In　order　to　make　full　use　of　the　redundant　and　complementary　information　and　to　assess　the　structural　health　states　from　a　structural　health　monitoring　system,　a　new　damage　identification　method　is　proposed　by　integrating　with　rough　set,　data　fusion　and　probabilistic　neural　network　(PNN).　In　this　method,　rough　set　is　used　to　reduce　attributes　so　as　to　decrease　spatial　dimensions　of　data　firstly,　then　PNN　is　utilized　to　fuse　redundant　and　uncertain　information　and　fusion　decision-making　and　damage　identification　results　are　made.　It　is　noteworthy　that　K-means　clustering　was　employed　to　discrete　data　during　the　attributes　reduction.　To　validate　the　efficiency　of　the　proposed　method,　multi-damage　patterns　from　two　numerical　examples　were　identified　finally,　and　a　comparison　was　made　between　the　proposed　method　and　a　PNN　classifier　without　data　processing　by　rough　set.　The　results　show　that　the　proposed　method　can　not　only　reduce　spatial　dimension　of　data,　but　also　have　good　damage　identification　accuracy.