In cutting hardened die steel with carbide end mill, the criterion for judging a tool failure are divided broadly into three categories, that is, wear, chipping or breakage. Among them, the prediction of the tool failure caused by wear is relatively easy, but it is difficult to estimate the tool failure caused by the remaining two. Thus it is desirable to select the cutting conditions which lead a tool failure caused by the wear. In this study a mathematical model for predicting the tool failure patterns is identified by applying simultaneous approach to fuzzy cluster, principal components and multiple regression analysis. In the approach the kind of machine, spindle speed, feed, radial depth of cut, axial depth of cut, free length and run out are chosen as predictor variables from the tool failure tests using carbide radius end mill, and the tool failure patterns (wear, chipping and breakage) as response variables. As the results, three clusters are obtained and from each cluster input variables affecting the tool failure patterns are extracted. According to the approach it is possible to predict the tool failure patterns and to select the cutting conditions leading to the tool failure caused by wear.
