Nana Li

Industrial visual anomaly detection is crucial for ensuring system reliability, safety, and production efficiency. However, most existing methods rely on static image modeling and ignore the temporal characteristics of industrial data, making it difficult to capture subtle and dynamically evolving anomalies in complex environments. To address this issue, we propose an online industrial visual anomaly detection method based on temporal aggregation representations. This method first converts consecutive image frames into temporally ordered spatiotemporal representations using a dimension-transformation-based temporal aggregation strategy, thereby suppressing factory environmental noise and enhancing subtle anomalous features. Subsequently, a Multiscale Wavelet-based Salient Attention (MWSA) module is designed to enhance the discriminability of subtle anomalous features through low-frequency structural enhancement and high-frequency noise suppression. Building upon this, a memory bank of normal features is constructed, and anomaly detection and localization are achieved based on nearest-neighbor distances. Experimental results demonstrate that the proposed framework achieves high-precision visual anomaly detection and localization, while maintaining stable performance under varying camera conditions and complex industrial environments. Actual deployment results demonstrate that the method exhibits excellent environmental robustness, achieving an event-level recall rate of 94.83\% and effectively reducing false triggers caused by complex background interference, thereby providing an efficient and robust solution for quality control in industrial automation.

Fabric defect detection faces significant challenges in identifying tiny defects and those with low contrast against complex background textures. Existing methods struggle with irregular shapes and texture interference. This paper proposes TEAAM-YOLOv8, an improved fabric defect detection framework based on YOLOv8 with three innovative modules. The Texture Enhanced Adaptive Attention Mechanism (TEAAM) in the backbone captures subtle defect features against similar backgrounds. The Adaptive Spatial Fusion Module (ASF) in the neck efficiently integrates multi-scale features for tiny defect detection. The Geometric Shape-aware IoU Loss (ShapeIoU) optimizes boundary fitting for irregular defects. Experimental results demonstrate significant improvements: mAP@0.5 reaches 96.0% (+3.8%), recall achieves 92.9% (+9.5%). The method maintains real-time performance with only 1.7% parameter increase and 7.4% computational overhead. This approach provides an effective solution for intelligent fabric quality inspection, combining high accuracy with high-speed processing.

Polymer-fiber weft-knitted textiles are widely used in functional apparel and moisture-management materials, but their complex loop topology, yarn-level porosity, and interlayer hierarchy make simulation-oriented geometric modeling challenging. Conventional control-point and interpolated-curve methods often have limitations in representing knitting actions, maintaining yarn-path continuity, and generating meshable geometries. This study proposes a rule-driven geometric modeling method for polymer-fiber weft-knitted structures using the yarn centerline as the basic geometric carrier. Knitting actions, including knit, tuck, float, plating, and double-needle-bed assignment, are converted into reusable local path-generation rules and integrated through pattern-matrix input, action recognition, parametric centerline generation, continuous stitching, and standardized output. The method represents single-bed and double-bed structures within a unified framework, including plain, jacquard, plated, tuck, rib, interlock, half-cardigan, full-cardigan, and purl structures. Compared with an interpolated-curve method, the curvature-jump rate of four representative structures decreases from 40.78–69.23% to 0–0.31%, with markedly reduced maximum bending angles. Mesh-generation results show continuous meshes with improved element quality for complex double-bed structures. A moisture-transfer simulation of a fully plated plain-knitted structure gives one-way transport indices of −122.3187 and 122.2472 for face- and back-side liquid entry, with relative errors of 1.74% and 0.50% compared with experiments. These results indicate that the proposed method provides reproducible and meshable geometric input for structure–property modeling and moisture-transfer prediction of polymer-fiber knitted textiles.

A bare polytetrafluoroethylene (PTFE) membrane has poor antifouling and antibacterial performance in the application process. In this work, the silver nanoparticle layer (AgNP-layer) was deposited onto the PTFE membrane surface via the magnetron sputtering process to overcome these limitations. Several spectra and microscopic analyses were taken to characterize the synthesized PTFE/AgNP-layer composite membranes. The effects of magnetron sputtering time on the AgNP-layer distribution, microstructure, antifouling, and antibacterial properties were systematically investigated. Our results showed the successful deposition and grafting of the AgNP-layer on the PTFE membrane surface and inside the cross-sectional pores near the membrane surface. The AgNP-layer not only improved the composite membrane wettability but also reduced its protein adsorption capacity from 88 to 39 μg, thus improving the antifouling ability of the membrane. The composite membranes could also achieve 99% and 88% antibacterial rates against Escherichia coli and Staphylococcus aureus, respectively, exhibiting good antibacterial activity.

It is widely recognized that polyester (PET) fibers with swelling properties are necessary for advanced applications of PET fibers in waterproof fabrics. This study presented the UV‐initiated methods to graft polyacrylamide (PAM) hydrogel on the surface of PET fibers. In the one‐step method, PET fibers are impregnated with a mixed solution of the photoinitiator benzophenone (BP) and the monomer acrylalkylamine (AM). Then, the PET fiber was directly exposed to UV light. In the two‐step method, PET fibers were first impregnated with a photoinitiator solution to generate BP radicals. Then, the fibers impregnated with a monomer solution before being grafted by UV light. Both two methods ultimately achieve PAM gelation‐modified PET fibers. The chemical composition and surface structure were detected by Fourier transform infrared spectroscopy and scanning electron microscope. The findings demonstrate that the PAM hydrogel was successfully grafted on the PET fibers through one‐step method or a two‐step method. In addition, the properties of the grafted PET fibers were evaluated by grafting yield, contact angle, and swelling test. The study indicated that the modified PET fibers via a two‐step method exhibited a superior grafting yield up to 29.3% (PET‐2n) and 54% (PET‐2) compared to the one‐step technique (PET‐1). It was shown that the grafting efficacy of PET fibers modified by two‐step method is better than one‐step method. The static contact angles of PET fibers by UV‐Induced were all reduced. For instance, the contact angles of PET‐2n were decreased from 115° to 62.9°; and the two‐step modified PET fiber after acetone cleaning (PET‐2) decreased from 115° to 60°. It was suggested that the hydrophilicity of the modified PET fibers significantly increased. In addition, the research shows that the gelation‐modified fibers were all possessed swelling properties.