Abstract: Tool failure can give a serious impact on the machining accuracy and surface quality of the parts. In order to control the machining quality, it is essential to know the tool wear condition. In this study, force and vibration signals were captured from milling experiments of titanium alloys on production field. Features extraction was carried out in time domain, frequency domain and time-frequency domain. Fisher's linear discriminant analysis based method was used to select the features which can better discriminate different tool wear conditions. The recognition of tool wear state was performed by using established support vector machine (SVM). High recognition accuracy using this method was shown in the results. The presented method can provide a reference to tool condition monitoring.