosity µ may be completely different to that of a (hypothetical)
fluid of no viscosity.

10
2.2 The Equations of Viscous Flow
Suppose we have an incompressible Newtonian fluid of constant density ρ and constant viscosity µ.
Its motion is governed by the Navier-Stokes equation
∂u 1
+ (u · ∇)u = − ∇p + ν∇2 u + g
∂t ρ (2.3)
∇·u=0

It is known that in viscous flow, both normal and tangential components of fluid velocity at a rigid boundary
must be equal to those of the boundary itself.
i.e. if the boundary is not moving, then u = 0.
The condition on the tangential component of velocity is known as the no-slip condition.

A Reynolds number is defined as


UL
R= (2.4)
ν
where

U :the characteristic flow speed


L :the characteristic flow scale

Note that the derivatives of the velocity components such as ∂u


∂x is of order

U speed
=
L length
2
and similarly, ∂∂xu2 is of order LU2 .
From the Navier-Stokes equation,
� �
U2
inertial term : |(u · ∇)u| = O
L
� � (2.5)
U
viscous term : |ν∇2 U| = O ν 2
L

so that � � � 2� �
� inertial term �
� � = O U ν U = O(R) (2.6)
� viscous term � L L2

i.e. the Reynolds number give a rough indication of the relative magnitude between inertial term and viscous
term in the Navier-Stokes equation.
For example, R � 1 implies that we have the motion of fluid of small viscosity. i.e. the viscous term is
negligible and the inertial term dominates.
However, viscous effects become important in thin boundary layers (due to the small value of L)

It is known that the thickness δ of a boundary layer is related to the Reynolds number as
� �
δ 1
=O √
L R
i.e. the larger the Reynolds number, the thinner the boundary layer.

11
2.3 Some Simple Viscous Flows. The Diffusion of Vorticity
Plane parallel shear flow

Consider a viscous fluid with velocity


u = (u(y, t), 0, 0).
Such a flow is called a plane parallel shear flow and we notice that
∇ · u = 0.
In the absence of gravity, the Navier-Stokes equation becomes
∂u 1
+ (u · ∇)u = − ∇p + ν∇2 u
∂t ρ
so that � � � �
∂u ∂ ∂ ∂ 1 ∂p ∂2 ∂2 ∂2
+ (u, 0, 0) · , , (u, 0, 0) = − +ν + + (u, 0, 0)
∂t ∂x ∂y ∂z ρ ∂x ∂x2 ∂y 2 ∂z 2
or
∂u 1 ∂p ∂2u
=− +ν 2 (2.7)
∂t ρ ∂x ∂y
with
∂p ∂p
= = 0.
∂y ∂z
∂p ∂2u
From (2.7), the term ∂x is the difference of two x-independent terms ( ∂u
∂t and ∂y 2 ), hence we conclude that
∂p
∂x is a function of t alone.

Example 2.1. Alternatively, we can derive the equation (2.7) from


du
τ =µ
dy
as follow.
In the absence of viscous forces and gravity, Euler’s equation for u = (u(y, t), 0, 0) is
∂u 1
+ (u · ∇)u = − ∇p
∂t ρ
or
∂u ∂p
ρ =− .
∂t ∂x

Consider an element of fluid of unit length in the z-direction and a cross section in the xy plane.

12
The net pressure force on the element in the x-direction is
� �
p(x + δx) − p(x) ∂p
p(x)δy − p(x + δx)δy = − · δxδy ≈ − δxδy.
δx ∂x
But by Newton’s second law, we have
∂p Du
− δxδy = ρ δxδy ·
∂x Dt
� �
∂u ∂u
= ρ δxδy · +u
∂t ∂x
∂u
= ρ δxδu ·
∂t
i.e.
∂u ∂p
ρ=− . (2.8)
∂t ∂x
Similarly, the net viscous shear forces on the top and the bottom of the element is
 � � 
� � ∂u � ∂u �
� δx − � δx
∂u �� ∂u ��  ∂y
y+δy
∂y
y 
µ δx − µ δx = µ   δxδy
∂y �y+δy ∂y �y δy

∂2u
≈µ δxδy.
∂y 2
Hence the net viscous shear forces per unit volume in the fluid element (δV = δxδy · 1) is

∂2u
µ .
∂y 2
From the equation (2.8), the Euler’s equation with viscosity is given by

∂u ∂p ∂2u
ρ =− +µ 2
∂t ∂x ∂y
which is equivalent to
∂u 1 ∂p ∂2u
=− +ν 2
∂t ρ ∂x ∂y
where the kinematic viscosity ν is given by
µ
ν= .
ρ

The flow due to an impulsively moved plane boundary

Suppose that viscous fluid lies at rest in the region 0 < y < ∞ and at t = 0, the rigid boundary y = 0 is
suddenly jerked into motion in the x-direction with constant speed U .
By the non-slip condition (the boundary layer moves with the boundary itself), the fluid elements in contact
with the boundary will move with velocity U immediately.

