In order to improve the ability of tool wear monitoring in the machining process, the spindle current and feed current are selected as the main information. The-multi feature extraction of current signal is based on the wavelet decomposition and soft-sensing model. The information of tool wear and break can be reflected from the processing of feed and spindle driving. On the basis of above all, the data fusion of the multi-feature information and the construction of intelligent alarm model can be implemented based on Parzen windows. The alarm boundary is determined based on Pa Ta criterion, so the intelligent alarm of tool condition can be achieved. When applying this technology to tool machining, it is shown from the experiment results that this technology can monitor tool condition in real-time and alarm in time.