A method of pattern recognition based on discrete hidden Markov model(DHMM) is proposed to monitor tool wears and predict tool failures in the cutting process. At first the FFT features are extracted from the signal of the tollholder vibratio n and the cutting force in cutting process, then the FFT vectors are presorted and coded into code book of int eger numbers by using the selforganizing feature map(SOM), and these code book s are introduced to DHMM for machine lea rning to build up 3HMMs for different tool wear stages. And then, pattern of h idden Markov model(HM M) is recognized by using maximum probability. Finally, the results of tool wear recognition and failure prediction experiments show that the me thod is effective.