Monitoring end-mill wear and predicting tool failure using accelerometers

Autoregressive models are fit to end-milling force and acceleration data. The machining modes are isolated and monitored for changes in energy attributed to the tool as it wears and approaches failure. The modal energies corresponding to these machining frequencies are shown to be closely linked to the condition of the tool.;Life tests are conducted using both a force dynamometer and an accelerometer to allow comparisons between the abilities of the two sensors. The acceleration signals are shown to provide more stable trends, therefore, only acceleration signals are investigated in detail. A monitoring scheme is developed that tracks the end-mill's wear and provides an early warning of failure. Six acceleration life tests are used to demonstrate the capabilities of the developed detection scheme under: standard conditions, extreme conditions, premature failure and different accelerometer locations. In all six cases, the scheme was able to provide a warning of impending failure several inches before the failure occurred.;However, the monitoring technique developed using univariate models is dependent on the cutting direction. In an attempt to develop a detection scheme that can overcome this imposition, multivariate models are fit to the data. By comparing the resulting trends, the multivariate models are shown to provide trends that have lower variability than when using univariate models.;Furthermore, using the parameters of the multivariate models, a monitoring technique is developed that is unaffected by changes in the cutting direction and provides an earlier warning of approaching failure than was possible with the univariate models. One additional advantage, the scheme is also independent of the accelerometers orientation. As long as the accelerometer is attached firmly to the spindle, the technique will be able to track the condition of the tool, regardless of the actual orientation of the accelerometer. Moreover, since the scheme is linked to the wear curve, the monitoring criterion can be incorporated on other systems where wear is the primary form of failure. Therefore, the technique has the ability to work on a wide variety of rotating equipment types.