In　order　to　solve　the　disadvantages　of　＇under　maintenance＇　or　＇over　maintenance＇　in　traditional　maintenance　methods　of　CNC　machine　tools,　a　preventive　maintenance　method　was　proposed　based　on　the　prediction　of　system　performance　degradation.　The　SPC　technology　was　used　to　analyze　the　product　quality　data　and　realize　the　prediction　of　the　working　states　for　the　machine　tools.　The　Wiener　degradation　model　of　a　key　component　was　established　to　predict　the　life　distribution,　so　that　the　optimal　preventive　maintenance　strategy　might　be　determined　by　residual　life　and　revenues.　Taking　the　abnormal　data　of　CNC　machine　tools　in　an　automobile　piston　production　line　as　an　example,　the　validity　of　proposed　method　was　verified.　©　2019,　China　Mechanical　Engineering　Magazine　Office.　All　right　reserved.