Research On Chatter Condition Monitoring Method In Turning

The stability of the cutting process is the key factor restricting the development of cutting technology.When turning,the intense chatter between the turning tool and the workpiece not only affects the machining accuracy of the machine tool,reduces the working efficiency of the machine tool,but also aggravates the wear of the turning tool,and even damages the spindle and the tool.The research on chatter monitoring and prevention is an important research topic to deal with the cutting stability problem and has important scientific research significance.In this paper,combined with metalcutting dynamics,signal processing,and deep learning methods,the phenomenon of regenerative chatter in turning is taken as the research object,focusing on signal feature recognition of turning chatter in different stages,and studying the chatter state recognition model based on deep learning method.The main research contents are as follows:1.To find the quantitative indicators of turning chatter and explore the mechanism of regenerative chatter in turning,three processing stages are divided according to the roughness level of the turning surface and the general process of chatter excitation:stable processing,transition processing,and intense chatter.A mechanical model of regenerative chatter in a single degree of freedom(DOF)outer circular turning was established to solve the blade diagram in the cutting stability domain of the target machining system.The tool acceleration vibration signal was selected as the acquisition and monitoring object,and the relationship between different development stages of chatter and the changing trend of workpiece surface roughness was verified.2.Aiming at the denoising problem of turning vibration signals,a joint denoising method based on complete EEMD with adaptive noise(CEEMDAN)and wavelet threshold denoising is designed.The original signal was decomposed by CEEMDAN and denoised by an improved threshold function with a scaling coefficient.The signalto-noise ratio(SNR)and root mean square error(RMSE)indexes verify that the improved threshold function can reduce signal characteristic loss,improve signal quality after denoising and reconstruction,and achieve a better denoising effect.3.To realize the classification and recognition of the turning chatter stage,a chatter state classification model based on a bidirectional long short term memory network(LSTM)was designed.The feature compression method based on autoencoder is studied,and the feature dimension reduction network based on denoising autoencoder(DAE)is designed,which is connected in series to the bidirectional LSTM network to reduce the learning difficulty of the network and improve the anti-interference ability.Sample test results show that the proposed chatter state recognition model can effectively distinguish stable processing,transition processing,and intense chatter stage.4.To realize online monitoring of chatter phenomenon in the machining process,an online monitoring system and related software for turning chatter were designed and developed,and an online monitoring and verification test-bed for turning chatter was set up to carry out cutting experiments.The comprehensive experimental evaluation results show that the proposed method can effectively identify the regenerative chatter in the turning process,and the accuracy and effectiveness meet the theoretical expectation.