Ensemble　Empirical　Mode　Decomposition　(EEMD)　is　a　new　noise-assisted　data　analysis　(NADA)　method,　which　utilizes　the　statistical　characteristics　of　white　noise　to　alleviate　mode　mixing　in　the　original　Empirical　Mode　Decomposition　(EMD)　method.　Finite,　not　infinitesimal,　amplitude　white　noise　is　necessary　to　force　the　ensemble　to　exhaust　all　possible　solutions　in　the　sifting　process,　thus　making　the　different　scale　signals　to　collate　in　the　proper　Intrinsic　Mode　Functions　(IMFs)　of　EEMD.　The　effect　of　EEMD　depends　on　two　key　parameters　which　are　the　amplitude　of　white　noise　and　the　ensemble　times.　However,　the　shortcoming　of　EEMD　is　that　it　lacks　adaptability　and　reliability　because　the　these　two　key　important　parameters　are　obtained　by　experience　and　human　intervention.　A　novel　approach　called　Adaptive　Ensemble　Empirical　Mode　Decomposition　(AEEMD)　is　proposed　in　this　paper,　by　adding　white　noise　and　ascertaining　ensemble　number　adaptively.　The　criterion　of　adding　white　noise　in　AEEMD　is　established,　by　which　a　composite　simulation　signal　could　be　adaptively　and　accurately　decomposed　into　IMFs　without　mode　mixing.　The　proposed　method　is　applied　to　an　early　rub-impact　fault　diagnosis　of　heavy　oil　catalytic　cracking　machine　set　and　a　gear　fault　detection　of　hot　strip　finishing　mills.　The　result　shows　that　AEEMD　can　obtain　more　precise　diagnosis　results　than　original　EMD　and　FFT　spectrum.