Positioning Accuracy of CNC Machine Tools : Thermal Effect Prediction Using Neural Network

• Published in: Nihon Kikai Gakkai rombunshuu. C hen Vol. 62; no. 599; pp. 2686 - 2691
• Main Authors: HIROTA, Yasuhiro, TSUTSUMI, Masaomi, NISHIWAKI, Nobuhiko, CHEN, Liang
• Format: Journal Article
• Language: Japanese
• Published: The Japan Society of Mechanical Engineers 1996
• Subjects: Back Propagation; CNC Machine Tools; Machining Center; Positioning Accuracy; Recurrent Neural Network; Thermal Effect
• ISSN: 0387-5024; 1884-8354; 1884-8354
• DOI: 10.1299/kikaic.62.2686

Methods and results are presented for applying neural networks to the prediction of positioning error due mainly to thermal effects. Experiments on a machining center show that the positioning error is affected by the running conditions, such as the temperature variation of the nut of the ballscrew and the feedrate. In this study, we have tried to identify the relationships between the thermal-induced positioning error and running conditions. Both recurrent and feedforward neural networks have been used in the identification and their learning effectiveness has been compared. The results have shown that the recurrent network performs better than the feedforward network in this case. To further improve the error prediction accuracy of the network, a training technique that uses additional input has been proposed and its effectiveness has also been proved.