A Q UA : A Fram e w ork for A utom atedQ ualitative A bstraction

In thisp ap e r,w e de alwith the p roble m ofabstracting be havio r m odelssuch thattheir le velofgranularity is ascoa rse asp ossible,butstill̄ ne e nough to car ry outa give n be havio r p r ediction o r diagnosistask.The fo cus ison dete r m ining task{de p e nde ntdistinctionswithin the dom ainsofvariables| i.e. q ualitative values| thata r e both ne cessa ry andsu±cie nt,give n a m odel com p osedfro m a library,a granularity ofp ossible obse rvations,anda granularity ofdesiredresults. W e p r ese nta for m alization ofthe p roble m ,p r ese ntfundam e ntalresultsre gardingthe e xiste nce andcharacte rization ofsolutionsto task{de p e nde ntq ualitative abstraction,anddevise a m ethodfor autom atically dete rm ining q ualitative value sbasedon a hie ra rchicalre p rese ntation ofthe device m odelthatallowsto e xp loit itssp e cīc structure. A p rincip ledap p lication isto turn r e al{valuedm odels, ascom m only usedin industry,into q ualitative m odelsto m ake the m accessible to m odel{basedre asoning m ethods.The r esulting to olsetthuse nhancesthe ability to use a be havior m odelofan e ngine e r eddevice as a com m on basisto sup p o rtdi®e r e nttasksalong itslife cycle.