This paper proposes a novel approach based on a Deep Residual Shrinkage Network (DRSN) and a Simple, Parameter-Free Attention Module (SimAM) for monitoring the tool wear state in computer numerical control (CNC) machines, using multi-sensor data as input. By collecting vibration, cutting force, and acoustic emission signals, the method converts them into time-frequency images using continuous wavelet transform (CWT) and employs parallel convolutional neural networks (CNNs) for multi-source feature extraction and fusion. On this basis, the DRSN-SimAM model is constructed, which enhances the feature selection and denoising capabilities of the model in strong noise environments through soft thresholding and the parameter-free SimAM attention mechanism. The results show that the proposed method significantly reduces the root mean square error (RMSE) in tool wear prediction to 11.73 μm, outperforming traditional ResNet and DRSN models. This proves its effectiveness in improving the accuracy and robustness of tool condition monitoring, providing reliable technical support for predictive maintenance of tools in intelligent manufacturing.
Wang et al. (Wed,) studied this question.