Estimation of tool wear based on adaptive sensor fusion of force and power in face milling

Sensor fusion of multi-sensory measurements is believed to enhance accuracy of tool condition monitoring methods for sensor signals which are at least partly complementary. In this paper, a novel statistic based on the least squared error criteria is used to evaluate performances of models arising from sensor fusion. An adaptive sensor fusion of cutting force and electrical power signals, which performs optimally even in case of failure in any acquiring channel, is proposed. Combinations of signal processing techniques are used to extract accurate and robust features, which are refined by feature space filtering techniques to obtain improved and robust estimators of tool wear. It is shown that the model using sensor fusion of force and power produces superior results to those using single measurements. Error bounds of the estimates are also provided as prediction limits. Significant improvements are obtained compared to the existing methods in terms of accuracy and robustness.