Abstract: During the machining process,it is difficult to obtain comprehensive information about tool wear characteristics using a single signal due to the complex and varied machining environment. A method is proposed that utilizes a combination of sound signals and surface image signals of workpieces through deep learning networks to recognize tool wear states. Initially,sound signals and surface image data from milling processes are synchronously collected,followed by the establishment of a dual-channel parallel convolutional neural network for feature extraction and fusion of one-dimensional sound signals and two-dimensional workpiece surface images. Finally,tool wear recognition is conducted. The results indicate that the use of information fusion methods can obtain more comprehensive tool wear characteristic information compared to single information recognition results,which is advantageous in improving the tool wear recognition effect.Moreover,the tool wear recognition accuracy and F1-score are above 95 %,allowing for effective recognition of the tool wear condition.

Key words: tool wear, wear recognition, information fusion, dual-channel convolutional neural network