Cavitations　which　is　one　of　the　worst　faults　forseawater　hydraulic　pump　should　be　diagnosed　and　prognosed.　In　this　paper,　a　improved　Hilbert-Huang　Transform　(　HHT)　is　proposed　to　extract　faults　feature　of　seawater　hydraulic　pump＇s　vibration　signal;　vibration　signal　is　decomposed　into　several　Intrinsic　Mode　Function　(IMF)　based　on　Ensemble　Empirical　Mode　Decomposition　(　EEMD　).Then　IMFincluding　the　features　of　failure　are　extracted　into　fault　feature　vectors　by　HHT.　Prognosis　model　is　established　based　on　support　vector　machine　(SVM)　method.　Fault　feature　vectors　are　acted　as　inputs　of　the　model　to　predict　cavitations　of　seawater　hydraulic　pump.　It　is　shown　that　the　method　of　this　paper　is　effective.　©　2019　IEEE.