On-line chatter detection and identification based on wavelet and support vector machine

Chatter is very harmful to precision machining process. To avoid cutting chatter effectively, a method based on wavelet and support vector machine is presented for chatter identification before it has fully developed. Wavelet transform, which can image the information in both the time and frequency domain, is applied as an amplification for the chatter premonition. Each wavelet packet's energy regularly changes during the development of the chatter, which has a time advantage for the identification. Therefore, a two-dimensional feature vector is constructed for chatter detection based on the standard deviation of wavelet transform and the wavelet packet energy ratio in the chatter-emerging frequency band. A support vector machine (SVM) is designed for pattern classification based on the feature vector. The intelligent recognition system, composed of the feature extraction and the SVM, has an accuracy rate of 95% for the identification of stable, transition and chatter state after being trained by the experiment data. The method can be applied in different machining processes.