Bayesian Nonparametric Methods in Textile Spinning:A Comprehensive Framework for Modeling Comparing Parameters in Quality Control

This paper introduces Bayesian nonparametric methods as a flexible and adaptive framework for quality control in textile spinning. Unlike traditional parametric models, BNP approaches allow model complexity to grow with data, capturing evolving process conditions. The study integrates Dirichlet Processes, Gaussian Processes, and Poisson-Gamma mixtures to model fibre properties, yarn evenness, and yarn fault counts. Numerical analyses demonstrate how BNP methods detect anomalies—such as deviations in fibre length, unevenness, and fault counts—providing early-warning signals of process degradation. The framework enhances process monitoring, predictive maintenance, and decision support, ultimately improving product consistency and operational efficiency. Limitations include computational demands, hyperparameter sensitivity, and interpretability challenges.