In　order　to　exactly　determine　the　fault　of　the　ventilator　and　realize　the　early　warning　of　it,　the　warning　principle　of　the　ventilator　failure　is　proposed.　The　fault　early　warning　software　based　on　the　information　fusion　technology　is　developed　using　VC++　6.0　combined　with　Matcom　4.5.　Information　fusion　technology　and　BP　neural　networks　and　evidence　theory　are　adopted　in　the　software,　in　which　non-linearity　mapping　between　the　fault　style　and　the　diagnostic　frequency　is　accomplished　by　the　BP　neural　network　and　multichannels　signal　fusing　and　analyzing　is　carried　out　using　the　theory　of　evidence.　Data　acquisition　and　conclusion　return　are　realized　by　using　the　OPC　mode　communication　between　the　King　View　running　system　and　the　developed　software.