Chatter Detection In Milling Process Based On Multicomponent Signal Decomposition Method

The weak rigidity of the thin-walled work piece leads to the chatter in milling process,which affects the machining quality and production efficiency.It is necessary to detect chatter at the early stage and take action to suppress it.Most chatter detection algorithms concentrate on off-line chatter detection,which cannot avoid negative affect.In practical machining,it is significant to detect chatter on-line.For the signal which contains chatter sign has a low signal to noise ratio,it increases difficulty extract chatter feature.A chatter detection algorithm based on multicomponent decomposition method is proposed in this paper,which focuses on three modules: signal processing,feature extraction and state identification.The main work is as follows:Milling force models based on linear and nonlinear cutting theory are respectively established.The frequency spectrum structures of vibration response under different models are compared.The vibration response characteristics under different cutting conditions are analyzed,which provides the evidence for extracting chatter characteristics.A multi-component signal decomposition method is proposed,which estimate instantaneous frequency based on Spectral Centralization Index and estimates amplitude based on Vold-kalman filter.The simulation results show that compared with the similar methods,the proposed method can effectively decompose the multi-component signal,suppress noise and it has an advantage to analyze non-stationary signals.A chatter feature extraction method based on vibration signal energy distribution is proposed.The milling vibration signal is decomposed into a set of single-component signals.Because the energy of vibration is concentrated on one single-component signal which contains certain frequency,the uniformity of energy distribution of all component signals in the set is reduced.Thus the energy of all component signals is normalized and entropy is extracted as chatter indicator.In addition,the existence of optimal parameter during entropy extraction is discussed theoretically.A method based on Gaussian Mixed Model is proposed for milling state identification.Because chatter feature in different milling states corresponds to different distribution intervals in Gaussian Mixed Model,it is feasible to detect chatter state by comparing the Gaussian Mixed Model of monitoring signal and normal cutting signal.A chatter detection algorithm integrated with the above modules is proposed.Firstly,multi-component signal decomposition method is used to process milling vibration signal,then entropy is extracted as chatter feature.And Gaussian Mixed Model is used to fit the feature to idendtificate cutting state.The optimal feature extraction parameters are obtained by milling experiments.It is verified that the algorithm under the optimal feature extraction parameter can detect chatter earlier,which is of great significance for chatter suppression.