Sei sulla pagina 1di 3

6.

DISCUSSION ( From thesis, June 1971)

An interesting and curious result occurs when applying the


current method to the single tube problem of Olsen and Shapiro.
The results of Figure 9 are obtained by extrapolating theoretical
results for mean Reynolds numbers below 2000. It has been noted
that the maximum Reynolds number in the experiments of Olsen and
Shapiro was about 2500. For calculations using the new method it
was found that the solution would eventually go unstable if the
mean flow Reynolds number exceeded approximately 2000. This is particularly
interesting since ordinary instability techniques

(33, 34) applied to flow in a rigid tube fail to verify the insta-
bilities observed experimentally. These techniques predict stable
tube flow for all Reynolds numbers. It should be noted that ordinary
instability theory neglects the non-linear convection terms in the
momentum equation. This assumption is made so that a disturbance
which is a linear solution can be applied. The non-linear terms might
be important (34) for prediction of laminar instability. Also,
ordinary instability theory pertains only to a perfectly rigid wall.

Small vibrations in the tube wall caused by a coupling with the flow
might offer the mechanism to produce strong disturbances. The amplifica-
tion mechanism might be similar to a resonance phenomena in that
a constructive reinforcement may be produced which outweighs the
dissipative effects of fluid viscosity.

Although these numerical results are not conclusive, in that


a detailed parametric study of the phenomenon was not performed, some
interesting observations were made. It was determined that the insta-
bilities were apparently not the result of accumulated numerical
error. This was done by holding the boundary conditions fixed for
a long period of time and observing numerical fluctuations.
Secondly, it was observed that changes in the mesh size for the numeric-
al calculation did not alter the results greatly. Last and perhaps
most interesting, it was noted that the flow would not remain stable
if a disturbance due to wall motion had caused an inflection point
to occur in the velocity profile as the flow approached a steady
state. This was true only if the Reynolds number of the flow exceed-
ed 2000. Such a phenomenon is associated with the predicted insta-
bility of planar flows using established techniques.
The correctness of the observed instabilities is supported by
recent literature related to the numerical solution of fluid flow
problems. The following observations were noted.

1. A text by Ames (35) discusses the numerical stability of the


method of characteristics. He states that at present there is no
general theory to predict the stability of the complete set of non-
linear hyperbolic equations used in this thesis. The linear equations,
however, are subject to stability analysis. It is shown that the
method of characteristics employing the mesh of characteristics is
stable. For this technique the numerical accuracy is also optimum.

2. Ladyzhenskaya (36) has conducted a series of studies to inves-


tigate the capability of the Navier-Stokes equations to describe
a physical flow at all Reynolds numbers. She notes, in particular,
that the non-linearity of the Navier-Stokes equations is responsible

for a multitude of solutions to exist above a finite Reynolds


number, Re1. Furthermore, it is stated that above a Reynolds
number, Re2 > Re1 , no solutions exist. This would serve to explain
why no stable solutions could be obtained above Re = 2000 using
the method in this thesis.

3. Several recent numerical studies have shed light on the numerical


and physical instabilities encountered when using the non-linear
Navier-Stokes equations. DiPrima et al. (37) discussed the necessity
of retaining the non-linear terms in the numerical approach. In
particular he also mentions the importance of the detailed form of
the velocity profile and the destabilizing effects of viscosity
in the process of laminar instability. Both of these effects have
been accounted for in the present theory.

Loer (38) also points to the importance of retaining the


non-linear terms in the study of the instability of the laminar
boundary layer. He states that these non-linear terms tend to ampli-
fy large disturbances imposed upon the boundary layer. He has
succeeded in showing that the laminar boundary layer is unstable at
high Reynolds numbers (Rex > 10**5 ). These observations support the

physical correctness of the instabilities observed in the present method.


Jain (39) discusses the numerical solution of viscous, incom-
pressible flow about a circular cylinder. He states that it is
known that a well behaved solution of an unsteady non-linear
problem based on the Navier-Stokes equations may not exist during

the entire interval of time, t > 0; the solution may become irregular
after a finite time period. These observations also tend to agree
with the results obtained with the present method.

Dr. Larry Cooper 1-15-2011


==============================================================
=

Potrebbero piacerti anche