An Optimal Weighted Wavelet Packet Entropy Method With Application to Real-Time Chatter Detection

Chatter detection is important for preventing harmful effects induced by chatter in machining processes, and for improving productivity and part quality. Chatter leads to decreased uniformity of energy distribution in the frequency domain. Such uniformity variation can be captured by the weighted wavelet packet entropy (WWPE). In this paper, an optimal WWPE method is proposed for a real-time chatter detection. We investigate the optimal weight, with which the relative distance between stable and chatter states is maximized in terms of the WWPE. Extreme value statistics with a visual method is proposed to compute a theoretically rigorous threshold by modeling tails of the distribution. Using the wavelet packet matrix, the proposed method achieves computational efficiency in milliseconds, and is thus suitable for real-time applications. Finally, machining experiments are conducted to derive the optimal weight and verify the effectiveness of the proposed method.