Prediction Of Diamond Tool Wear Based On Generalized Regression Neural Network

In the mass production process,the processing of high-angle chamfered surface of the mobile phone shell is taken as an example.Due to the abrasion of the tool,the highlight chamfering phenomenon such as drawing,film collapse,and fogging occur.In recent years,there have been many studies on tool wear prediction,but few tools are used for high-flute chamfer surfaces.Existing tool wear prediction methods can only solve the problem of tool wear with large depth of cut.The processing of shell high-angle chamfering has a depth of cut of only 0.02 mm,which requires a very high surface finish.The wear of 8um may seriously affect the surface quality of the workpiece.Therefore,this paper proposes a tool wear prediction method based on vibration signals.The research content includes the following points:The relationship between the quality defects and tool wear in the process of production and processing of high-angle chamfered surface of mobile phone case was analyzed.The selection test of the sensor was designed according to its processing characteristics,and the current signal and vibration signal monitoring method were preliminarily determined.According to the actual production and processing process standards,a test plan has been prepared,and a current and vibration signal data acquisition platform have been built and the test have been completed.The short-term energy method used in speech recognition was used to intercept the signal,and the signal was denoised by wavelet analysis.The time-domain,frequencydomain,and wavelet-packet decomposition features are extracted,the magnitude differences between features are reduced by normalization processing,and feature vectors related to tool wear are obtained through correlation analysis.The BPNN network model and GRNN network model were established.The comparative analysis found that GRNN was obviously due to BPNN in training time and generalization ability.the advantages and disadvantages of the current signal and the vibration signal are also compared and analyzed.It is concluded that the vibration signal has better prediction accuracy than the current signal.A model between the vibration signal and the tool wear amount have been established,set the threshold value to identify the corresponding product defect and the tool change time point of each wear stage,and finally realize the tool radius compensation in real time through the tool wear prediction model.