The　reliability　of　machining　accuracy　is　of　great　significance　to　performance　evaluation　and　optimization　design　of　the　machine　tools.　Different　geometric　errors　have　various　influences　on　the　machining　accuracy　of　the　machine　tools.　The　main　emphasis　of　this　paper　is　to　propose　a　generalized　method　to　distribute　geometric　accuracy　of　component　for　improving　machining　accuracy　reliability　under　certain　design　requirements.　By　applying　MBS　theory,　a　comprehensive　volumetric　model　explaining　how　individual　errors　in　the　components　of　a　machine　affect　its　volumetric　accuracy　(the　coupling　relationship)　was　established.　In　order　to　reflect　the　ability　to　reach　the　required　machining　accuracy,　the　concept　of　machining　accuracy　reliability　is　proposed　in　this　paper.　Based　on　advanced　first　order　and　second　moment　(AFOSM)　theory,　reliability　and　sensitivity　with　single　failure　modes　were　obtained　and　the　model　of　machining　accuracy　reliability　and　the　model　of　machining　accuracy　sensitivity　with　multiple　failure　modes　were　developed.　By　taking　machining　accuracy　reliability　as　a　measure　of　the　ability　of　machine　tool　and　taking　machining　accuracy　sensitivity　as　a　reference　of　optimizing　the　basic　parameters　of　machine　tools　to　design　a　machine　tool,　an　accuracy　distribution　method　of　machine　tools　for　improving　machining　accuracy　reliability　with　multiple　failure　modes　was　developed　and　a　case　study　example　for　a　five-axis　NC　machine　tool　was　used　to　demonstrate　the　effectiveness　of　this　method.　It　is　identified　that　each　improvement　of　the　geometric　errors　leads　to　a　decrease　in　the　maximum　values　and　mean　values　of　possibility　of　failure,　and　the　gaps　among　reliability　sensitivity　of　geometric　parameter　errors　improved　also　decreased.　This　study　suggests　that　it　is　possible　to　obtain　the　relationships　between　geometric　errors　and　specify　the　accuracy　grades　of　main　feeding　components　of　mechanical　assemblies　for　improving　machining　accuracy　reliability.　©　2014,　Polish　Academy　of　Sciences　Branch　Lublin.　All　rights　reserved.