According　to　the　bridge　damage　information,　identifications　are　random　and　the　Bayesian　network　analysis　expresses　information　potential　probability　relationship　as　flexible.　Based　on　Bayesian　network　and　seismic　theory,　the　method　of　bridge　seismic　hazard　grade　identification　is　established.　Based　on　the　complete　investigation　data　of　the　Wenchuan　earthquake　bridges,　eight　representative　factors　such　as　PGA　etc.　are　selected.　The　influence　of　bridge　seismic　vulnerability　classification　variables　and　discrete　quantization　value,　parameters　learning　using　the　maximum　likelihood　(MLE)　method　is　to　get　the　nodes　probability　table.　The　grade　of　bridge　seismic　is　identified　by　using　BN　reverse　reasoning　technology　and　it　has　good　prediction　accuracy　of　95.5%　and　can　identify　the　reasons　for　damaging.　Through　the　model　validation　of　the　Baihua　Bridge　and　10　typical　bridges,　the　classification　imbalance　of　the　sample　data　had　impacts　on　the　model　prediction.　There　is　a　deviation　between　the　damage　states　of　the　bridge　with　the　survey　data,　the　results　verify　the　feasibility　and　effectiveness　of　the　proposed　method.　The　traffic　capacity　of　the　bridge　after　the　earthquake　the　traffic　flow　changes　and　the　transit　time　in　the　Lishui　Bay　Bridge　are　evaluated.　The　results　show　that　the　evaluation　results　provide　some　reference　values　for　post-earthquake　disaster　reduction.