The　optimizing　setting　values　of　the　key　controlled　variables　are　difficult　to　find　in　the　shaft　furnace　roasting　process　for　the　hematite　ores,　which　results　in　difficulty　to　control　production　indices　within　the　target　ranges.　A　multivariable　intelligent　optimizing　setting　method　was　developed　by　the　combination　of　case-based　reasoning　(CBR),　variables　prediction　and　expert　system　(ES).　First　a　presetting　model　presented　the　presetting　points　for　the　control　loops.　Next,　the　presetting　points　were　evaluated　and　adjusted　respectively　by　a　case　evaluating　model　and　a　case　revising　model.　Thereby　the　set　points　of　control　loops　of　process　were　automatically　adjusted　online.　The　industrial　application　has　proved　that　it　can　adapt　the　frequently　changed　working　conditions　while　fulfilling　the　optimal　control　for　the　production　indices.　Prominent　effects　have　been　observed　by　the　proposed　method.