Real-time chatter detection using the weighted wavelet packet entropy

The weighted wavelet packet entropy (WWPE) is presented for real-time detection of chatter before it has fully developed in turning processes. This method is based on low dimensionality of process dynamics shown by chatter. The energy of frequency band that contains chatter information is weighted. Compared with the WPE, the WWPE is more sensitive to energy variation caused by the onset of chatter. Therefore, chatter can be detected in an earlier stage. From computational point of view, the wavelet packet matrix is employed instead of the Mallat algorithm, which is computationally expensive due to its complex convolution operations. A moving average filter combined with a threshold algorithm is also introduced to improve the performance of the WWPE. Finally, numerical simulations are conducted to demonstrate the efficiency of the proposed method. Moreover, the WWPE with different weights is experimentally explored and the proper value is obtained.