The　disadvantages　of　model　update　based　on　artificial　neuron　network　(ANN)　are　improved.　Because　too　much　samples　for　training　ANN　are　demanded,　uniform　design　method　is　used　to　produce　samples　in　order　to　reduce　the　amount　of　samples　in　the　process　of　structural　damage　simulation,　and　the　computational　efficiency　is　enhanced.　Genetic　algorithm　is　used　to　optimize　the　initial　weight　of　back　propagation　network　and　the　operation　velocity　is　enhanced.　A　recurrence　update　method　based　on　substructures　and　ANN　is　presented　especially　for　spatial　structures.　A　space　structure　is　divided　into　multilayer　substructures;　the　updating　is　realized　step　by　step.　Fist,　frequencies　are　selected　as　damage　factors　and　the　effect　is　verified.　Then　the　energy　of　wavelet　frequency　bands　is　proposed　as　a　better　damage　factor.