In　this　paper　wavelet　packet　technique　is　used　to　decompose　the　vibration　signals　of　grinding　process.　Since　these　decomposed　signals　in　every　band　have　different　energy,　they　can　be　used　to　represent　the　vibration　exception　with　fuzzy　ART　neural　network.　The　fuzzy　ART　neural　network　can　generate　a　new　clustering　to　give　an　alarm　when　vibration　exception　of　unknown　pattern　appears.　Compared　with　traditional　methods　this　method　can　identify　the　known　and　unknown　pattern　exceptions　of　grinding　process.　The　result　of　practical　application　shows　that　this　method　is　very　efficient.