The method of BP neural network and DS evidence theory combination was used in tool wear monitoring. The wavelet packet decomposition method was used to extract the acoustic emission signals from the tool wear process, feature vectors was constructed and tool wear state was determined using BP neural network. Basic probability assignment of DS evidence theory was calculated through the output of BP neural network and training error, and level fusion was decided using recognition results of DS evidence theory to BP neural network. Experimental results show that this method avoids misdiagnosis of neural network and improves recognition accuracy and reliability of the tool wear monitoring system.