some digging into history and an analysis of what we find. For example,
while we generally consider that the loom was simply mechanized during
the dawn of industrial revolution, the seeds of computer control were
sown by Jacquard with punched-card programming of the needles in his
loom. Neither is it sufficient to say that the present spectacle of
automated pushbutton machines producing many commodities is no different
from the introduction of mass-produced tractors. Tractors, after all,
displaced horses; the computer-controlled factory is displacing men who
don’t always want to be put out to pasture.
Automation is radically changing our lives. It is to be hoped that
intelligent and humane planning will facilitate an orderly adjustment to
this change. Certainly workers now toil in safer and pleasanter
surroundings. It is reported that smashed toes and feet, hernia, eye
trouble, and similar occupational accidents have all but disappeared in
automated automobile plants. Unfortunately other occupational hazards
are reportedly taking the place of these, and the psychological trauma
induced by removal of direct contact with his craft has given more than
one worker stomach ulcers. Let us investigate this transfer of contact
from man to computer-controlled machine.
A paper presented at the First Congress of the International Federation
of Automatic Control, held in Moscow in 1960, uses as its introductory
sentence, “Automatic control always involves computing.” The writer then
points out that historically the computing device was analog in nature
and tied so closely with the measuring and control elements as to be
indistinguishable as an actual computer. In more recent history,
however, the trend has been to separate the computer. With this trend is
another important change, that of using the digital computer in
automatic control.
One of the first papers to describe this separate computer function is
“Instrument Engineering, Its Growth and Its Promise,” by Brown,
Campbell, and Marcy, published in 1949. “Naturally,” the authors state,
“a computer will be used to control the process.” Not a shop foreman or
an engineer, but a computer. Watt’s “flyball” governor pioneered the
field; more recent and more obvious examples of control by computers
include ships guided by “Iron Mike” and airplanes flown by the automatic
pilot. These were analog devices, and the first use of a digital
computer as a control was in 1952, quite recently in our history. This
airborne digital control computer was built by Hughes and was called
“Digitac.”
Since most industries have been in existence for many years, far
antedating aviation, electronics, and the modern computer, the general
incorporation of such control has been difficult both because of the
physical problem of altering existing machines and the mental phenomenon
of inertia. Factory management understandably is slow to adopt a
revolutionary technique, and most control systems now in use in industry
are still analog in nature. However, where new plants are built from the
ground up for computer control, the results are impressive. Designed by
United Engineering, the Great Lakes 80-inch hot strip mill automatically
processes 25-ton slabs of steel. More than 1,000 variables are
controlled, and 200 analog signals and 100 digital computer-generated
signals are used in the process. The steel sheets are shot out of the
rolls at some 45 miles an hour, or about 66 feet a second! A human
supervisor would have a difficult job just watching the several hundred
signals related to thickness, temperature, quality, and so on, much less
trying to think what to do if he noticed something out of
specifications. This would be roughly analogous to an editor trying to
proofread a newspaper as it flashes by on the press and making
corrections back in the linotype room before any typographical errors
were printed. The new computer-controlled mill has an output of about
450,000 tons a month, twice that of the next largest in operation.
American control experts who attended the Moscow conference brought back
the information that Russian effort in computer control is greater than
that in the United States, and that the Russians are more aware of what
we are doing in the field than we are of their progress. Their
implementation of modern computer control may be made easier because
their industries are newer and do not represent such a long-established
and expensive investment in hard-to-modify existing equipment.
Basically, at least, computer control is simple and can be compared to
the feedback principle that describes many physical systems including
the workings of our own bodies. In practice, the computer can be put in
charge of producing something, and by sampling the output of its work
can constantly make corrections or improvements that are desired. This
is of course an extreme simplification, and the control engineer speaks
of “on-line” operation, of adaptive systems that adjust to a changing
environment, of predictive control, and so on. One vital requirement of
the computer involved in a control process, obviously, is that it cannot
take its time about its computations. The control computer is definitely
operating “on the line”; that is, in real time, or perhaps even looking
ahead by a certain amount so that it can not only keep up with
production but also predict forthcoming changes and make corrections in
time to be of use.
The human process controller is stuck with methods like those of the
cook who mixes up his recipe with a spoonful of this, and three pinches
of that, sniffs or tastes the batter subjectively, and may end up with a
masterpiece or a flop. Computer control processes the same batter
through the pipes at a thousand gallons a minute and catches
infinitesimal variations in time to correct them before the hotcakes are
baked. In effect it makes hindsight into foresight by compressing time
far more than man could hope to do.
Early applications of the computer in industrial processes were simply
those of data “loggers,” or monitors. It was still up to the human
operator to interpret what the computer observed and recorded, and to
throw the switch, close the valve, or push the panic button as the case
demanded. Actual computer control, the “closing of the loop” as the
engineers call it, is the logical next step. This replaces the human
operator, or at least relegates him to the role of monitor.
