Rolling　bearing　is　not　only　one　of　the　most　common　components　in　machinery　and　equipment,　but　also　prone　to　fault.　So　it　is　very　important　for　the　recognition　of　bearing　faults.　Aiming　at　identifying　the　common　faults　of　bearing,　for　the　problem　about　isolated　points　or　noises　mixed　in　vibration　signal,　in　this　paper,　a　classification　model　is　proposed,　which　is　based　on　the　affinity　fuzzy　support　vector　machine.　And　compared　with　the　traditional　support　vector　machine　(SVM),　the　introduced　model　of　fuzzy　support　vector　machine　(FSVM)　works　better.