Research On Tool Wear Monitoring Techniques Based On Wavelet Analysis And Integrated Neural Network

In the system of modern manufacture, in order to ensure the safety and machining quality of the automated processing equipment, it is urgent to solve the problem of monitoring. Cutting tool is one of the most important factors in the machining process. Tool wear will not only affect the quality and precision of the products, but also may destroy the machine and the workpieces, even endanger workers' lives. In addition, due to the diversity of processing conditions, the variability of cutting parameters and tool wear, it makes the condition monitoring of the cutting tools become a major tache in the entire manufacture process.In this paper, the multi-sensor fusing technology was applied in the tool wear monitoring. There are three sensors ,the force measuring instrument of Kistler 9257 B, the broadband sound launching sensor of SR12 and vibration sensor of B&K4370 ,using to sensing signal. We also did some research about signal processing, feature extraction and pattern recognition to monitor the condition of tool wear.Wavelet analysis method, which is very effective to analyze instability signals, was used to process cutting force signal and vibration signal in this paper. Wavelet-packet transform was not only used to analyze low-frequency signal, but also the high-frequency signal. And it can decompose any signals including sinusoidal signal to independent and arbitrarily precise frequency band orthogonally without redundance and omissions. For this reason, we process the frequency band energy of the cutting forces and vibration signals, and can obtain the sensitive frequency band features of tool wear. Based on the changes of corresponding frequency band energy, we can monitor the condition of the tools effectively.Intelligent Diagnostic System is based on the wavelet analysis and neural network. In the first instance, BP neural network is used to recognize the characteristics of the cutting forces and vibration signals. Aiming at the disadvantages of it, the tool wear identification system is presented which is based on integrate neural network. It is made up of one vibration neural network and two cutting force neural networks. This method utilizes all the effective information sufficiently in the machining process. The result shows that the identification efficiency and speed are improved greatly. It is proved to be practical and it is significant for the tools monitoring in the machining process.