Multivariate analysis of fiber properties and their relation to yarn properties

Date
2014-12
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract

Natural variation in cotton quality within and between samples impacts the performance of cotton fiber in an industrial setting. Industrial processes used to transform the fiber from field to fabric have the potential to break fiber and degrade fiber quality. A set of 37 commercial cotton bales exhibiting a wide range in fiber length and maturity are used to investigate the impact of fiber breakage on the within sample distribution of fiber quality. A multivariate partitioning technique is used to demonstrate the AFIS individualizer has a significant impact on the within sample distribution of fiber quality. The impact is then characterized in terms of how it alters the within sample distribution in fiber quality. The AFIS individualizer is found to preferentially break longer fibers in the sample, likely due to the structure of the fiber individualizer mechanism. Variation in within sample fiber quality imparted by the AFIS individualizer is used to model the breakage process. The impact of future stages of processing are predicted with the model. AFIS must first individualize the fibers from a sliver in order to characterize the within sample distribution of fiber quality. The breakage model provides an estimate of the within sample distribution of fiber quality prior to AFIS fiber individualization. This technique can be extended to model breakage experienced in other types of processing. Understanding the relationship between fiber and yarn quality is critical at many junctures in the cotton industry. Yarn is spun from 110 commercial bales representing a wide range of fiber quality. Information statistics are used to construct a multimodel composed of all possible models formed by combinations of 15 AFIS and HVI fiber quality parameters. Multimodel inference based on information theory is used to compare the set of all possible yarn quality models composed of bundle measurements to those composed of fiber quality parameters based on the within sample distribution in fiber quality. Models constructed of parameters based on the complete within sample distribution in fiber quality are better supported by the observed data. The within sample distribution of fiber quality holds important information needed for understanding the multivariate relationship of fiber and yarn quality.

Description
Keywords
Cotton, Fiber quality
Citation