ameen@techunix.BITNET (Ameen Abu_Hanna) (04/12/88)
In model based troubleshooting, probing into the diagnosed
system to examine some terminal's output, is _one way_ to
discriminate between suspect components (competing hypothesis).
Clearly, *choosing* a "good" terminal/port for examination is
vital for efficiency. I need suggestions for heuristics to
estimate how "good" is a terminal examination (i.e. how
discriminatory power it might yield in case such a test
succeeds/fails).
The diagnosed system in my case is concerned about the
electrical/digital domain and modeled (structurally) by a
hierarchical representation where a component might be either a
primitive or a module consisting of other (sub)components.
Aspects like number of pins a chip has (more pins of a
suspected component raise probability of it's "failure belief",
hence an affected terminal by such component might be worth
considering), price of "observability" of the expected output at
some terminal, number of possible contributor suspect components
to the terminal, terminal accessibility etc. are some criteria to
be considered. Any suggestions ? (partial/conceptual ones are
welcomed).
Thanks,
Ameen Abu-Hanna,
Domain: ameen@techunix.technion.ac.il UUCP: ...!ucbvax!ameen@techunix.bitnet
BITNET: ameen@techunix ARPANET: ameen%techunix.bitnet@wiscvm.arpa