The Great Lakes hot-rolling steel mill has been mentioned as an example
of complete computer control. In Hayange, France, the first European
completely automated steel-beam mill is slated to go into operation late
in 1962. The Jones & Laughlin Steel Corporation in this country uses a
digital computer system to control continuous annealing in its
Aliquippa, Pennsylvania, plant, and is evaluating an RCA
computer-controlled tin-plating line operating at 3,000 feet a minute.
Newer computer-control applications in the offing include sintering and
other metal production operations.
[Illustration:
_Minneapolis-Honeywell_
Boston ice cream makers, H. P. Hood & Sons, use computer to make
pushbutton ice cream. Analog computer thinks out recipes, punches
them on cards to operate valves.
]
To those of us who consume it, ice cream may not seem a likely candidate
for computer control. However, the firm of H. P. Hood & Sons uses
computer control in its blending operation, finding it 20,000 times as
fast, and more accurate than when handled by human operators, since
computer controls hold mixes within one-tenth of 1 per cent accuracy.
Automation is a significant breakthrough in this industry, whose history
goes back 110 years, and in baking, which is a little older. The Sara
Lee bakeries use the computer too in assembling the ingredients for
their goodies. To bake such cakes, Mother will have to get herself a
computer.
Minneapolis-Honeywell furnished the computer for the ice-cream control;
this same company delivered a system for the Celanese Corporation of
America’s multimillion dollar acetyl manufacturing plant at Bay City,
Texas. The new plant produces a petrochemical used in plastics, paint,
synthetic rubber, dye, fibers, and other products. Going “on-stream” in
1962, the Celanese plant will produce half a billion pounds of chemicals
annually.
Russia has been mentioned as active in industrial computer control. A
case in point is the soda plant at Slavyansk in the Donets Basin, which
was recently test-operated for a continuous period of 48 hours by
computer. An unusual feature of this test was that the computer was in
Kiev, almost 400 miles away. A wire link between the two cities
permitted monitoring and control of the plant from Kiev in what the
Russians claim as the first remote automatic operation of such a plant.
Other Soviet achievements include two large-scale automatically
controlled installations. In oil-field operation at Tataria, gas and oil
outputs from many wells are monitored and controlled from a central
station, dropping the work force required from 600 to 100. The other
installation controls irrigation servicing 9,000 acres. A desktop
control handles the pumping of water from the Syr Darya River through
underground pipes, and distribution to Uzbekistan cotton fields. The
Russians have also designed an automatic distillation unit for the
Hungarians. With an annual capacity of a million tons, the unit was
installed in the large Szoeny refinery and scheduled for operation by
1962.
Refineries in the United States are also employing automatic controls in
their operations. Phillips Petroleum installed a digital computer
control system in its Sweeney, Texas, plant to achieve maximum
efficiency in its thermal cracking process. In the first step of an
experimental program, Phillips, working with Autonetics computer
engineers, used a digital computer to plan optimum furnace operation. An
initial 10 per cent improvement was achieved in this way, and a further
6 per cent gain resulted when a digital computer was installed on-line
to operate the cracking furnace.
The Standard Oil Company of California is using an IBM 7090 in San
Francisco to control its catalytic or “cat” cracking plant in El
Segundo, some 450 miles away. The need for computer speed and accuracy
is shown by the conditions under which the cracking plant must operate
continuously with no shutdowns except for repair. Each day, two million
gallons of petroleum is mixed in the cracker with the catalyst, a
metallic clay. The mixing takes place at incandescent heat of 1,000° F.,
and the resulting inferno faces operators with more than a hundred
changing factors to keep track of, a job feasible only with computer
help.
Another use of computer control in the petroleum industry is that of
automatic gasoline blending, as done by the Gulf Oil Corporation. A
completely electronic system is in operation at Santa Fe Springs,
California. The system automatically delivers the prescribed quantities
of gasoline for the desired blend. In case of error or malfunction of
equipment, the control alerts the human supervisor with warning lights
and an audible alarm. If he does not take proper action the control
system automatically shuts itself off.
From the time the war-inspired industry of synthetic rubber production
began in 1940 until very recently, it has been almost entirely a manual
operation. Then in 1961 Goodyear Tire & Rubber introduced computer
control into the process at its Plioflex plant in Houston, Texas.
Goodyear expects the new system to increase its “throughput” and also to
improve the quality of the product through tighter, smoother control of
the complicated operation. Other chemical processors using computer
control in their plants include Dow Chemical, DuPont, Monsanto, Union
Carbide, Sun Oil, and The Texas Company.
