Accurate prediction of cotton ring-spun yarn quality from high-volume instrument and mill processing data

Abstract

The derivation and performance of yarn quality prediction models in a program called Cottonspec is reported. Cottonspec incorporates a large database of fiber and yarn data from commercial spinning mills, a series of regression-based models predicting yarn quality from measured cotton fiber quality parameters and yarn specifications and a user interface. The inclusion of independent variables into prediction equations was dependent on the criteria that their inclusion was statistically significant and that variables had a theoretically direct influence on yarn structure. Yarn data was corrected to allow for twist, yarn count and yarn irregularity before correlation with fiber properties. Differences in yarn testing results between mills could be corrected by a Mill Correction Factor. Adherence to these criteria and the ability to draw on the very large database meant prediction ability of the models was excellent, as demonstrated in a series of cross-validations.