| Diamond roller is a kind of high precision superhard diamond tool with complex surface,which is mainly used in shaping grinding wheel dressing.Arc and straight line constitute the complex profile of diamond roller.The precision of diamond roller directly determines the precision of the dressing grinding wheel and indirectly affects the precision of the workpiece processed by the grinding wheel.At present,diamond wheel grinding method is often used for precision modification of diamond roller.This method relies heavily on manual operation,high cost,low efficiency,the industry is extremely urgent demand for automatic processing.To solve this problem,this paper proposed to identify the grinding contact state by establishing the cut grinding contact model and the longitudinal grinding contact model so as to optimize the machining parameters.Based on this requirement,acoustic emission was used to monitor the grinding contact process.The acoustic emission signals of longitudinal grinding and cut grinding were stabilized by EEMD modal decomposition and VMD variational modal decomposition,and the cut grinding contact model was established by VMD energy reconstruction method.The time series analysis method in deep learning was used to establish the longitudinal grinding contact model.The main work of this paper is as follows:(1)Combine with the actual situation that the long straight line segment of complex diamond profile is modified by longitudinal grinding method on optical curve grinder at present,and the arc surface of complex profile is modified by multiple cut point grinding.The contact state model of the cut grinding process was established to realize the function of eliminating the headway so as to optimize the feed speed in the cut grinding process and reduce the headway time in automatic grinding.The contact state model of the longitudinal grinding process was established to optimize the feed rate so as to ensure the continuous contact grinding of diamond wheel and diamond roller.(2)Several commonly used grinding monitoring methods are analyzed,and the effectiveness of acoustic emission signals used to monitor grinding contact state is proved.Acoustic emission signals collected from longitudinal grinding process and cutting grinding process are analyzed and processed by various methods,including time-frequency domain analysis method,etc.(3)The contact model of the cut grinding was established based on the VMD energy reconstruction method.The original acoustic emission signal was decomposed into several components by VMD method,and the IMF4 component was selected to replace the original signal and reconstruct the signal energy to judge the grinding contact state.Compared with the conventional contact identification methods,the time domain threshold method and the effective threshold method,it is found that the VMD energy reconstruction method is significantly better than other methods to identify the contact state of diamond roller in the grinding of diamond wheel.(4)A longitudinal grinding contact model based on BILSTM-ATTENTION was established.Firstly,the collected non-stationary nonlinear acoustic emission time series signals are stabilized by EEMD modal decomposition method and decomposed into several components.Then,the components with low correlation are removed by correlation analysis.The remaining modal components are windowing with 50%overlap Hanning window,and eigenvalues are extracted according to the frame.The framework of BILSTM-ATTENTION algorithm model was constructed to establish the contact state recognition model.On the basis of BILSTM,this model introduces the Attention mechanism to extract implicit information.The experimental results show that this model is significantly better than other models. |
