ME 563 - Intermediate Fluid Dynamics - Su Lecture 8 - More Viscous Flow Examples

ME 563 - Intermediate Fluid Dynamics - Su
Lecture 8 - More viscous flow examples

Reading: Acheson, 2.32.5.
In the last class we illustrated some basic viscous flow ideas by finding the velocity field for the
steady flow of viscous, incompressible fluid driven by a constant pressure gradient between two
parallel plates (plane Poiseuille flow). In this lecture we will look at a more sophisticated example.
The added sophistication doesnt arise particularly in the physics of the problem, but rather in the
mathematical tools necessary for solving the differential equations.
Flow due to an impulsively moved plane boundary
(This example is from the book. Well work it out in full detail here.)
Consider the system in Fig. 1. A viscous, incompressible fluid occupies the region y > 0, which
is bounded by a solid wall at y = 0. The wall is initially at rest, but at t > 0 the wall has velocity
U in the x-direction. The flow is not pressure-driven. We want to compute the velocity field
u(x, t) = (u, v, w) for t > 0.
Figure 1: An impulsively moved plane boundary.

As usual we want to simplify the problem as much as possible before looking at equations.
First of all, the geometry of the problem doesnt vary in the x- and z-directions, so we assume
that the velocity components are only functions of y. Also, there is nothing driving the flow in the
z-direction, so we assume that w = 0. This gives us u = (u(y, t), v(y, t), 0).
Because the flow is incompressible, u = u/x + v/y = 0. We have u/x = 0 since
u = u(y, t), so v/y = 0. We know v = v(y, t), so v/y = 0 means v is constant everywhere
but we have as a boundary condition that v = 0 at y = 0, so finally, we get v = 0 everywhere.
Thus
u(x, t) = (u(y, t), 0, 0).
(1)
(This is a plane parallel shear flow.)

The Navier-Stokes momentum equation is:
u
p
+ (u )u =
+ 2 u + g.
t
We are only interested in the u-component of this equation, which, for the velocity field of (1),
works out as
u
2u
= 2.
t
y
(2)
We have eliminated the pressure gradient term because the flow is not pressure-driven. This
equation is called the one-dimensional diffusion equation, and arises with diffusive properties such
as heat. (The thermodynamic analogue of this problem would be if our fluid had initially constant
temperature, and suddenly the temperature of the boundary was changed.)
The conditions on this problem are:
Initial condition: u(y, 0) = 0 for y > 0.
Boundary conditions: u(0, t) = U and u(, t) = 0 for t > 0.
(3)
We are going to look for something called a similarity solution. There is a lot of lore about
similarity solutions in the mathematical literature, but unfortunately there seem to be no general
rules about when you can and cant look for similarity solutions and even if there were, the math
would be way too complicated for us. Instead, well rely on reasoning.
Lets consider whats happening in this system. Initially the fluid and the wall are at rest
(Fig. 1). Suddenly, the wall is set in motion at velocity u = U . Because the fluid is viscous, the
fluid in contact with the wall also moves at speed U by the no-slip condition. This establishes a
velocity difference, or gradient, with the layer of slower fluid immediately above. The resulting
viscous shear stress acts to accelerate that layer of slower fluid. As its velocity increases, this layer,
in turn, acts to accelerate the slower fluid above it, and on and on. At any time in the process,
we get a velocity distribution that looks something like that on the right in Fig. 1. Basically, the
moving wall introduces momentum to the system, which gets diffused upward by the fluid viscosity
as time proceeds.
What can we say about the shape of the velocity profile as time proceeds? One thing we can
assume is that U is a scale factor for the velocity field, so a change in U would have a direct, linear
effect on the velocity values u(y). For example, if we double U , then we would just double all of
the velocity values in the fluid for a given t.
Another thing we notice is that the process by which the different layers of fluid continually
accelerate slower fluid above them has no stopping point, because the fluid domain is infinite in
the positive y-direction. So there is no fixed length scale set by the boundary conditions in the
y-direction. This suggests that if we look at the velocity profile at successively later times, the
shape will be the same but the profile will just be continually stretched in the y-direction. What
is the stretching factor for the profile height?
The text describes one way to find this stretching (or normalizing) factor, . Another way
is to use dimensional analysis. The height of the profile can only depend on the time, t, and the
viscosity, . The only other parameter in the problem is U , and weve already reasoned that it just
scales the velocity values. (Looking at the u profile in Fig. 1, this means that U scales the profile
2
horizontally, and doesnt affect the height in y.) The viscosity, , has units of (length)
/(time), so
the only way to form a variable, , with units of length from and t is to set = t.
So we are looking for a velocity field of the form
u = U f (), where =
y
y
= .
By the chain rule, we have (using the notation df /d = f 0 ())

