An efficientheuristicforreliabilitydesignoptimizationproblems

This paperdevelopsanefficientheuristictosolvetwotypicalcombinatorialoptimizationproblemsfre- quently metwhendesigninghighlyreliablesystems.Thefirstoneistheredundancyallocationproblem (RAP) ofseries-parallelbinary-statesystems.ThedesigngoaloftheRAPistoselecttheoptimalcombi- nation ofelementsandredundancylevelstomaximizesystemreliabilitysubjecttothesystembudget and tothesystemweight.Thesecondproblemistheexpansion-schedulingproblem(ESP)ofmulti-state series-parallel systems.Inthisproblem,thestudyperiodisdividedintoseveralstages.Ateachstage,the demand isrepresentedasapiecewisecumulativeloadcurve.Duringthesystemlifetime,thedemandcan increase andthetotalproductivitymaybecomeinsufficienttoassumethedemand.Toincreasethetotal system productivity,elementsareaddedtotheexistingsystem.TheobjectiveintheESPistominimize the sumofcostsoftheinvestmentsoverthestudyperiodwhilesatisfyingavailabilityconstraintsateach stage. TheheuristicapproachdevelopedtosolvetheRAPandtheESPisbasedonacombinationofspace partitioning, geneticalgorithms(GA)andtabusearch(TS).Afterdividingthesearchspaceintoasetof disjoint subsets,thisapproachusesGAtoselectthesubspaces,andappliesTStoeachselectedsubspace. Numerical resultsforthetestproblemsfrompreviousresearcharereportedandcompared.Theresults show theadvantagesoftheproposedapproachforsolvingbothproblems.