In　dealing　with　the　deficiency　of　traditional　wavelet　analysis　in　extracting　fault　feature　of　bearing,　the　second　generation　of　wavelet　transform　(SGWT)　is　adopted　in　the　fault　diagnosis　of　bearing.　The　prediction　operator　and　update　operator　of　SGWT　are　constructed　based　on　the　interpolating　subdivision　principle.　Furthermore,　the　first-order　smoothing　boundary　constitution　technology　is　used　to　solve　the　mutation　problem　in　the　calculating　boundary.　As　a　result,　the　reconstructed　signal　can　be　smoothed.　The　denoising　technology　based　on　SGWT　can　subtract　the　background　noise　to　obtain　the　fault　feature　successfully,　and　the　effect　is　much　better　than　that　of　the　traditional　wavelet　denoising.　Finally,　the　incipient　fault　features　have　been　extracted　successfully　by　integrating　the　SWGT　with　correlation　analysis　and　demodulation　technology.