13
Assume that the flow is driven only be the motion of the boundary so that there is no external pressure
gradient.
i.e.
p(x = −∞) = p(x = ∞).
Hence we have
∂p
= 0.
∂x
Then the equation (2.7) becomes one-dimensional diffusion equation

∂u ∂2u
=ν 2 (2.9)
∂t ∂y

with the initial condition


u = (y, 0) = 0, y>0
and the boundary conditions
u(0, t) = U u(∞, t) = 0
for the time t > 0.
Note that if we apply a transformation for some constant α such as

y −→ αy and t −→ α2 t

so that
u(y, t) −→ u(y � , t� ) = u(αy, α2 t)
then the new function u(y � , t� ) still satisfy the equation (2.9).
i.e. if u(y, t) is a solution of (2.9), so is u(αy, α2 t).

To see this, we have y � = αy and t� = α2 t.


Then
∂u ∂u ∂t� ∂u
= � = α2 �
∂t ∂t ∂t ∂t
and � � � �
∂2u ∂ ∂u ∂ ∂u ∂y � ∂y � ∂2u
= = = α2 �2
∂y 2 ∂y ∂y ∂y � ∂y � ∂y ∂y ∂y
So
∂u ∂2u ∂u ∂2u
=ν 2 implies �
= ν �2 .
∂t ∂y ∂t ∂y
1
This fact suggests that the solution for the equation (2.9) is a function of y and t of the combination of y/t 2 ,
since this similarity variable is unchanged by such a transformation.

We assume that the solution is of the form


u = f (η)
for
y 1
η = √ = y(νt)− 2
νt
Then we have (� ≡ dη )
d

∂u y 1 1
= −f � (η) 1 3 = − f � (η) · η ·
∂t 2ν 2 t 2 2 t

14
and

∂u ∂η 1
= f � (η) = f � (η) √
∂y ∂y νt
� �
∂2u ∂ 1 ∂η 1 1 1
2
= f� √ = f �� √ · √ = f ��
∂y ∂η νt ∂y νt νt νt

Then the equation (2.9)


∂u ∂2u
=ν 2
∂t ∂y
becomes
1 1 1
− f � · η · = f ��
2 t t
or
1
f �� + ηf � = 0.
2
By integrating with respect to η, we obtain
1 2
f � = Be− 4 η

and � η
1 2
f =A+B e− 4 s dx
0
for some constant A and B.

From the boundary condition u(0, t) = U (η = 0 for y = 0),

f (0) = A + B · 0 = U

so that A = U .
And the boundary condition u(∞, t) = 0 (η → ∞ for y → ∞),
� ∞
1 2 √
f (∞) = U + B e− 4 s dx = U + B · π = 0
0

so that B = − √Uπ .
Therefore � �

1 η
− 14 s2
u=U 1− √ e ds (2.10)
π 0

Note.

1. At√time t1 , the velocity u is a function of y/ νt1 and at time t2 , the velocity u is a same function of
y/ νt2 and so on.
As the time t goes on, the velocity profile u(y) becomes stretched out while the geometrical shape of
solution remain similar, in the absence of an upper boundary.
2. The viscous effects gradually communicate the motion of the boundary to the whole fluid. √
i.e. the effects of the motion of the plane boundary are largely confined√to a distance of order νt
from the boundary for a fixed time t. (e.g u is less than 1% of U at y = 4 νt)

15
This leads the idea of the diffusion of vorticity.
Recall the in tow-dimensional flow with u = (u, v, 0), the vorticity ω is

∂v ∂u
ω= − .
∂x ∂y

In this case, we have ∂v


∂x = 0 and
� � η �
∂u ∂ U 1 2
ω=− =− U−√ e− 4 s ds
∂y ∂y π 0
� η
U ∂ � − 1 η2 �
=√ e 4 dη
π 0 ∂y
� η
U d � − 1 η2 � dη
=√ e 4 dy (2.11)
π 0 dy dy
� η � �
U 1 1 2
=√ ·√ d e− 4 η
π νt 0
U y2
=√ · e− 4νt
πνt
for η = √y .
νt √
This is exponentially small beyond a distance of order νt from the boundary.
The diffusion of vorticity from a plane boundary occurs so that the vortex sheet forms an infinite
concentration of vorticity at y = 0 and t → 0 while no vorticity elsewhere (y > 0 and t → 0).

