Tracking the health of cutting tools under typical wear conditions is advantageous to the speed and efficiency of many manufacturing processes. Although techniques exist to monitor tool performance, most are dependent on cutting direction, sensor orientation, cutting procedure, or the use of computationally intensive mathematical models. In this work, several solutions are proposed such that real-time analysis of signal variance and frequency magnitudes is possible through the identification of trends in the transient behavior of tri-axial force dynamometer signals. Moreover, the nature of both the transient variance (the temporal change in signal variance) and the frequency magnitude trends are shown to be independent of direction through the observation of both linear and pocketing mill sequences. Ten autonomous methods for failure prediction are discussed, tested, and presented. Analysis methods are demonstrated through the implementation of procedures derived from autocorrelation, FFT type, and Autoregressive models. Methods demonstrating high levels of success are subsequently contrasted for forecast success and computational requirements. Forecast success is demonstrated by both FFT type and Autoregressive Models. Methods discussed have the potential for on-line implementation using current commercially available computational capacity.
All Science Journal Classification (ASJC) codes
- Materials Science(all)
- Mechanical Engineering
- Industrial and Manufacturing Engineering