In　this　paper,　a　new　eigen-level　data　fusion　model,　whereby　rough　set　data　and　a　probabilistic　neural　network　(PNN)　are　integrated　using　a　data　fusion　technique,　is　proposed　for　structural　damage　detection.　This　model　is　used　for　structural　damage　detection　and　identification,　particularly　for　cases　where　the　measurement　data　has　many　uncertainties.　More　specifically,　structural　modal　parameters　derived　from　vibration　responses　are　first　discretized　by　the　K-means　clustering　technique　and　the　rough　set　technique　is　then　employed　to　deal　with　the　great　volume　of　data　and　to　extract　optimal　feature　parameters.　After　that,　the　processed　data　and　information　are　input　to　the　fusion　centre　of　the　data　fusion　technique　and　fused　with　the　PNN　to　give　a　fusion-based　damage　detection　result.　To　verify　the　proposed　method,　two　numerical　examples　are　presented　to　identify　both　single　and　multi-damage　case　patterns.　The　effects　of　measurement　noise　and　of　non　pre-processed　rough　set　data　on　the　damage　detection　results　are　also　discussed.　The　results　show　that　the　proposed　model　not　only　has　good　damage　detection　capability　and　noise　tolerance,　but　also　significantly　reduces　data　storage　memory　requirements　and　saves　runtime　as　a　consequence　of　the　data　fusion　processing.