Tool wear directly affects the machining quality and dimensional accuracy of the workpiece. Correctly grasping the tool wear status and changing tools in time will reduce machine downtime, directly improve processing efficiency. In order to improve the accuracy of tool wear state recognition, an improved particle swarm optimization PNN (IPSO-PNN) neural network based on parameter strategy was presented to identify the tool wear state. BP neural network has the disadvantage of slow convergence speed and easy to fall into local optimal.Compared with BP neural network, IPSO-PNN has simple structure and simple and fast training. Compared with the simulation results of BP neural network and standard PNN neural network, the results show that IPSO-PNN neural network has high recognition accuracy, fast convergence speed and short simulation time. It can effectively improve the accuracy of tool wear recognition.