Performance Reliability Prediction Of Tools In Metal Cutting Using The Validity Index Neural Network
Ratna Babu Chinnam, William J. Kolarik, Chowdary V. Manne · International Journal of Modelling and Simulation · 1996
Previous research indicates that neural network technology is applicable to tool condition monitoring in metal-cutting processes. The backpropagation, competitive learning, and adaptive resonance- based approaches have been used in tool classification problems (classifying a given tool as either fresh or worn). However, previous research has not addressed confidence intervals in tool monitoring with respect to neural networks technology. In this paper the authors present an alternative approach that uses the validity index neural network that incorporates radial basis functions as nodal functions to predict the real-time conditional performance reliability measure of risk relative to tool wear. The specific application considered is in-procen monitoring of the condition of the drill bit in a drilling process. The promising results demonstrated by application examples show that this technology seems to have a high potential in automated tool condition monitoring.