ARS | Publication request: USING PARTIAL LEAST SQUARES REGRESSION TO OBTAIN COTTON FIBER LENGTH DISTRIBUTIONS FROM THE BEARD TESTING METHOD

Submitted to: National Cotton Council Beltwide Cotton Conference
Publication Type: Proceedings
Publication Acceptance Date: February 8, 2010
Publication Date: April 28, 2010
Citation: Cai, Y., Cui, X., Belmasrour, R., Li, L., Delhom, C.D., Rodgers III, J.E., Martin, V., Watson, M. 2010. USING PARTIAL LEAST SQUARES REGRESSION TO OBTAIN COTTON FIBER LENGTH DISTRIBUTIONS FROM THE BEARD TESTING METHOD. Proceeding of the 2010 National Cotton Council Beltwide Cotton Conference, January 5-7, 2010, New Orleans, Louisiana. p. 1411-1413. 2010 CDROM.

Interpretive Summary: The beard testing method can rapidly determine certain fiber length parameters, such as Upper Half Mean Length (UHML) and Uniformity Index (UI). Cotton is a natural product and its length has a wide variation. Knowing the length distribution enalbles a better assement of the cotton quality. In the instrumental implementations, the engineering complexity alters the original fiber length distribution observed by the instrument, and this causes challenges in obtaining the entire original length distribution of a cotton sample by the beard method. For the purpose of improving cotton fiber length measurements and expanding the application of the test results, new approaches were explored to obtain the original length distribution from the rapid beard testing method. A new method (Partial Least Squares Regression) was applied to study the relationship between the observed length distribution in a fiber beard and the original length distribution. The new approach showed promising potential to estimate the original length distribution from the observed length distribution of a fiber beard.

Technical Abstract: The beard testing method for measuring cotton fiber length is based on the fibrogram theory. However, in the instrumental implementations, the engineering complexity alters the original fiber length distribution observed by the instrument. This causes challenges in obtaining the entire original length distribution of a cotton sample. For the purpose of improving cotton fiber length measurements and expanding the application of the test results, new approaches were explored to obtain the original length distribution from the rapid beard testing method. The Partial Least Squares Regression algorithm was applied to study the relationship between the observed length distribution in a fiber beard and the original length distribution. The new approach showed promising potential to estimate the original length distribution from the observed length distribution of a fiber beard.