Adept at controlling the flow of material through pipes, the computer
can also control the flow of electricity through wires. An example of
this application is the use of digital computers in electric-utility
load-control stations. A typical installation is that of the
Philadelphia Electric Company in Philadelphia, the first to be
installed. Serving 3-1/2 million customers, the utility relies on a
Minneapolis-Honeywell computer to control automatically and continuously
the big turbine generators that supply electric power for the large
industrial area. The memory of the computer stores data about the
generators, transmission-line losses, operating costs, and so on.
Besides controlling the production of power for most economy, the
computer in its spare time performs billing operations for exchange of
power carried on with Pennsylvania-New Jersey-Maryland Interconnection
and Delaware Power & Light Company and Atlantic City Electric Company.
Other utilities using computer control are the Riverside Power Station
of the Gulf States Utilities Company, Southern California Edison, and
the Louisiana Power & Light Company’s Little Gypsy station in New
Orleans.
Another industry that makes use of a continuous flow of material is now
being fitted for computer control, and as a result papermakers may soon
have a better product to sell. IBM has delivered a 1710 computer to
Potlach Forests, Inc., in Idaho for control of a paperboard machine 500
feet long. Papermaking up to now has been more art than science because
of the difficulty of controlling recipes. With the computer, Potlach
expects to make better paper, have less reject material, and spend less
time in changing from one product run to another.
Showing that automatic control can work just about anywhere, the English
firm of Cliffe Hill Granite Company in Markfield, Leicestershire,
controls its grading and batching of granite aggregate from a central
location. Besides rock-crushers, cement plants like Riverside Cement
Company use computer control in the United States.
Thus far most of the computer control operations we have discussed are
in the continuous-processing fields of chemicals or other uniform
materials. The computer is making headway in the machine shop too,
although its work is less likely of notice there since the control panel
is less impressive than the large machine tool it is directing. Aptly
called APT, for Automatically Programmed Tools, the new technique is the
brainchild of M.I.T. engineer Douglas Ross. Automatic control eliminates
the need for drill jigs and other special setup tools and results in
cheaper, faster, and more accurate machine work.
[Illustration:
_International Business Machines Corp._
Controlled by instructions generated by IBM’s AUTOPROMPT, a Pratt &
Whitney Numeric-Keller continuous-path milling machine shapes a raw
aluminum block (upper left) into the saddle-shaped piece shown at
right. The surface is a portion of a geometric shape called a
hyperbolic paraboloid.
]
A coded tape, generated by a computer, controls the milling machine,
drill press, or shaper more accurately than the human machinist could.
In effect, the computer studies a blueprint and punches out instructions
on tape that tell the machine what it is to do, how much of it, and for
how long. Huge shaping and contouring machines munch chunks of metal
from blanks to form them into complex three-dimensional shapes.
Remington Rand UNIVAC and IBM are among the companies producing
computers for this purpose. The trend is to simpler, more flexible
control so that even small shops can avail themselves of the new
technique. In a typical example of the savings possible with “numerical”
tape control, these were the comparative costs:
[Illustration:
_Control Engineering_
Operation of computer-controlled freight yard in England.
]
_Conventional_ _Tape
Control_
Tooling $755 $45
Setup time 15 min. 15 min.
Work time 15 min. 11 min.
Cost per $2.96 $1.81
part
Beyond the automated single- or multipurpose tool is the completely
computer-controlled assembly line. Complete automation of products like
automobiles may be some distance off, but there is nothing basically
unworkable about the idea. Simpler things will be made first, and to
promote thinking along these lines, Westinghouse set up an automatic
assembly line for paperweights. An operator typed the initials of
manufacturing department managers on a computer, which transferred the
instructions to a milling machine. The machine cut the initials in
aluminum blocks which were then automatically finished, painted, and
packaged for shipment as completed paperweights.
Another firm, Daystrom, Inc., is designing a computer control system for
assembly lines which will adjust itself for the “best” product as an
output. President Tom Jones described the principle in which the
computer will begin production, then move valves, switches, or other
controls a small amount. Measuring the finished product, it will decide
if the change is in the right direction, and proceed accordingly. Once
it finds the optimum point, it will lock in this position and settle
down to business.
An excellent example of the computerized assembly line is the Western
Electric Company carbon resistor production line at its Winston-Salem
plant. A digital computer with a 4,096-word memory is used for the
programming, setup, and feedback control of the eleven-station line. It
can accept a month’s scheduling requirements for deposited carbon
high-quality resistors in four power ratings and almost any desired
resistance values. Production rate is 1,200 units per hour.
The computer keeps track of the resistors as they are fabricated,
rejecting those out of specification and adjusting the process controls
as necessary. Operations include heating, deposition of carbon, contact
sputtering, welding, grooving, and inspecting.
