To　efficiently　identify　structural　damage,　a　two-stage　damage　detection　method　with　genetic　algorithm　is　presented.　The　frequency　multiple　damage　location　assurance　criterion　(MDLAC)　method　and　mode　shape　MDLAC　method　are　applied　to　obtain　original　decisions,　respectively,　and　the　evidence　theory　is　employed　to　integrate　both　frequency　identification　information　and　mode　shape　identification　information,　which　is　more　accurate　than　MDLAC　methods,　thus　the　damage　sites　are　preliminarily　localized　according　to　the　combined　information.　The　improved　genetic　algorithm　is　finally　applied　to　detect　damage　extents.　The　simulation　results　demonstrate　that　the　evidence　theory　approach　containing　both　frequencies　and　mode　shapes　can　more　successfully　identify　multiple　damage　sites　or　determine　the　group　of　potential　damage　sites　than　MDLAC　methods,　and　the　improved　genetic　algorithm　can　perfectly　detect　damage　extents.