The　study　of　the　reliability　of　road　network　during　incident　plays　an　important　role　to　construct　effective　traffic　emergency　response　plan.　First　the　conception　of　recovery　reliability　during　traffic　accidents　is　clearly　defined.　The　influence　of　accident　duration.　severity　and　average　saturation　to　the　time　of　traffic　recovery　is　analyzed　through　the　use　of　DYNASMART　simulation.　With　the　data　collected　by　DYNASMART　simulation,　the　model　of　recovery　time　based　on　BP　neural　network　is　established.　At　last,　Monte　Carlo　simulation　method　is　introduced　to　calculate　the　reliability　of　recovery　time,　and　an　example　demonstrates　the　efficiency　of　the　above　algorithm.