Vibration　source　information　(source　number,　source　waveforms,　and　source　contributions)　of　gears,　bearings,　motors,　and　shafts　is　very　important　for　machinery　condition　monitoring,　fault　diagnosis,　and　especially　vibration　monitoring　and　control.　However,　it　has　been　a　challenging　to　effectively　extract　the　source　information　from　the　measured　mixed　vibration　signals　without　a　priori　knowledge　of　the　mixing　mode　and　sources.　In　this　paper,　we　propose　source　number　estimation,　source　separation,　and　source　contribution　evaluation　methods　based　on　an　enhanced　independent　component　analysis　(EICA).　The　effects　of　nonlinear　mixing　mode　and　different　source　number　on　source　separation　are　studied　with　typical　vibration　signals,　and　the　effectiveness　of　the　proposed　methods　is　validated　by　numerical　case　studies　and　experimental　studies　on　a　thin　shell　test　bed.　The　conclusions　show　that　the　proposed　methods　have　a　high　accuracy　for　thin　shell　structures.　This　research　benefits　for　application　of　independent　component　analysis　(ICA)　to　solve　the　vibration　monitoring　and　control　problems　for　thin　shell　structures　and　provides　important　references　for　machinery　condition　monitoring　and　fault　diagnosis.