Cutting tool wear state assessment during the manufacturing process is extremely significant. The primary purpose of this study is to monitor tool wear to ensure timely tool change and avoid excessive tool wear or sudden tool breakage, which causes workpiece waste and could even damage the machine. Therefore, an intelligent system, that is efficient and precise, needs to be designed for addressing these problems. In our study, an end-to-end improved fine-grained image classification method is employed for workpiece surface-based tool wear monitoring, which is named efficient channel attention destruction and construction learning (ECADCL). The proposed method uses a feature extraction module to extract features from the input image and its corrupted images, and adversarial learning is used to avoid learning noise from corrupted images while extracting semantic features by reconstructing the corrupted images. Finally, a decision module predicts the label based on the learned features. Moreover, the feature extraction module combines a local cross-channel interaction attention mechanism without dimensionality reduction to characterize representative information. A milling dataset is conducted based on the machined surface images for monitoring tool wear conditions. The experimental results indicated that the proposed system can effectively assess the wear state of the tool.

Tool wear status is crucial for maintaining high workpiece quality and enhancing productivity. Numerous deep learning-based approaches have been developed to monitor tool wear. However, a single deep learning model often face limitations in feature extraction. To address this issue, we propose a hybrid global and local neural network with lightweight multiscale information fusion for tool wear prediction task. Firstly, Stockwell transform is applied to comprehensively extract time-frequency features from raw signals. Next, the lightweight multiscale feature extraction module arranges spatial separable convolution blocks with different receptive fields in parallel. This design allows the model to fully exploit spatial features at multiple scales to enhance dynamic signal sensitivity. Subsequently, a hybrid global and local feature extraction module comprises two parallel branches, one focused on extracting fine-grained local detail and the other on modeling global long-range dependencies. Finally, a novel orthogonal attention feature fusion strategy is proposed to enable effective interactions among features, and capture comprehensive highlevel representation of wear-related characteristics. To validate the superior predictive capability of proposed model, a milling experiment was designed and conducted. The experiment results demonstrate that the proposed model achieves an average root mean square error of 9.369 and a coefficient of determination of 0.907. Moreover, ablation experiments demonstrated the effectiveness of the proposed multiscale feature extraction module and hybrid global and local architecture.