A　metamodel　can　be　represented　by　the　forms　of　mathematical　equations　(response　surface　models)　or　neural　networks　(black-box　models)　in　the　interest　of　correlating　the　inputs　(parameters)　with　the　outputs　(responses)　of　a　physical　system.　In　view　of　little　relevant　research,　this　paper　attempts　to　use　response　surface　models　as　surrogates　for　FE　models　due　to　the　provision　of　explicit　equations　and　easy　implementation　in　the　aspect　of　modeling　updating.　Metamodeling　is　achieved　by　using　the　design　of　experiment　involving　the　2k　factorial　design　and　the　central　composite　design　for　parameter　screening　and　structural　input-response　modeling　respectively.　The　feasibility　of　the　proposed　method　has　been　demonstrated　using　a　numerical　beam　example　which　proves　the　satisfactory　performance　of　employing　such　metamodels　in　model　updating　based　damage　identification.　©　(2011)　Trans　Tech　Publications.