proceedings

Aerodynamic Lattice Calculations
Using Punched Cards
W HAT I am going to present to you is by no means a
finished product. It is riot ev·en something I am exceedingly
pr.oud of. I would like to show it to you in some detail,
and hope that I will receive from you some criticisms and
help on how we could have this computation done in a
much shorter time than it takes us now. It is possible that
we are on the wrong track entirely.
The problem of the turbine engineer with all its complications
is essentially as shown in Figure 1.
Nozzle Velocity
Relative
Bucket Velocity
FIGURE 1
HANS KRAFT
General Electric Company
Momentary
Streamlines
Most of you know that a turbine is essentially a windmill.
A very fast flow .of steam or gas issues from stationary
passages which are called nozzles. The moving blades
we of the General Electric Company call buckets. The
steam flow streaming from the nozzles passes between
them and is deflected. This process generates mechanical
power which is removed by the rotating shaft.
We have had a long history of experimentation. We
have experimented very intensively since 1920. We would
like to do some theoretical computations in addition. We
feel that we are somewhat against a blank wall with only
experimental approach. It is my own honest, private opinion
that further improvement in the performance of the
60
modern turbine-and it performs very well already-will
be made when, and only when, we are able to follow by
calculation the flow through this nozzle and bucket combination
with the buckets moving at high speed.
This means that we have to compute a flow through a
row of nozzle profiles. In aerodynamic language such a
row of equally spaced profiles is called a lattice. We will,
in addition, need to kn.ow the flow through the bucket lattice.
Furthermore, there is an interaction between these
nozzles and buckets. This interaction appears as a time
variation. As the buckets move past the nozzles, different
configurations of the available flow space result.
\Ve cannot rely much on the well-known approximati.on
of the flow by that of an incompressible fluid. Our velocities
are such that we always have to consider the fluid
as compressible. Thus, we must first of all learn to compute
compressible flow through a stationary, two-dimensionallattice
; later on we must study interference between
the two lattices as one passes by the other. Theoretically,
we think We know more or less how to handle the problem,
although as far as actual computation is c.oncerned, we
still have a long distance to go.
I should like to discuss some of the initial work which
we have done to describe a simple flow through a row of
buckets. It has been performed for flow of an incompressible
fluid, but was done in a manner identical to that to
be followed to give us the compressible counterpart of
this incompressible calculation. The compressible computation
still awaits the completion of a set .of input functions
before it can be performed.
To solve the incompressible problem Laplace's equation
must be solved. We do not attempt, however, to solve a
boundary value problem. \Ve try to learn to build up profiles
from given functions and accept the resulting shape
if it seems to be one which we actually do want, i.e., a
shape which will perform well.
We use the representation of a flux function t/t given as
a function over a field with the coordinates..X' and y. In the
compressible case we will not have this simple Laplace's