A　RBF　neural　network　based　Monte　Carlo　method　is　proposed　to　evaluate　the　reliability　of　existing　large-span　cable-stayed　bridges　for　bridge　assessment　and　management.　A　fast　simulation　RBF　neural　network　model　is　established　for　the　Zhaobaoshan　bridge,　and　the　training　sample　is　obtained　according　to　uniform　design　and　using　the　ANSYS　software　for　considering　geometrical　nonlinearities.　The　reliability　analysis　of　the　Zhaobaoshan　bridge　under　two　types　of　failure　modes　and　three　limit　states　and　the　sensitivity　analysis　of　the　live　load　modes,　the　limit　state　functions　and　the　detection　periods　to　the　reliability　indices　are　carried　out.　The　results　show　that　several　limit　states　can　be　considered　simultaneously　by　using　the　presented　method.　The　accuracy　and　the　efficiency　of　the　RBF-MC　method　is　verified　from　the　simulation.　The　results　of　the　evaluation　are　influenced　by　the　live　load　modes　and　the　limit　state　functions　considered　in　the　analysis.　The　Zhaobaoshan　bridge　is　safety　during　the　detection　period.