The rapid development of modern industry has significantly raised the demand for workpieces. To ensure the quality of workpieces, workpiece surface defect detection has become an indispensable part of industrial production. Most workpiece surface defect detection technologies rely on cloud computing. However, transmitting large volumes of data via wireless networks places substantial computational burdens on cloud servers, significantly reducing defect detection speed. Therefore, to enable efficient and precise detection, this paper proposes a workpiece surface defect detection method based on YOLOv11 and edge computing. First, the NEU-DET dataset was expanded using random flipping, cropping, and the self-attention generative adversarial network (SA-GAN). Then, the accuracy indicators of the YOLOv7–YOLOv11 models were compared on NEU-DET and validated on the Tianchi aluminium profile surface defect dataset. Finally, the cloud-based YOLOv11 model, which achieved the highest accuracy, was converted to the edge-based YOLOv11-RKNN model and deployed on the RK3568 edge device to improve the detection speed. Results indicate that YOLOv11 with SA-GAN achieved mAP@0.5 improvements of 7.7%, 3.1%, 5.9%, and 7.0% over YOLOv7, YOLOv8, YOLOv9, and YOLOv10, respectively, on the NEU-DET dataset. Moreover, YOLOv11 with SA-GAN achieved an 87.0% mAP@0.5 on the Tianchi aluminium profile surface defect dataset, outperforming the other models again. This verifies the generalisability of the YOLOv11 model. Additionally, quantising and deploying YOLOv11 on the edge device reduced its size from 10,156 kB to 4,194 kB and reduced its single-image detection time from 52.1ms to 33.6ms, which represents a significant efficiency enhancement.

In modern manufacturing, cutting tools are essential for cutting processes, and their wear state directly affects the processing accuracy, production efficiency, and product quality. Identification of the tool-wear state using a single sensor is insufficient to satisfy the requirements of high-precision, high-efficiency machining. To address this problem, this paper proposes a novel approach to identify the tool-wear state using information fusion technology and the sparrow search algorithm (SSA)–backpropagation (BP) neural network framework. This method uses a principal component analysis (PCA) to fuse multi-domain features extracted from three-way vibration signals, power signals, and temperature signals. Subsequently, the optimal initial threshold and weight of the BP neural network are optimized using the SSA to prevent the network from falling into the local optimum and accelerate the convergence of the algorithm. Lastly, a tool-wear-state identification model based on the SSA–BP neural network is constructed. Experimental results show that the proposed method has an identification accuracy of 98.33%, precision rate of 98.81%, recall rate of 97.96%, and F1 score of 98.36%.

Integer-order models cannot characterize the dynamic behavior of the flexible two-link manipulator (FTLM) system accurately due to its viscoelastic characteristics and flexible oscillation. Hence, this paper proposes a fractional-order modeling method and identification algorithm for the FTLM system. Firstly, we exploit the memory and history-dependent properties of fractional calculus to describe the flexible link’s viscoelastic potential energy and viscous friction. Secondly, we establish a fractional-order differential equation for the flexible link based on the fractional-order Euler–Lagrange equation to characterize the flexible oscillation process accurately. Accordingly, we derive the fractional-order model of the FTLM system by analyzing the motor–link coupling as well as the symmetry of the system structure. Additionally, a system identification algorithm based on the multi-innovation integration operational matrix (MIOM) is proposed. The multi-innovation technique is combined with the least-squares algorithm to solve the operational matrix and achieve accurate system identification. Finally, experiments based on actual data are conducted to verify the effectiveness of the proposed modeling method and identification algorithm. The results show that the MIOM algorithm can improve system identification accuracy and that the fractional-order model can describe the dynamic behavior of the FTLM system more accurately than the integer-order model.

In mechanical manufacturing, tool wear directly affects machining quality and production efficiency. This study proposes a tool-wear-state identification method that combines continuous wavelet transform (CWT), coordinate-attention-weighted CWT (CA-CWT), a convolutional neural network (CNN), spatial and channel synergistic attention (SCSA), and a long short-term memory (LSTM) layer. The vibration signal from each sensor axis is divided into one-second windows and converted into 64 × 64 grayscale time–frequency images. In the implemented network, the CNN output has shape B × 128 × 8 × 8, is flattened to B × 8192, and is supplied to the LSTM as B × 1 × 8192. Because the sequence length is one, the LSTM is interpreted as a gated nonlinear feature-refinement layer rather than as a multi-step temporal model. On the fixed image-level validation split, the representative CA-CWT CNN-LSTM-SCSA run achieved 98.15% accuracy, 98.24% macro precision, 98.13% macro recall, and 98.16% macro F 1-score. Across 24 archived runs using the same split and configuration, the corresponding mean ± standard deviation was 97.92 ± 1.35%, 97.98 ± 1.29%, 97.92 ± 1.35%, and 97.92 ± 1.34%, respectively. These results characterize within-tool, fixed-condition image classification and do not establish tool-independent or condition-independent generalization.