On-line tool condition monitoring in face milling using current and power signals

The vast majority of tool condition monitoring systems use the cutting force as the predictor signal. However, due to prohibitive cost to performance ratios and maintenance and operational problems, such methods are not favoured by industries. In this paper, a method for continuous on-line estimation of tool wear, based on the inexpensive spindle motor current and voltage measurements, is proposed for the complex and intermittent cutting face milling operation. Sensors for these signals are free from problems associated with the cutting forces and the vibration signals. Novel signal processing strategies have been proposed for on-line computation of useful features from the measured signals. Feature space filtering is introduced to obtain robust and improved predictors from the extracted features. A multiple linear regression model, built on the filtered features, is then used to estimate tool wear in real-time. Very accurate predictions are achieved for both laboratory and industrial experiments, surpassing earlier results using cutting forces and estimation methods based on complex methodologies such as artificial neural networks.