Xin Ru

High-value recycling of waste textiles requires accurate fiber-composition identification, yet similar blends often exhibit overlapping near-infrared absorption bands and substantial within-class variation. Using 291 waste-textile swatches, this study develops ConvPatchTST, a convolution-enhanced Patch Transformer that integrates Raw and SNV (SG) spectra as complementary input channels. Overlapping convolutional patch embedding captures local continuous absorption features, while a Transformer encoder models cross-band dependencies. Spectral bands spanning 958.5–1646.6 nm were selected for model training and analysis. Under a single fabric-swatch-grouped internal train-validation split with pixel-level evaluation, ConvPatchTST achieves an accuracy of 0.966 and a macro F1 of 0.970, outperforming several baseline models. The Raw + SNV (SG) representation also surpasses single-channel, three-channel, and alternative dual-channel inputs. Pairwise analysis identifies statistically discriminative and highly attributed intervals for four similar material pairs. For polyester/polysp, the intervals near 1369.0–1379.6 nm show clear statistical-model agreement. These results indicate that complementary spectral representations and convolution-enhanced Patch modeling support accurate classification and provide band-level clues for interpreting model decisions. Generalization to external samples, batches, and instruments requires further validation.

Conventional overload verification of bimetallic protection elements requires long-duration current loading and considerable time and energy consumption. This study proposes a rapid verification method combining a first-order thermodynamic equivalence model with image-based deflection measurement. Local linear interpolation between adjacent measured frames was used to determine the energization time corresponding to a prescribed target deflection. A current-dependent empirical correction factor was introduced to compensate for systematic deviations caused by simplified lumped thermal parameters and test-specific thermal boundary conditions. The model was identified using one target-deflection dataset and evaluated at two additional target deflections without parameter re-identification. Repeated tests were also conducted to assess cycle-to-cycle stability. The results show that the corrected model substantially reduced the temperature-rise deviation, maintained good prediction consistency across different target deflections, and exhibited low cycle-to-cycle variability. The method provides a physics-informed semi-empirical framework for rapid evaluation of bimetallic-strip overload characteristics. Further re-identification and assembled-product validation are required when the specimen configuration or thermal boundary conditions change.

In recent years, vortex spinning has garnered significant attention owing to its high efficiency and superior yarn quality. However, the drafting process involves multiple interrelated parameters, and different combinations of parameters can considerably influence subsequent spinning performance. To address this, the present study introduces a novel hybrid optimization algorithm to enhance spinning quality by rationalizing the coordination of drafting parameters. First, orthogonal experiments were conducted with the draft ratio and roller center distance as variables, using the mean grayscale value and grayscale standard deviation of the post-experiment silver images as multi-objective functions to evaluate drafting effectiveness. Subsequently, a regression model between drafting parameters and drafting outcomes was constructed using the Particle Swarm Optimization–Backpropagation Neural Network (PSO-BP) algorithm, followed by multi-objective optimization via the Non-dominated Sorting Genetic Algorithm II (NSGA-II) genetic algorithm to obtain a Pareto-optimal solution set. Finally, the entropy-weighted Technique for Order of Preference by Similarity to Ideal Solution (TOPSIS) method was applied to comprehensively evaluate the Pareto-optimal set and determine the optimal combination of process parameters. The results demonstrate that, under the optimal parameter combination, the deviation between the measured quality indicators of the drafted sliver and the predicted values remains within 6%, confirming the effectiveness of the proposed model as a viable approach for optimizing drafting parameter configurations.

Motor starter calibration requires long-distance rotation of a setting cam to locate current graduation points, generating substantial non-value-added mechanical travel time on production lines. This paper proposes a cam pre-adjustment angle prediction method that integrates a phenomenological gray-box bimetallic model with a hierarchical combinatorial algorithm framework. A generalized lumped-parameter model incorporating heat dissipation correction and mechanical gap compensation is constructed to describe the electrothermal–mechanical coupling of the bimetallic strip. An improved fuzzy C-means (IFCM) algorithm addresses the cold-start problem for new material batches, and an adaptive particle swarm optimization (APSO) algorithm performs online parameter identification. To handle the process asymmetry arising from the unidirectional cam rotation mechanism, an optimized gray wolf optimizer with one-sided error control (GWO-OSE) based on an asymmetric loss function is employed to inversely determine the optimal pre-adjustment angle while actively suppressing over-prediction. Validation on 1200 production line samples across three material batches demonstrates an over-prediction rate of only 2.8%, a mean absolute angle prediction error of 23.9°, a reduction in single-product calibration time of approximately 12 s, and an improvement in overall production line efficiency of 24.5%. This efficiency gain results from the process-level redesign facilitated by the pre-adjustment strategy rather than from minimizing absolute prediction error, and the proposed method provides an engineering-applicable optimization strategy for reducing non-value-added calibration time in motor starter production lines.