u
y
= f 0 ()
= f 0 () 1/2 3/2 ,
t
t
2 t
u
1
= f 0 ()
= f 0 () 1/2 1/2 , and
y
y
t
2
u
1
= f 00 () .
y 2
t
2
Inserting into (2), we get

f 0 ()
y
2 1/2 t3/2
1
= f 00 () , or
t
1
f 0 = f 00 .
2
(4)
Observe that this is a first-order differential equation, because one term involves the first derivative
and the other term involves the second derivative. So we can define g() = f 0 (), and inserting
into (4), we get
1
dg
g =
.
2
d
We can use the separation of variables to write this as
1
dg
d =
.
2
g
Integrating both sides, we get
1
2 + C = ln(g),
4
where C is an arbitrary constant. Taking the exponential of both sides gives us
g = eC e
2 /4
= Be
2 /4
df
,
d
where B = eC and we have re-substituted g = df /d.

Integrating (5) once, we get
Z
2
f () = A + B
es /4 ds.
(5)
(6)
We use the dummy variable of integration s, because we want to evaluate the integral at specific
limits so we can impose boundary conditions, and there is no analytical expression for the integral
2
of es /4 . Our boundary conditions were given in (3). Since the similarity function f was defined
by u = U f (), our conditions on f are f (0) = 1 and f () = 0. The condition f (0) = 1 is easy,
because the integral in (6) goes to zero, so A = 1. To evaluate f at = , we exploit the fact that
the definite integral
Z
2
es /4 ds =
0
is found in integral tables. So for f to satisfy the boundary condition at = requires that
B = 1 . So finally, we get as our solution for u

u(y, t) = U
1
1
s2 /4
e
0
y
ds , where = .
t
The velocity field defined by (7) basically looks like the one depicted in Fig. 1.
(7)
Diffusion of vorticity/flow with circular streamlines
We said before that viscosity was responsible for the diffusion of both momentum and vorticity.
This can be seen explicitly by looking at the equation for the evolution of the vorticity field, (x, t):
+ (u ) = ( ) u + 2 ,
t
(8)
in which the term 2 describes the viscous diffusion of vorticity, analogously to the viscous term
in the Navier-Stokes momentum equation.
To illustrate the viscous diffusion of vorticity, it is convenient to consider flows with circular
streamlines. These are most naturally described in cylindrical coordinates. In particular, we are
interested in velocity fields of the form
u = u (r, t) e .
(9)
The book gives the Navier-Stokes equations in cylindrical coordinates in 2.4. The one were
interested in is the equation for the time evolution of the u velocity component, which is

u
u r u
2 ur
1 p
u
2
+ (u ) u +
=
+ u + 2
2 ,
t
r
r
r
r
u
where u = ur
+
+ uz ,
z
r r 2
1
2
1
and 2 =
r
+ 2 2 + 2.
r r
r
r
z
For the velocity field of (9), we then have (u ) u = 0 and 2 u =
u
1 p
=
+
t
r
1 u
2 u
u
+
2
r 2
r r
r
2 u
r 2
1 u
r r ,
which gives us
(10)
Consider the pressure term. Every other term in (10) is a function of r and t only, so we can write
p
= P (r, t).
Integrating, we get
p = P (r, t) + C(r, t)
where C(r, t) is a constant of integration with respect to the variable , which means that it can
vary with r and t. We can say that P (r, t) has to be zero, because otherwise p() 6= p( + 2), i.e.
p would have multiple values at a given position. Thus p is not a function of , so (10) becomes

2
u
1 u
u
u
(11)
+
=
2 .
t
r 2
r r
r
We will use this to explore particular flows with circular streamlines in the next lecture.

ME 563 - Intermediate Fluid Dynamics - Su Lecture 8 - More Viscous Flow Examples

Uploaded by

Copyright:

Available Formats

ME 563 - Intermediate Fluid Dynamics - Su Lecture 8 - More Viscous Flow Examples

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ME 563 - Intermediate Fluid Dynamics - Su Lecture 8 - More Viscous Flow Examples

Uploaded by

Copyright:

Available Formats

ME 563 - Intermediate Fluid Dynamics - Su

Lecture 8 - More viscous flow examples

Flow due to an impulsively moved plane boundary

Figure 1: An impulsively moved plane boundary.

(This is a plane parallel shear flow.)

By the chain rule, we have (using the notation df /d = f 0 ())

Inserting into (2), we get

where B = eC and we have re-substituted g = df /d.

Diffusion of vorticity/flow with circular streamlines

You might also like