Abstract: In view of the problem of on-line identification of cutting wear of roadheader, a multi-feature signals recognition method based on PNN neural network is proposed.The vibration and acoustic emission characteristic signals of the picks with different pick wear degrees during the cutting process are extracted and analyzed separately.The four characteristic parameters of the vibration acceleration, the peak value of acoustic emission signal and the root mean square of the frequency domain diagram of two characteristic signals were used to obtain the variation of vibration signal, acoustic emission signal and pick of different wear degrees.A Multi-feature signal sample database for five different pick wear degree were established.The multi-feature signal samples were used to study and train the PNN neural network, the recognition model of pick wear degree was established to realize the accurate identification of pick wear degree.The results show that the recognition accuracy of model based on PNN neural network has higher recognition accuracy, and the accuracy of recognition and prediction are about 93.3% and 95%, which is 3.3% and 15% higher than that of BP neural network.The results provide an important technical means for identifying the wear degree of picks precisely and improving the work efficiency.