Tool condition monitoring technique for deep-hole drilling of large components based on chatter identification in time–frequency domain

Abstract This report presents a tool condition monitoring technique for deep-hole drilling that will ensure the quality of precision large-scale components. To avoid catastrophic tool failure, in-process identification of chatter is essential, as is identification of transient and spontaneous events that are highly non-Gaussian and nonlinear. Therefore, we propose a novel system that combines short-time Fourier transform and spectral kurtosis analysis in the time–frequency domain to determine relevant chatter and transient vibration information from an accelerometer signal. A dynamic model of a long drill was used to provisionally set a target frequency band (TFB) in which chatter is predominant, and then a verification experiment was performed on a drilling process involved in steam turbine rotor manufacturing. In the tests, chatter appeared abruptly near the TFB and rapidly caused tool failure, while transient events were less influential. This paper also presents an energy-based chatter criterion that can be applied throughout long-term manufacturing. The proposed method is not only highly capable of preventing tool failure by detecting the onset of chatter, but also provides comprehensive information about the tool’s state.