Abstract: | Automatic monitoring tool condition becomes more important and necessary for preventing work_pieces and tools from damage in automatic manufacturing. A significant amount of research has been performed in this field during the past decades. However, the reliability for tool condition monitoring still need to be improved because of the high uncertainty of machining processes. This paper presents a new adaptive tool condition monitoring system based on wavelet packet and neural network. In this system, wavelet packet is used to decompose the cutting vibration signal into different frequency bands, multiple features are extracted as the inputs to the neural network, and a fuzzy neural network (FNN) estimates tool wear states based on these features. It is shown, from the experimental results that this system can identify the tool wear states with higher accuracy and can improve the reliability of neural network by learning human knowledge and learning sampled training data. |