Lightweight and Robust Chatter Detection Algorithms for Milling

Abstract

Most machining processes suffer from self-excited chatter vibrations that destroy the work surface and affect machine tool’s health. Therefore, accurate and timely detection of chatter vibrations is critical for productivity and longevity of the manufacturing equipment. This paper presents two different chatter detection algorithms for milling process by monitoring power spectrum of the vibration signal measured on the machine. Firstly, a real-time suitable computationally efficient approach is presented, which computes total spectral power of the vibration signal using time-domain variance operator and forced vibration power using moving Fourier transform. Power spectrum of chatter vibration is then evaluated by deducting the forced vibration component from the total signal power. The second method is based on Principal Component Analysis (PCA), which extracts the dominant harmonics of the measured vibration signal. Forced and chatter vibration harmonics are then clustered and labeled to evaluate the forced and chatter vibration powers based on their eigenvectors and eigenvalue magnitudes. For both methods, the power ratio (PR) is used to detect chatter. Once PR exceeds 0.5, chatter starts to dominate whole process, and it is detected. Both algorithms are experimentally tested in machining flexible workpieces. Results show that although both methods perform well on detecting chatter timely, computationally efficient algorithm gives false alarms for flexible workpiece while PCA based algorithm provides more robust chatter detection.