Extracting standardized digital design patterns from real knitted fabric images is critical for textile reverse engineering and digital archiving. Unlike smooth graphics, knitted fabrics exhibit high-frequency textures from yarn loop interlacing, introducing significant grayscale variations within same-color regions. Existing algorithms struggle to distinguish these from pattern edges, causing color quantization and segmentation failures. To suppress yarn texture while preserving edges between color blocks, we propose an adaptive pattern extraction method using Bayesian-optimized bilateral filtering. The primary contribution lies in providing a domain-specific, application-focused integrated framework. Specifically, (1) a knitting-texture-aware multidimensional evaluation parameter is constructed by integrating physical-cause-based texture features (gray-level co-occurrence matrix (GLCM) contrast, homogeneity, and Laplacian variance) with perception-based edge preservation metrics (the Sobel operator and the structural similarity index (SSIM)), enabling accurate discrimination between yarn-level texture noise and pattern-level color block boundaries—a distinction that generic image quality metrics cannot make. (2) Then, this domain-specific objective function is embedded within a Bayesian optimization framework to achieve automatic, zero-shot, per-image parameter adaptation across different knitting processes, without requiring any external training data. K-means color quantization maps in continuous tones to discrete classes, generating standardized patterns meeting knitting requirements. Experiments on 316 samples covering six processes show our method outperforms standard denoising and advanced algorithms like relative total variation (RTV), achieving an average SSIM of 0.83 and PSNR of 26.92 dB, reducing processing time from 15–30 min to 21 s per image, providing efficient automation for knitted Computer-Aided Design (CAD) systems.

A method for simultaneous classification of fabric material and color based on hyperspectral imaging and visual detection is proposed. Fabric material classification is performed using hyperspectral imaging (HSI) combined with a one-dimensional convolutional neural network (1D-CNN), while fabric color recognition is achieved using an red-green-blue (RGB) camera and a color classification model. Material and color features from the same fabric sample are matched to realize synchronous classification. Experiments were conducted on three fabric materials (cotton, polyester, and cotton–polyester blend) and eight colors. At a conveyor speed of 1 m/s, the sorting success rates reach 95.0% for cotton, 97.5% for polyester, and 85.0% for cotton–polyester blended fabrics. The proposed method demonstrates reliable performance for single-material fabrics and good industrial applicability for automated fabric sorting.

Traditional custom-made socks primarily focus on ensuring the accuracy of key metrics such as foot length and height, failing to address the demand for more intricate customization. The human foot model encapsulates precise data essential for designing custom socks. This study proposes a methodology to derive accurate process parameters based on user-specific foot models and input them into sock-knitting machinery to produce customized socks. Initially, this study introduces an algorithm designed to generate customized sock models derived from foot models. The proposed algorithm leverages the alpha wrapping technique, augmented by grid smoothing and segmentation methods, to construct the wrapping model of the foot. By analyzing the customized socks, this study introduces fit and style indicators to assess the generated wrapping model. Based on these metrics, an evaluation function for the simulated annealing algorithm is constructed, enabling automated adjustment of alpha wrapping algorithm parameters to create customized sock models that meet predefined criteria. The system subsequently simulates the knitting process by iteratively sampling the model’s surface, treating sampling points as coils to progressively generate knitting paths aligned with the model’s geometric attributes. These paths are then converted into machine-knitting process parameters. Finally, the process parameters were input into a sock machine to fabricate a knitted sock. A comparison of the sample’s dimensions with the foot model revealed an error margin below 6.5%. In addition, density changes between the flat state and the wearing state of the knitted sock averaged under 7.7%, affirming the method’s validity.

Texture mapping of weft-knitted fabrics plays a crucial role in virtual try-on and digital textile design due to its computational efficiency and real-time performance. However, traditional texture mapping techniques typically adapt pre-generated textures to deformed surfaces through geometric transformations. These methods overlook the complex variations in yarn length, thickness, and loop morphology during stretching, often resulting in visual distortions. To overcome these limitations, we propose Knit-Pix2Pix, a dedicated framework for generating realistic weft-knitted fabric textures directly from knitted unit mesh maps. These maps provide grid-based representations where each cell corresponds to a physical loop region, capturing its deformation state. Knit-Pix2Pix is an integrated architecture that combines a multi-scale feature extraction module, a grid-guided attention mechanism, and a multi-scale discriminator. Together, these components address the multi-scale and deformation-aware requirements of this task. To validate our approach, we constructed a dataset of over 2000 pairs of fabric stretching images and corresponding knitted unit mesh maps, with further testing using spring-mass fabric simulation. Experiments show that, compared with traditional texture mapping methods, SSIM increased by 21.8%, PSNR by 20.9%, and LPIPS decreased by 24.3%. This integrated approach provides a practical solution for meeting the requirements of digital textile design.

The motion anomalies of needle-selection blades in electromagnetic needle selectors are subtle and rare. Limited by insufficient observation, feature extraction difficulty, and severe class imbalance due to scarce fault samples, data-driven models suffer in training and generalization. To address this, we propose a physics-guided anomaly detection method based on visually captured trajectory data. A system dynamics model generates physically consistent virtual anomaly samples to mitigate data imbalance, while normal-data pre-training enhances rare-fault recognition. Building on physics-informed data augmentation, the proposed enhanced weighted random forest achieves 97% in WAP, WAR, and WAF, with F1-scores of 0.84 for two rare anomaly classes and recall improved to 92% and 80%, outperforming mainstream data augmentation and classification methods, enabling efficient rare-fault detection under small-sample conditions.