3. Vorticity diffuses a distance of order νt in time t.

y y2
η=√ or t=
νt η2 ν
i.e. the time taken from vorticity to diffuse a distance of order L is of the order
� 2�
L
viscous diffusion time = O
ν

The flow on inclined plane

Consider the steady flow under gravity down on inclined plane.

16
By the non-slip condition, we must have

u=0 when y = 0.

So u must depends on y (otherwise whole flow would not move relative to the boundary) and u is independent
of time t since it is steady ( ∂u
∂t = 0).
Since this is two-dimensional flow, we can assume that

u = (u(y), v(y), 0)

But since it is incompressible flow, we must have

∇·u=0

so that
dv
=0
dy
Hence the initial assumption of u reduced to

u = (u(y), 0, 0).

From Navier-Stokes equation


∂u 1
+ (u · ∇)u = − ∇p + ν∇2 u + g,
∂t ρ
we have
1 ∂p d2 u
0 = − + ν + g sin α
ρ ∂x dy 2
(2.12)
1 ∂p
0 = − + − g cos α
ρ ∂y
in x and y component respectively.

By integrating the second equation with respect to y, we have

p = −ρgy cos α + f (x) (2.13)

where f (x) is an arbitrary function of x.

Since all the streamlines are parallel to the plane, at y = h, the tangential stress must be zero and the
pressure p must be equal to atmospheric pressure p0 .
i.e. �
du ��
µ =0 and p(h) = p0 .
dy �y=h
From (2.13),

p(y) = −ρgy cos α + f (x)


p0 = −ρgy cos α + f (x).

By subtracting, we have
p − p0 = pg(h − y) cos α
or
p = ρg(h − y) cos α + p0

17
i.e. p is a function of y only.
∂p
Hence we have ∂x = 0 and (2.12) becomes

d2 u
ν = −g sin α (2.14)
dy 2
with boundary conditions �
u=0 when y = 0
dy = ν dy = 0
µ du when y = h
du

By integrating (2.14) with respect to y


du
ν = −gy sin α + A
dy
and from the second boundary condition, we obtain

A = gh sin α

so that
du
ν = −gy sin α + gh sin α.
dy
By integrating again with respect to y,
1
νu = − gy 2 sin α + ghy sin α + B
2
and form the first boundary condition, we obtain B = 0.
Hence
� �
1 1
u= ghy sin α − gy 2 sin α
ν 2 (2.15)
g
= y(2h − y) sin α.

i.e. the velocity profile u(y) is parabolic as in the diagram.

The volume flux down the plane per unit length in the z-direction is given by
� h � h
g g 2 gh3
Q= u du = sin α (2hy − y 2 ) dy = sin α · h3 = sin α.
0 2ν 0 2ν 3 3ν

Example 2.2. Consider the viscous fluid lies at rest in the region 0 < y < h.
Suppose that at t = 0, the rigid boundary y = 0 is suddenly moved with speed U while an upper boundary
y = h is at rest.
As in the previous discussion, we have
u = (u(y, t), 0, 0)
which satisfies the diffusion equation
∂u ∂2u
=ν 2 (2.16)
∂t ∂y
subject to the initial condition
u(y, 0) = 0, 0<y<h

18
and the boundary conditions �
u(0, t) = U t>0
u(h, t) = 0 t>0
equilirium
The complementary solution uc for (2.16) can be found by setting

∂2u
ν =0
∂y 2
so that
u = Ay + B
for some constants A and B.
From the boundary conditions, we obtain

u(0) = B = U and u(h) = Ah + U = 0

so that A = − Uh .
Hence �
U y�
uc = − y+U =U 1− .
h h
The general solution for (2.16) is � y�
u = uc + up = U 1 − + up
h
where up is the particular solution satisfying

∂up ∂ 2 up
=ν (2.17)
∂t ∂y 2
with corresponding initial conditions and boundary conditions as
 � �
y
up (y, 0) = −U 1 − h
 0<y<h
up (0, t) = 0 t>0


up (h, t) = 0 t>0

Note that the boundary conditions are now homogeneous.

