Online machining chatter forecast based on improved local mean decomposition

Machining chatter is the self-excited vibration between the cutting tool and work piece that can deteriorate surface quality and increase cost. A reliable machining chatter forecast method is needed to recognize chatter symptom, while the chatter is developing. In this paper, an online machining chatter forecast framework is proposed. The improved moving average algorithm is designed to improve the performance of the local mean decomposition (LMD) method in dealing with non-stationary force and vibration signals. According to the result of signal character analysis, some sensitive features are extracted to build characteristic vectors. A machining chatter forecast model is put forward based on the hidden Markov model (HMM). The Viterbi algorithm is used to resolve the optimum cutting transient sequence. The mapping relationship between vector sequences in the input space and images in the pattern space has been developed based on the posterior transient transform probability. An experimental platform has been implemented to illustrate and validate the proposed method. Case studies showed that machining chatter could be forecasted effectively in advance during its developing stage. Quantitative analysis and comparison verified that both the precision and timeliness could be improved.