PADIN@FNALCDF.BITNET (10/17/86)
My interest in AI was peaked by a blurb on EXPERT SYSTEMS which I
read in the DEC PROFESSIONAL. I immediately saw the possible use of EXPERT
SYSTEMS in my work here at FERMILAB. However, in reading more about the
development of an ES, it appears to be a very long process and useful only
under certain circumstances as outlined by Waterman in his book 'A Guide to
Expert Systems'. He states
"Consider expert systems only if expert system development
is possible, justified, and appropriate."
By 'possible' he means
if [ (task does not require common sense) &
(task requires only cognitive skills) &
(experts can articulate their methods) &
(genuine experts exist) &
(experts agree on solutions) &
(task is not too difficult) &
(task is not poorly understood) ]
then
[ expert system development is POSSIBLE ]
By 'justified' he means
if [ (task solution has a high payoff) or
(human expertise being lost) or
(human expertise scarce) or
(expertise meeded in many locations) or
(expertise needed in hostile environment) ]
then
[ expert system development is JUSTIFIED ]
By 'appropriate' he means
if [ (task requires symbol manipulation) &
(task requires heuristic solutions) &
(task is not too easy) &
(task has proctical value) &
(task is of manageable size) ]
then
[ expert system approach is APPROPRIATE ]
As OPERATORS at FERMILAB we take the Protons that are extracted from
our MAIN RING and maneuver them to experimental targets. There are several
areas which I see the possible application of ES in our work.
1) troubleshooting help -- we are responsible for maintaining
a multitude of systems: water,cryogenic,computer,CAMAC,
electrical,safety interlock, and more. quick solutions
to problems save money, time, and maximize data flux to
experiments.
2) operator training -- we have both rapid turnover and a long
training season, i.e., it takes at least a year for an
operator to be trained. thus, we need a large knowledge
base and a sophisticated simulator/tutorial.
3) data aquisition -- we monitor large amounts of status data
and must have out-of-bounds alarms for many devices.
our alarm displays need to be centralized and smart so
that they display actual problems.
4) control system -- we control the path which the Protons
take by controlling the magnetic field strengths of
magnets though which the Protons travel. 'TUNING' a
BEAM LINE (targeting protons onto experimental
apparatus) is an art and as such is subject to the
frailty of human judgement. proper tuning is mandatory
because it increases data flux to experiments,
minimizes radiation intensities, and reduces equipment
damage.
? Are Waterman's criteria reasonable ones on which to make a decision
about pursueing ES application?
? I've read that the creation of an ES would take about 5 man-years,
does that sound right?
? If an ES is recomended, what would be the next step? Do I simply
call a representative of some AI company and invite them to make
a more informed assessment?
First I must convince myself that an ES is something that is really necessary
and useful. Next I must be able to convince my superiors. And finally, DOE
would need to be convinced!
thanks for any info
Clem <Padin@fnal.bitnet>