Set
up = Y (y)T (t).
Then (2.17) becomes
dT d2 Y
Y = νT 2
dt dy
or
1 dT ν d2 Y
= .
T dt Y dy 2
So we have, for some positive constant k

1 dT 1 d2 Y
= −νk 2 and = −k 2 (2.18)
T dt Y dy 2

From the first of (2.18), we obtain


2
T (t) = e−νk t .

19
From the second of (2.18), we obtain

Y (y) = C cos ky + D sin ky

for some constant C and D.


By the homogeneous boundary conditions for up ,

Y (0) = C = 0 and Y (h) = D sin ky = 0.

We set D = 1 for the simplicity, we obtain



k= n = 1, 2, 3, · · ·
h
so that � nπu �
Y (y) = sin
h
and

up = T (t)Y (y)
� 2 2 t� � nπy �
n π νy
= exp − · sin .
h2 h
Hence the general solution for up can be written as
�∞ � 2 2 t� � nπy �
n π νy
upn = An exp − 2
· sin .
n=1
h h

But from the initial condition for up , we have


∞ An �
� nπy � � y�
sin = −U 1 − .
n=1
h h

By using the Fourier’s theory


� h� � nπy � � � h
mπy � � y� � mπy �

An sin · sin sin dy = −U 1 − · sin dy.
0 n=1 h h 0 h h

So
� �
2 h
y� � nπy �
An = − U 1− · sin dy
h h h
�0� �h � h �
2 h � nπy � � y� h U � nπy �
=− − cos ·U 1− + − · cos dy
h nπ h h 0 nπ 0 h h
�� � �� � h �
2 h U � nπy �
=− 0− − U − cos dy
h nπ nπ 0 h
� �
2 hU U h � � nπy ��h
=− − · · sin
h nπ nπ nπ h 0
2U
=− .

Then
�∞ � � � 2 2 � � nπy �
2U n π νt
up = − exp − 2
sin .
n=1
nπ h h

20
Finally, the general solution for u = uc + up is
� � 2 2 �
y � 2U � 1 � nπy �

n π νt
u(y, t) = U 1 − − exp − sin . (2.19)
h π n=1 n h2 h

Note.
h2
For a sufficiently long time later t � ν , the exponential term in (2.19) can be neglected and the flow almost
reach its steady state so that � y�
u≈U 1−
h
And the vorticity ω is almost distributed uniformly throughout the fluid.

21
2.4 Flow with Circular Streamlines
We have Navier-Stokes equation without gravity
∂u 1
+ (u · ∇)u = − ∇p + ν∇2 u
∂t ρ
∇ · u = 0.

In cylindrical polar coordinates (r, θ, z), we have

∂Φ 1 ∂Φ ∂Φ
∇Φ = er + eθ + ez
∂r r ∂θ ∂z
1 ∂ 1 ∂uθ ∂uz
∇·u= (rur ) + +
r ∂r � � r ∂θ 2 ∂z 2 (2.20)
1 ∂ ∂ 1 ∂ ∂
∇2 = r + 2 2+ 2
r ∂r ∂r r ∂θ ∂z
∂ uθ ∂ ∂
(u · ∇) = ur + + uz .
∂r r ∂θ ∂z
Since     
er cos θ sin θ 0 i
eθ  = − sin θ cos θ 0  j 
ez 0 0 1 k
we have
∂er
= (− sin θ, cos θ, 0) = eθ
∂θ
∂eθ
= (− cos θ, − sin θ, 0) = −er (2.21)
∂θ
∂ez
=0
∂θ
and
∂ei ∂ei
= =0 for i = r, θ, z. (2.22)
∂r ∂z
So the term (u · ∇)u becomes
� �
∂ 1 ∂ ∂
(u · ∇)u = ur + uθ + uz (ur er + uθ eθ + uz eθ )
∂r r ∂θ ∂z
� �
∂er uθ ∂er ∂er
= (u · u)ur er + ur ur + + uz
∂r r ∂θ ∂z
� �
∂eθ uθ ∂eθ ∂eθ
+ (u · ∇)uθ eθ + uθ ur + + uz
∂r r ∂θ ∂z
� �
∂ez uθ ∂ez ∂ez
+ (u · ∇)uz ez + uz ur + + uz (2.23)
∂r r ∂θ ∂z
� uθ �
= (u · ∇)ur er + ur 0 + eθ + 0
� r �

