Monitoring of tool wear in CFRP drilling based on acoustic emission

Abstract Carbon fiber–reinforced composites (CFRP) are widely used in the aerospace industry due to their excellent properties. The quality of drilling in CFRP components has a significant impact on their structural integrity and in-service performance. Tool wear is a key factor determining drilling quality. To achieve data-driven process control, this study developed an integrated signal acquisition platform capable of simultaneously capturing acoustic emission (AE) and axial force signals during the drilling of CFRP. Under controlled and repeatable drilling conditions, acoustic emission and axial force signals were systematically recorded at each stage. A five-layer decomposition thresholding method based on the db4 wavelet was applied to the acoustic emission signals. This method effectively suppresses broadband noise while preserving the transient impact characteristics associated with the progressive degradation of the cutting edge. By analyzing acoustic emission signals and axial force in the time and frequency domains, among other methods, characteristic parameters highly correlated with the tool’s wear state were identified based on the Passion correlation coefficient. Finally, a classification method based on a convolutional neural network (CNN) under batch processing conditions was proposed to enable effective identification of tool wear under various operating conditions. This lightweight convolutional neural network achieved 94.44 % identification accuracy for tool wear, demonstrating strong potential for engineering applications.