Tool wear monitoring using a novel parallel BiLSTM model with multi-domain features for robotic milling Al7050-T7451 workpiece

Industrial robots have great potential to machine large parts. However, the vibration or chattering induced by their inherent weak stiffness can easily damage or break the tool. Therefore, this paper introduced a novel tool wear monitoring method for robotic milling. Firstly, a multi-domain features extraction method was proposed to obtain local features. Then, a novel deep learning model with two parallel a parallel bidirectional long short-term memory networks (BiLSTM) (Vibration branch and Cutting Force branch) was introduced to fuse the multi-domain features and learn the time dependence patterns. The proposed method was verified on both based on robot milling Al7050-T7451 workpiece dataset and the 2010 prognostics health management (PHM) dataset. The experiment results showed that the proposed method acquired an excellent predication accuracy and strong adaptability to the change of cutting parameters. The results show that the average root mean square error (RMSE) for wear recognition based on the robot milling dataset is 10.61, with an average mean absolute error (MAE) of 9.104. The average RMSE for wear recognition based on the computerized numerical control (CNC) milling dataset is 7.83, with an average MAE of 6.62.