+ (u · ∇)uθ eθ + uθ 0 + (−er ) + 0
r
+ (u · ∇)uz ez + uz (0 + 0 + 0)
� � �
u2θ ur uθ �
= (u · ∇)ur − er + (u · ∇)uθ + eθ + (u · ∇)uz ez
r r

22
Similarly, the term ∇2 u becomes

∇2 u =∇2 (ur er + uθ eθ + uz ez )
� � � �
1 ∂ ∂ 1 ∂ ∂ ∂2
= r u r er + 2 u r er + 2 u r er
r ∂r ∂r r ∂θ ∂θ ∂z
� � � �
1 ∂ ∂ 1 ∂ ∂ ∂2
+ r u θ eθ + 2 u θ eθ + 2 u θ eθ
r ∂r ∂r r ∂θ ∂θ ∂z
� � � �
1 ∂ ∂ 1 ∂ ∂ ∂2
+ r u z ez + 2 u z e z + 2 u z ez
r ∂r ∂r r ∂θ ∂θ ∂z
� � � � � �
1 ∂ ∂ur 2
∂ ur 1 ∂ ∂
= r + e r + u e
r r
r ∂r ∂r ∂z 2 r2 ∂θ ∂θ
� � � � � �
1 ∂ ∂uθ ∂ 2 uθ 1 ∂ ∂
+ r + eθ + 2 u θ eθ
r ∂r ∂r ∂z 2 r ∂θ ∂θ
� � � � � �
1 ∂ ∂uz ∂ 2 uz 1 ∂ ∂
+ r + ez + 2 u z ez
r ∂r ∂r ∂z 2 r ∂θ ∂θ
� � � � � �
1 ∂ ∂ur ∂ 2 ur 1 ∂ ∂ur ∂er
= r + er + 2 er + u r
r ∂r ∂r ∂z 2 r ∂θ ∂θ ∂θ
� � � � � �
1 ∂ ∂uθ 2
∂ uθ 1 ∂ ∂uθ ∂eθ
+ r + eθ + 2 eθ + u θ + ∇ 2 u z ez
r ∂r ∂r ∂z 2 r ∂θ ∂θ ∂θ
� � � � � �
1 ∂ ∂ur ∂ 2 ur 1 ∂ ∂ur
= r + er + 2 er + u r eθ (2.24)
r ∂r ∂r ∂z 2 r ∂θ ∂θ
� � � � � �
1 ∂ ∂uθ ∂ 2 uθ 1 ∂ ∂uθ
+ r + eθ + 2 eθ + uθ (−er ) + ∇2 uz ez
r ∂r ∂r ∂z 2 r ∂θ ∂θ
� � � � � �
1 ∂ ∂ur ∂ 2 ur 1 ∂ 2 ur ∂ur ∂er ∂ur ∂eθ
= r + er + 2 er + + eθ + u r
r ∂r ∂r ∂z 2 r ∂θ2 ∂θ ∂θ ∂θ ∂θ
� � � � � �
1 ∂ ∂uθ ∂ 2 uθ 1 ∂ 2 uθ ∂uθ ∂eθ ∂uθ ∂er
+ r + e θ + e θ + + (−e r ) − u θ
r ∂r ∂r ∂z 2 r2 ∂θ2 ∂θ ∂θ ∂θ ∂θ
+ ∇ 2 u z ez
� � � � � �
1 ∂ ∂ur ∂ 2 ur 1 ∂ 2 ur ∂ur
= r + e r + e r + 2 e θ − u e
r r
r ∂r ∂r ∂z 2 r2 ∂θ2 ∂θ
� � � � � �
1 ∂ ∂uθ ∂ 2 uθ 1 ∂ 2 uθ ∂uθ
+ r + e θ + e θ − 2 e r − u e
θ θ
r ∂r ∂r ∂z 2 r2 ∂θ2 ∂θ
+ ∇ 2 u z ez
� � � �
1 ∂ ∂ur 1 ∂ 2 ur ∂ 2 ur ur 2 ∂uθ
= r + 2 + er − 2 er − 2 er
r ∂r ∂r r ∂θ2 ∂z 2 r r ∂θ
� � � �
1 ∂ ∂uθ 1 ∂ 2 uθ ∂ 2 uθ uθ 2 ∂ur
+ r + 2 + eθ − 2 eθ + 2 eθ
r ∂r ∂r r ∂θ2 ∂z 2 r r ∂θ
+ ∇ 2 u z ez

