Tool wear monitoring can greatly enhance the ability of machine tool fault diagnosis and improving accuracy of manufacturing. Though many associated researches have been done by some specific methods, none satisfied results have been obtained. Artificial neural network is a rapidly developing area in the recent years, and has demonstrated a great application in manufacturing process monitoring. However, standard bp neural network has disadvantages of slow convergence rate, long period of learning time and low accuracy. Thus in this paper, the mathematical model of network is established, and an improved algorithm for bp neural network is presented, meanwhile how to decide the number of units in hidden layer is also discussed. The advantages of improved bp algorithm are confirmed with the help of simulation. Simulation results show that tool wear monitoring method based on improved bp neural network is applicable and effective.
