Transformer-Based Multi-Stage Quality Prediction in Cotton Spinning

The cotton spinning process encompasses multiple interconnected stages where raw cotton properties fundamentally influence downstream product quality. This study addresses the complexities inherent in multi-stage quality prediction by developing a novel deep learning framework that synergistically combines Transformer-based feature extraction with multilayer perceptron prediction capabilities. The proposed model effectively integrates static cotton-blending parameters with dynamic process indicators captured across pre-spinning, spinning, and winding operations. Comprehensive experimental validation demonstrates the framework’s superior performance in the classification task of internal quality level prediction, substantially outperforming conventional machine learning approaches in terms of precision and mean squared error metrics. This research contributes a scalable and interpretable methodology for advancing intelligent quality control systems within the cotton spinning sector.