23
Hence, in cylindrical polar coordinates, the Navier-Stokes equation can be written in components as
� �
∂ur u2θ 1 ∂p ur 2 ∂uθ
+ (u · ∇)ur − = − + ν ∇ ur − 2 − 2
2
∂t r ρ ∂r r r ∂θ
� �
∂uθ ur uθ 1 ∂p 2 ∂ur uθ
+ (u · ∇)uθ + = − + ν ∇ uθ + 22
− 2 (2.25)
∂t r ρr ∂θ r ∂θ r

∂uz 1 ∂p
+ (u · ∇)uz = − + ν∇2 uz
∂t ρ ∂z

Example 2.3. Consider a flow which has the circular streamline as

u = uθ (r, t)eθ .

Note that this flow is incompressible as


∇ · u = 0.
We have
� �
∂ uθ ∂ ∂
(u · ∇)u = ur + + uθ (0 + uθ (r, t)eθ + 0)
∂r θ ∂θ ∂z
uθ ∂ u2 ∂eθ u2
= (uθ eθ ) = θ = − θ er
�r ∂θ � � r ∂θ 2 r �
1 ∂ ∂ 1 ∂ ∂2
∇ u=
2
r + 2 2 + 2 (uθ eθ )
r ∂r ∂r r ∂θ ∂z
� � � �
1 ∂ ∂ 1 ∂ ∂
= r u r eθ + 2 u θ eθ
r ∂r ∂r r ∂θ ∂θ
� 2 �
∂ uθ 1 ∂uθ 1 ∂
= + eθ − 2 (uθ er )
∂r2 r ∂r r ∂θ
� 2 �
∂ uθ 1 ∂uθ uθ
= + − eθ
∂r2 r ∂r r2
So the Navier-Stokes equation in components becomes
u2θ 1 ∂p
0 + 0 − = − + 0
r ρ ∂r
� �
∂uθ 0 · uθ 1 ∂p ∂ 2 uθ 1 ∂uθ uθ
+ 0 · uθ + = − + ν 2
+ − 2
∂t r ρr ∂θ ∂r r ∂r r

1 ∂p
0 + 0 + 0 = − + 0
ρ ∂z
∂p
Since uθ is a function of r and t only, from the second equation, the same must be true for ∂θ .
i.e.
∂p
= P (r, t)
∂θ
for some function P depends only on r and t.
By integrating with respect to θ,
p = P (r, t)θ + f (r, t)

24
for some function f (r, t).
But we must have P (r, t) = 0, otherwise the pressure p would be different for the different value of θ.
Hence p is constant and ∂p ∂θ = 0. Therefore, the second equation becomes
� �
∂uθ ∂ 2 uθ 1 ∂uθ uθ
=ν + − 2 (2.26)
∂t ∂r2 r ∂r r

Example 2.4. Consider the flow as in the previous example. Suppose the flow is steady ( ∂u ∂t = 0) and the
fluid occupied the gap r1 ≤ r ≤ r2 between two circular cylinder which rotate with angular velocities Ω1 and
Ω2 respectively.
We have
d2 uθ duθ
r2 2 + r − uθ = 0.
dr dr
This is a Cauchy-Euler equation and an auxiliary equation is given by

am2 + (b − a)m + c = 0

or (a = 1, b = 1 and c = −1)
m2 − 1 = 0.
So m = ±1 and the general solution is
B
uθ = Ar +
r
for some constant A and B.
By the non-slip conditions, we have
B
uθ (r1 , t) = Ar1 + = r1 Ω 1
r1
B
uθ (r2 , t) = Ar2 + = r2 Ω 2
r2
And we obtain
Ω2 r22 − Ω1 r12 (Ω1 − Ω2 )r12 r22
A= and B=
r22 − r12 r22 − r12
It is known that if Ω1 is too large, the flow becomes unstable and superbly regular and axisymmetric Taylor
vortices appear.

25

Vous aimerez peut-être aussi