Monitoring End-Mill Wear and Predicting Tool Failure Using Accelerometers

Abstract

Autoregressive models are fit to end-milling acceleration data and the Data Dependent Systems methodology is utilized to isolate the modal energies of the first and second multiples of the tooth pass frequency. The modal energies are shown to be closely linked to the wear curve and a detection scheme is developed that is capable of tracking the end-mill’s wear and providing an early warning of impending failure. Six life tests are conducted under varying conditions to demonstrate the capabilities of the detection scheme: standard cutting conditions, extreme cutting conditions, premature catastrophic failure and accelerometer placement. In all six cases, the detection scheme was able to provide a warning of impending failure several centimeters before the failure occurred.