Continuous Online Tool Wear Estimation in Face Milling using Adaptive Sensor Fusion of Force and Power

Accuracy, reliability and robustness of tool condition monitoring methods could be enhanced by sensor fusion of multi-sensory measurements. In this paper, an estimation method for continuous online monitoring of tool wear in face milling through adaptive sensor fusion of cutting force and electrical power signals using multiple linear regression models is proposed. Force and power signals are processed in parallel to extract useful features, which are insensitive to random process variations and power supply fluctuations. Features are refined with the help of Isotonic regression and Exponential smoothing techniques to obtain improved and robust predictors of tool wear. The model using sensor fusion is found to be superior to those using single measurements according to a statistical model selection criterion. Prediction limits are also provided to indicate the probabilistic worst case predictions of tool wear. The proposed method gives improvement over the existing models in accuracy and robustness and is suitable for industrial application.