The　rolling　bearings　often　suffer　from　compound　faults　in　practice,　moreover,　several　faults　add　together　and　interfere　with　each　other,　which　make　it　difficult　to　separate　week　fault　signals　from　them　through　conventional　ways.　In　order　to　improve　the　compound　faults　diagnosis　of　rolling　bearings　via　signals＇　separation,　the　paper　proposes　a　new　method　to　identify　compound　faults　from　mixed-signals,　which　is　based　on　parallel　dual　Q-factors　method　and　the　adaptive　maximum　correlation　kurtosis　deconvolution　(AMCKD)　method.　With　the　approach,　the　vibration　signal　is　firstly　decomposed　into　high　and　low　resonance　components　by　the　sparse　decomposition　based　on　dual　Q-factors　method,　which　can　sparsely　represent　the　signal　and　extract　the　fault　impact　signal.　Then,　the　low-resonance　component　is　processed　by　AMCKD,　and　the　AMCKD　can　adaptively　optimize　the　selection　of　parameters　M　and　L.　Finally,　the　compound　faults　can　be　separated　effectively　by　the　method,　which　makes　the　fault　features　more　easily　extracted　and　more　clearly　identified.　Simulated　analysis　result　validates　the　effectiveness　of　the　proposed　method　in　compound　faults　separating.　©　2018　IEEE.