Research On The Gringing Performance Degradation Of Diamond Wheel Based On Acoustic Emission Signal

Optical glass is a hard and brittle material that is widely used in various technical fields such as high-power lasers,large astronomical telescopes,medical equipment,and cameras.Precision and ultra-precision grinding are important processing methods for optical surfaces,and an important process to ensure the high-quality,high-precision,and efficient production of optical glass components.In ultra-precision grinding,the grinding state of the grinding wheel greatly affects the processing quality and efficiency of the optical element.Dressing the wheel in time can ensure stable quality grinding.Grinding wheel wear condition monitoring is an important guarantee for precision and ultra-precision grinding quality.In the process of precision and ultra-precision grinding,the generation of micro-cracks on the surface of the grinding wheel and the changes in the abrasive particles are accompanied by acoustic emission phenomena.Because the acoustic emission signal is the most timely and sensitive characterization of the grinding interference.Therefore,in this paper,the wear state of the grinding wheel is monitored based on the acoustic emission signal.The whole experiment process is to collect the acoustic emission signals and grinding force signals of the whole life cycle grinding process to analyze the macro and micro changes of the grinding wheel,and the topographic image of the fixed position of the grinding wheel is used for verification.The purpose is to study the correspondence between the wear pattern of the grinding wheel and the acoustic emission signal,to realize the identification and prediction of the wear state of the grinding wheel.The main contents of this article follows:(1)Based on the three-axis precision machine tool,build a grinding experiment platform and signal acquisition system.Carry out the full life cycle grinding experiment of resin diamond grinding wheels,and the time-domain signal amplitude of the acoustic emission increases slightly as the grinding progresses,and the time-domain signal changes periodically,It can be seen from the surface morphology of the grinding wheel that the abrasive particles on the surface of the grinding wheel are gradually worn out and clogged.(2)Research the relationship between wheel wear and acoustic emission signals.The acoustic emission signals of the grinding wheel during the life of the grinding wheel are divided at equal time intervals,All samples were analyzed by wavelet packet,and the sensitive frequency band of the acoustic emission signal during the grinding cycle of the grinding wheel is[0-250]KHz.The spectrum analysis of the acoustic emission signal in the sensitive frequency band shows that the spectrum sequence of the sensitive frequency band gradually increases with the grinding.(3)Research the theory of dimensionality reduction algorithm to reduce the dimensionality of the spectrum sequence of sensitive frequency bands.By comparing the visualization results of PCA and LDA dimensionality reduction selection features,it can be seen that LDA compared with PCA dimensionality reduction selection features can better reflect the difference in acoustic emission signals during grinding.Auto_encoder separates the inherent pattern of the acoustic emission signal from random interference through feature self-learning to obtain a self-learning reconstructed spectrum of the sensitive frequency band.The waveform of the reconstructed spectrum can be used to make a preliminary judgment on the wear state of the grinding wheel.(4)Research the theoretical model of learning vector quantization(LVQ)and BP neural network algorithm.The selected features of dimensionality reduction are used as training data to train the LVQ and BP neural network grinding wheel wear state recognition model.Through the accuracy of LVQ and BP neural network model prediction,the characteristics of LDA dimensionality reduction selection can reflect the changes of grinding wheel wear state more than PCA and Auto encoder dimensionality reduction selection features,and can make better identification of different wear states of grinding wheel.Finally,the BP neural network regression model is trained based on the characteristics selected by LDA,and the grinding wheel wear fitting curve is obtained.