Here,　an　adaptive　redundant　lifting　wavelet　denoising　method　was　proposed　to　realize　effective　identification　of　weak　fault　information　of　roller　bearings　in　mechanical　equipments.　According　to　the　different　features　contained　in　scale　coefficients　to　be　decomposed,　wavelets　matching　optimally　these　features　were　adaptively　selected　based　on　the　norm　criteria.　Meanwhile,　a　multi-hole　algorithm　was　introduced　here　to　guarantee　sufficient　information　in　each　scale　coefficient　and　wavelet　coefficient　after　every　decomposition.　Then,　the　variable　scale　threshold　denoising　algorithm　was　applied　to　process　each　wavelet　coefficient,　the　reconstruction　and　envelope　spectral　analysis　of　these　coefficients　were　performed　to　extract　fault　features　of　roller　bearings.　Analysis　results　with　the　method　presented　above　for　compound　fault　signals　of　a　bearing　test　table　and　field　practical　signals　showed　that　bearing　fault　identification　can　be　well　realized　and　the　proposed　method　is　effective.