FUZZY THINKING

A CHEMTEC extraction

BY: Wilson Smith

FROM: SPARTANBURG COMMODORE USERS GROUP


NNNN Computers give blunt answers.Yes or no,black or white. Researchers in
artificial intelligence are trying to teach their machines a little subtlety:
to encode the shades of gray characteristic of human thought. One approach
that is producing good results is FUZZY LOGIC, the creation of Lotfi Zadeh of
the University of California at Berkeley. His brainchild is now running cement
kilns and making decisions for corporate managers. When a conventional expert
system is given the task of deciding whether a man is tall, it will reply yes
to anybody above, say, six feet, and no to everybody else. This is doing an
injustice to an English word. Tall has a less precise (and more useful)
meaning in conversation--taller than the local average, tall for your age,
taller than broad, etc.
NNNN In fuzzy logic, if the computer's conception of "tall" is seven feet
high, the truth value of the statement "he is tall" willl be 100% if a man is
seven feet and close to 0% if he is four feet high. In between, the truth
value varies smoolthly between 0 and 100.
NNNN The fact that somebody is tall hardly requires a computer. However,
exactly the same logic applies to deciding whether an economy is healthy,
except in this case a vast array of separate statistics must be taken into
account--inflation, balance of payments, growth, and so on. Fuzzy logic is a
way of considering all these factors simultaneously in a consistent way.
NNNN Conceived over twenty years ago, artificial intellegience it is still
controversial. Skeptics say it is no more than a restatement of probability
theory and has nothing to do with logic. They note that it flies in the face
of the law of the excluded middle--an object either belongs to a set or does
not--which is a cornerstone of conventional logic. Yet there is no reason that
logic should be restricted to two values (yes and no); mathematicians have
been experimenting with more values than that for 50 years. Fuzzylogic has an
infinite number of values, but an intriguing thing about it is that it and all
other systems with more than two values can be derived mathematically from a
three-value system (yes, no, and maybe). In its philosophical status, fuzzy
logic is being put to work. It's ideal application is in expert
systems--computer programs that regurgitate human expertise. One of the most
successful is a microprocessor controller for cement kilns. A cement kiln is a
cylindrical chamber about 300 feet long in which limestone and clay react at
about 1000C to form small nuts of "clinker" minerals that are later ground
down into cement. Only a half-dozen measurements can be made to determine the
state inside the kiln, but experienced operators have learned to fine-tune the
kiln using this limited information.
NNNN Researchers interviewed operators to obtain fuzzy-logic membership curves
for such notions as "lime content high" or "kiln drive torque low." The
microprocessor measures the truth value of complex statements based on these
curves and decides how much the lime content or the drive torque should be
increased or decreased. Although the microprocessor will only be as good as
the original human operator, it can (unlike a human) work tirelessly around
the clock.
NNNN Operators are now running kilns in Europe and the United States.
Experience so far indicates that the machine pays for itself within a year of
installation. One microprocessor is controlling a similar process in a paper
mill. A California-based firm, has begun marketing a decision-making software
package called Reveal that uses fuzzy logic to sort through large data bases
to find, for example, companies with high sales and large profit margins for
possible acquisition. Other applications of fuzzy logic are on the way.
Hitachi in Japan is testing a fuzzy train driver. It adjusts the train's speed
to optimize such things as safety, comfort, and energy consumption.
 The Institute of Industrial Cybernetics in Sofia, Bulgaria, fuzzy computer
vision guides a robotic arc welder along a metal seam. Fuzzy logic determines
how edge-like each point in the field of view is. This works more accurately
than using a fixed threshold on a gray-scale to trigger perception of an edge.
In China, meteorologists use fuzzy logic to encode low annual temperature and
moderate wind velocity to determine the best regions in China for growing
rubber trees. Perhaps fuzzy logic can be used to reconcile old communist
ideals with new capitalist incentives?  
[PRESS RETURN]: