Control of roundness on turned cylindrical bars using artificial neural network

The emergence of government legislations in environmental matters has lead to improvements in machining processes in the recent years. Dry machining is one such attempt in which the use of fluids has been minimized. In this sense, the study of dry turning process is important. In a study by Marcos-Barcena et. al. (2005), the macrogeometric deviations (roundness) of UNS A92024 (Al-Cu) cylindrical bars turned dry under specified conditions of imposed cutting (cutting speed and feed) were reported. They established an exponential relation between roundness and the cutting parameters. For the ongoing research related online control of turning process, neural network was selected due to its capability to adapt to changes in surrounding environment. Hence, the relationship established in earlier study has been used for training and comparative study of back propagation (BP) and quick propagation (QP) neural networks. The results show that, overall, BP was the best algorithm for the application tested. The present study serves as an input for predicting the required process parameters for on-line control of roundness in turning process.