Abstract: Permutation entropy of a time series is a conceptual simple, robust and extremely fast complexity measure method for distinguishing conditions of a physical system. Considering the deficit of single scale permutation entropy in reflecting the specifics of milling force signals, a milling chatter on-line detection method based on indictors of single scale permutation entropy and scales 4 permutation entropy is proposed through experimental analysis. The experimental results show that this method can identify the chatter accurately and effectively. Meanwhile, this method is insensitive to noise and process parameters variation. Therefore, multi-scale permutation entropy based method possesses excellent potential for practical and on-line application.

Key words: milling chatter, multi-scale permutation entropy, on-line monitoring