Gearbox　vibration　signal　of　high-speed　wire　rolling　mills　is　complex　and　has　much　complex　frequency　structure　and　various　disturb.　It　leads　to　the　difficulty　of　diagnosis　work　especially　some　or　other　early　impulse　fault　hidden　in　existence.　The　paper　presents　a　new　comprehensive　analysis　method　which　integrates　the　statistical　analysis,　the　frequency　analysis　and　the　time-frequency　analysis　organically.　First,　time　domain　wave　analysis,　frequency　analysis,　statistical　trend　analysis　and　characteristic　frequency　amplitude　trend　analysis　are　used　to　judge　the　status　of　the　machine　and　fault　parts　primarily,　and　then　precision　diagnosis　of　wavelet　packets　decomposition　and　reconstruction　based　on　the　multi-resolution　analysis　is　used　to　make　certain　the　fault　quality　and　fault　position　quantitatively.　The　diagnosis　analysis　results　of　one　application　example　indicate　the　validity　and　the　practicability　of　the　method　proposed　here.　©　2006　IEEE.