_The Robots_
Most of these automated factory operations are doing men’s work, but it
is only when we see the robot in the shape of ourselves that cold chills
invade our spines. Children’s Christmas toys lately have included
mechanical men who stride or roll across the floor and speak, act, and
even “think” in more or less humanoid fashion, some of them hurling
weapons in a rather frightening manner. There is an industrial robot in
operation today which may recall the dread of Frankenstein, though its
most worried watchers are perhaps union officials. Called Unimate, this
factory worker has a single arm equipped with wrist and hand. It can
move horizontally through 220 degrees, and vertically for 60 degrees,
and extend its arm from 3 feet to 7 feet at the rate of 2-1/2 feet a
second. Without a stepladder, it can reach from the floor to a point
nearly 9 feet above it. Unimate can pick up 75 pounds, and its 4-inch
fingers can clamp together on an iron bar or a tool with a force of up
to 300 pounds.
The robot weighs close to a ton and a half, but can be moved from job to
job on a fork-lift truck. Its designers have turned up a hundred
different jobs that Unimate could do, including material loading,
packaging, welding, spray painting, assembly work, and so on. The robot
has a memory and can retain the 16,000 “bits” of information necessary
for 200 operations. To teach it a new task, it is only necessary to
“help” it manually through each step one time. Unimate can be instructed
to wait for an external signal during its task, such as the opening of a
press or a furnace door.
Advantages of a robot are many and obvious. Pretty girls passing by will
not distract it, nor will it require time for lunch or coffee breaks, or
trips to the washroom. If necessary it will work around the clock
without asking for double power for overtime. High temperatures, noxious
gases, flying sparks, or dangerous liquids will not be a severe hazard,
and Unimate never gets tired or forgets what it is doing.
But Unimate has some drawbacks that are just as obvious. It can’t tell
one color from another, and thus might paint parts the wrong color and
never know the difference. It is not readily movable, and not very
flexible either. It costs $25,000, and will need about $1,300 in
maintenance a year. Some industry spokesmen say that this is far too
much, and Unimate has a long way to go before it puts any people out of
work. Others say it is a step in the right direction, and this is
probably a fair evaluation.
Apparently United States Industries, Inc., whose AutoTutor teaching
machines are pacing the field, has made another step in the right
direction with its “TransfeRobot 200.” This mechanical assembly-line
worker is an “off-the-shelf” item, and currently in use by about fifty
manufacturers. TransfeRobot uses its own electronic brain, coupled with
a variety of magnetic, mechanical, or even pneumatic fingers to pick up,
position, insert, remove, and do other necessary operations on small
parts.
Besides these capabilities, TransfeRobot controls secondary operations
such as drilling, embossing, stamping, welding, and sealing. It is now
busy building things like clocks, typewriters, automobile steering
assemblies, and electrical parts. No one-job worker, it can be
re-programmed for other operations when a new product is needed, or
quickly switched to another assembly line if necessary. Billed as a new
hand for industry, TransfeRobot obviously has its foot in the door
already. United States Industries estimates current yearly sales of its
small automation equipment at about $3 million.
[Illustration:
_Massachusetts Institute of Technology_
Dr. Heinrich Ernst, Swiss graduate student at MIT, watches his
computer-controlled “hand” pick up a block and drop it in the box.
]
The robots in Čapek’s play _R.U.R._ looked like their human makers, but
scientist Claude Shannon is more realistic. “These robots will probably
be something squarish and on wheels, so they can move around and not
hurt anybody and not get hurt themselves. They won’t look like the
tin-can mechanical men in comic strips. But you’ll want them about
man-size, so their hands will come out at table-top or assembly-line
level.” Since Professor Shannon is the man who sparked the
implementation of symbolic logic in computers, his ideas are not
crackpot, and the Massachusetts Institute of Technology’s Hand project
is a good start toward a real robot. Dr. Heinrich Ernst, a young Swiss,
developed Hand with help from Shannon. Controlled by a digital computer,
the hand moves about and exercises judgment as it encounters objects.
Such research will make true robots of the remotely manipulated machines
we have become familiar with in nuclear power experiments, underwater
exploration, and so forth. Hughes Aircraft’s “Mobot” is a good example,
and it is obvious that the robot’s bones, muscles, and nerves are
available. All they need is the brain to match.
While we wait fearfully for more robots which look the way we think
robots should, the machine quietly takes over controlling more and more
even bigger projects. The computer does a variety of tasks, from the
simple one of cutting rolling-mill stock into optimum lengths to
minimize waste, to that of running an electronic freight yard in which
cars are classified and made up automatically. The computer in this
application not only measures the car and weighs it, but also computes
its rollability. Using radar as its eyes, the computer gauges the speed
and distance between cars as they are being made up and regulates their
speed to prevent damaging bumps. To the chagrin of veteran human
switchmen, the computer system has proved it can “hump” cars—send them
coasting to a standing car for coupling—without the occasional
resounding crash caused by excessive speed.