Research On Tool Wear And Tool Breakage Condition Of CNC

As the executor of the metal cutting process, the tool will inevitably suffer from tool wear or even breakage during the cutting process. The changes in the tool conditions due to tool wear will have direct effect on the quality of products and lead to increased production costs, and in turn will reduce the competitiveness of their products. Therefore, it is of great importance to conduct real-time monitoring on the tool wear condition in order to improve the quality of products, to lower down the costs of production as well as to improve the efficiency of production.In order to solve the problem of tool wear, we would like to build a signal acquisition system based on the Lance LC0151T vibration sensor and the Adlink PCI-9221acquisition card, so that we would be able to acquire the signals for different tool wear conditions through the orthogonal experimental design method, followed by a series of time domain analysis, frequency domain analysis and wavelet packet analysis. In particular, the wavelet packet transformation is not only able to decompose the low-frequency signal, but also able to do decomposition of the high-frequency signal. Due to such advantages of the wavelet packet, we would choose the energy value of each band based on the wavelet packet decomposition as the characteristic values for various tool wear condition, and hence obtain the bands corresponding to the tool wear conditions.Making use of the strong nonlinear mapping function of neural networks, we would also build a BP neural network, which is expected to help build up the mapping relationship between the tool conditions and the characteristic vector of the vibration signals. Furthermore, a monitoring system for the tool wear conditions would be constructed eventually based on the wavelet packet analysis and BP neural network, with specifying the number of input nodes, hidden layers and output nodes in the BP neural network.At last, we develop a tool wear condition identification system based on C++ Builder and Matlab. From the experimental results we know that the method is feasible.