To　extract　effectively　features　and　improve　damage　identification　precision,　this　study　proposed　an　intelligent　data-fusion　model　by　integrating　fractal　theory,　probabilistic　neural　network　(PNN)　and　data　fusion.　A　two-span　concrete-filled　steel　tubular　arch　bridge　in　service　was　used　to　validate　the　intelligent　model　by　identifying　both　single-　and　multi-damage　patterns.　The　results　show　that　the　intelligent　model　proposed　can　not　only　reliably　identify　damage　with　different　noise　levels,　but　also　have　an　excellent　anti-noise　capability　and　robustness.