Fan　plays　an　irreplaceable　role　in　the　ventilation　of　underground　environment.　It　will　cause　serious　consequences　once　the　fan　fails.　So　it　is　essential　to　do　fault　diagnosis　of　fan　to　prevent　the　emergence　of　break　down.　Spectral　analysis　method　is　a　simple　and　effective　method　for　signal　analysis　and　it　has　been　widely　used　in　fault　diagnosis　of　rotating　machinery.　This　paper　has　researched　the　bearing　fault　of　fan　on　the　base　of　in-depth　study　of　the　frequency　characteristics　of　typical　faults　fan.　High-frequency　noise　of　vibration　signal　has　been　removed　through　wavelet　soft　thresholding　method.　Get　its　spectrum　characteristic　of　the　reconstructed　signal,　and　then　the　failure　category　of　signal　is　discriminated　by　comparing　its　characteristic　frequency　to　typical　faults　frequency　in　theory.　Combining　the　advantage　of　LabVIEW　and　MATLAB,　we　implement　the　algorithm　through　two　of　them,　accomplishing　denoising　and　spectral　analysis　of　vibration　signals,　completing　fault　diagnosis　of　fan.