A methodological approach ball bearing damage prediction under fretting wear conditions.
Abatract-ReIiableon-line tool conditioning monitoring is an essential feature of modern sophisticated and automated machine tools. Appropriate and timely decision for tool-change is urgently required in the machining systems. Ample researches have been carried out in this direction. Recently artljicial neural networks (NN) are applied for thzk purpose in conjunction with suitable sensory systems. Its fwt processing capability is well-suited for quick estimation of tool condition and corrective measure to be taken. The present work uses back-propagation type training and feed-forward testing procedures for the neural networkx Three moa%lsusing dl~erentforce parameters are tried to monitor tool wear on-line. The close estimation of the modeled output to the actual wear value a%monstrates the possibility of successful tool wear monitoring. Copyright ~ 1996 Elsevier Science Ltd.