Neural network supported inverse parameter identification for stability predictions in milling

A new approach for the inverse identification of structural dynamics and process parameters in milling operations is presented. An algorithm is fed with process information and the stability state of recorded cuts. By comparing stability model predictions with experimental data, the underlying relationships and governing parameters are estimated with the help of Artificial Neural Networks. The capabilities of the approach are demonstrated using a simulative example and two experimental studies. The algorithm shows to be capable of approximating unknown relationships such as spindle speed dependent dynamics, which in turn allows for accurate stability predictions without the need for extensive dedicated measurements. (C) 2020 CIRP.