Unstable vibrations during machining can harm both the tool and the workpiece,requiring careful selection of process parameters to avoid them. These parameters are usually set based on vibration models of the machining process. However, due to unmodeled dynamics or process variations, chatter can still occur, highlighting the need for online chatter monitoring systems. Existing methods often detect chatter only after it occurs, so there is a need for monitoring systems that can predict chatter before it occurs to ensure high-quality machining.

This thesis presents a new method to identify the dynamics of regenerative chatter from the measured process vibrations in milling. This method combines the synchronous once-per-revolution sampling of stable process vibrations with Operational Modal Analysis to estimate the Floquet multipliers of the delayed linear time-periodic dynamics in milling, all from the natural process vibrations without external excitation. The identified multipliers quantify vibration stability, enabling chatter prediction before it occurs. Additionally, they can be used to calibrate physics-based chatter models based on vibration measurements solely within the stable region.

The method’s accuracy in identifying Floquet multipliers is validated through extensive numerical simulations and two experimental case studies. The results show that chatter due to both Hopf and period-doubling bifurcations can be predicted from the process vibrations during stable cuts. Moreover, the experimental case studies demonstrate a vibration measurement system for implementing the presented method in standard milling operations and confirm its effectiveness in practice.