Accurate　uncertainty　estimation　for　parameters　and　responses　of　real-world　structures　may　improve　the　reliability　of　analysis　results.　In　this　study,　by　using　the　probabilistic　collocation　method　and　regression　analysis,　stochastic　response　surface　models　are　firstly　constructed　for　expressing　the　complex　implicit　relationship　between　the　uncertain　parameters　and　responses　of　a　structure.　Then　the　statistical　features　of　the　responses　can　be　quickly　estimated　using　the　constructed　models.　Meanwhile,　a　two-stage　updating　strategy　is　employed　which　separately　updates　the　parameter　means　and　standard　deviations　in　the　interest　of　problem　simplification　and　improvement　in　updating　efficiency.　Lastly,　fifty-five　nominally　identical　steel　plates　were　tested　in　the　laboratory.　The　means　and　standard　deviations　of　the　measured　natural　frequencies　under　a　free-free　boundary　condition　have　been　adopted　to　estimate　the　statistical　features　of　the　plate　thicknesses　and　material　properties.　By　this　means,　the　feasibility　and　reliability　of　the　proposed　method　has　been　validated.　©　2016,　Nanjing　Univ.　of　Aeronautics　an　Astronautics.　All　right　reserved.