|
INTRODUCTION
Time is short today, and I want to get directly into
the technical improvements in our product lines which can be made by
utilizing the power of electronic computing devices. But I cannot resist
the temptation to make a few generalizations about the inter-relation of
problem and equipment. The structure of this program today, for instance,
reflects a common but not universally accepted opinion that the problems
of engineering, of accounting, of sales analysis, and of production
control require a different approach, and eyen different computing
equipment.
I should like to point to the possibility - indeed the probability
-that data processing systems of the future will be so flexible, so fast,
so capacious and so well utilized by professions programmers that they
will take on the characteristics needed by various problem areas when
required, and change over to different apparent characteristics
when applied in another function.
This is a real handy idea, since it relieves we of the necessity of describing
the type of machine best suited for engineering and technical
calculations, makes it unnecessary for me to tackle the question of
whether each function should have its own equipment because of specialized
requirements, and introduces enough "blue sky" to enable me to
wriggle off the hook during the question period. For instance, I
need no longer to distinguish between the power of digital and analog
equipment for engineering applications in much detail, since it is already
possible to point to digital simulation of analog equipment, and to analog
devices which embody many digits components and principles. It would
probably be fair to both sides of the analog-digital rivalry; to say that
at the present state of the art, analog computation is noted for its
convenience and digital for its power.
SHORT RANGE APPLICATIONS
When an engineering section first has access to electronic calculation
equipment, it most frequently develops simple applications first. And
these usually are done from the design area. For short, one might refer to
such problems as involving formula evaluation. in that the design
techniques are not new, have been expressed in simple mathematical
equations already, and are familiar to the customer (pardon me for the use
of the term "customer" - I have been operating in a service
capacity, and charging for those services, so I think of the engineers and
scientists whom we serve in those terms!). The power of the electronic
computer, analog or digital, here resides in the economies of time and
money it makes possible. The work could be done just as well by
slide rule, by graphical techniques, or by desk calculators. Indeed, it
could probably be accomplished in many cases by construction of
experimental models, thus avoiding all mathematization. But it can be done
more cheaply and more rapidly by using electronic equipment. In military
work, in new fields of advanced technology, and in areas where the
competition is particularly keen, time considerations commonly outweigh
money economies. In more conservative areas -the lamp business,
transformers, and so on - time may be available, and the important
consideration will be reduction of engineering costs.
An allied application, and one frequently used as an introduction, is
that of test data reduction. All through our business are engineers who
want to record and reduce the results of tests on physical equipment. Some
are writing down ammeter readings on the back of envelopes with a blunt
pencil, while at the other extreme completely automated data systems may
be listening to guided missiles in flight, or to reactors nearing the
critical phase, Computing equipment can make little contributions to the
single-setup test - the Research Laboratory, for instance, makes little
use of automatic data reduction techniques. But where simple tests are
done ever and ever, many economies are definitely possible, and where
tests are complicated and the data reduction problem is difficult,
economies of time make it possible to increase the utilization of million
dollar facilities. so that time economies lead, as they always must, to
profit rewards. This has been the case in the jet engine business; raw
test data is transmitted from engine cells in Lynn to our computing
facilities in Evendale, and reduced there with very expensive equipment;
we may be shooting sparrows with a coast defense gun, but it pays off tremendously
by making the results of the previous day's test available immediately to
those conducting the next one. .
All engineering sections facing the problem of major testing costs
should immediately consider the advisability of recording their results in
digital form suitable for insertion directly into a general purpose
digital computer, or the construction or purchase of analog equipment
designed to transform the readings directly into final form as part of the
test setup, ready for engineering analysis.
Both the substitution of numbers into designed formulas and the
reduction of test data present little conceptual difficulty. When we
consider the problems of aerodynamics, of heat transfer. of neutron flux -
in general, one might say "field" problems - design
techniques turn out to revolve around the solution ~f partial differential
equations and systems of ordinary differential equations, usually
non-linear and often time-dependent. Here - only the most powerful new
electronic' systems are capable of making a contribution at all. Design
techniques in these areas had to wait the invention and development of
computer tools; the work was never done by hand, and I believe it is safe
to say that the flight problems of supersonic aerodynamics and the
shielding problems for a nuclear aircraft engine could not be solved if
our computers were taken away.
LONG RANGE APPLICATIONS
The items I have just mentioned, although they differ
considerably in difficulty, share one characteristic: they involve
analysis but no synthesis. It is a Common situation to use a computer 'Of
some sort for the analysis of a proposed design; it is much less frequent,
and a much more challenging possibility, to also ask the machine to improve
the design and arrive at a specified optimum. To put it bluntly, I
might say that mechanizing design analysis is a way of getting around the
engineering assistant shortage, while mechanizing design synthesis
mitigates the shortage of engineers.
A requirement for this sort of thing - in fact the most
important requirement - is a complete understanding - of the physical
fundamentals underlying your product. I 'Was personally responsible, many
years ago, for the first steps towards mechanizing the design of optical
lenses; Pittsfield and Port Wayne engineers have gone far in automatic
design techniques for transformers, and much has been done in the optimization
of power distribution networks. These problems could be tackled earlier
because 'We understand things like lenses and transformers very well; on
the other hand, the problems of aerodynamics, combustion, and mechanical
design posed by jet engines and rockets are much less clearly understood,
and we are years from synthesis of such engines.
The message here, of course, is that those of you who are
engineering technical products that are clearly understood should immediately
consider the use of computers for automatic optimization of designs. There
will be additional payoff to the ones already mentioned - lower costs, shorter
time -in that you can be certain of a more perfectly optimized design. And
this is important when your product competes on the basis of a few tenths
of a per cent of efficiency, as does a steam turbine or a transformer.
A related application is that of systems design. Until the
advent of large computers, it 'Was necessary to design the component of a
system individually, and to then assemble the components by a separate
design process. This will, of course, always be the case in many areas,
for the good engineer can dream up a system so huge as to swamp the
computing equipment available at any instant including human brain
power! But the big digital or analog machine does make it possible
to consider much larger chunks of a system; for instance, we are already
able to analyze the propulsion system, the structure, the guidance system,
and the armament capabilities of a missile in the most powerful computers
now available. Since computers are improving at a faster rate than 'the
complexity of missiles is increasing, we shall shortly be able to analyze
a complete missile - a complete weapons setup - in one computer
setup. Then the interaction of jet engine compressor design and the
alternative arming devices for a nuclear warhead can be examined
explicitly, instead of intuitively!
Let's get down to earth again for a moment. You aren't all
concerned with ICBM! But every engineering manager in this audience, and
every manufacturing manager is concerned, with the way engineering
information sets turned into metal. So I want to mention in passing the
great potential that we must explore in replacing the engineering
blueprint and manual operation of tools in the shop. As certainly as day
follows night, engineering design calculations will some day result, not
in a draftsman's pretty picture, but in a reel of magnetic tape (or some
more economical or Bore durable alternative) which will directly control a
machine tool making the experimental part, the tools and dies for mass
production, or the master for a printed circuit.
CONCLUSION
I have talked philosophy today because the interests of this audience
are so diverse. Finance managers don't want to hear about the details of
steam turbine stress calculations, just as the engineering managers don't
particularly enjoy the fine details of mechanized inventory control. I am
conversant, however, with many of the detailed engineering and scientific contributions
made by use of electronic computers in the Company, and would be glad to answer
questions now or later. I should also like to invite those of you who come
through Evendale to inspect the new Computation Building and its
equipment.
|
