The modernengineeringdesignoptimizationprocessoftenreplaceslaboratoryexperimentswith computers imulations, which leads to expensive black-box optimization problems. Such problem soften containcandidatesolutionswhichcausethesimulationtofail,andthereforetheywillhaveno objectivevalueassignedtothem,ascenariowhichdegradesthesearcheffectiveness.Toaddressthis, this paperproposesanewcomputationalintelligenceoptimizationalgorithmwhichincorporatesa classifierintotheoptimizationsearch.Theclassifierpredictswhichsolutionsareexpectedtocausea simulationfailure,anditspredictionisusedtobiasthesearchtowardssolutionsforwhichthe simulationisexpectedtosucceed.Tofurtherenhancethesearcheffectiveness,theproposedalgorithm continuouslyadaptsduringthesearchthetypeofmodelandclassifierbeingused.Arigorous performanceanalysisusingarepresentativeapplicationofairfoilshapeoptimizationshowsthatthe proposedalgorithmoutperformedexistingapproachesintermsofthefinalresultobtained,and performedasearchwithacompetitivelylownumberoffailedevaluations.Analysisalsohighlightsthe contributionofincorporatingtheclassifierintothesearch,andofthemodelandclassifier selection steps