Improvement of the cotton fiber length measurements using High Volume Instrument (HVI) fibrogram
MetadataShow full item record
Within-sample variation in cotton fiber length is an important parameter to consider when explaining variation in yarn quality. However, the most widely used cotton fiber length parameters, Upper Half Mean Length (UHML) and Uniformity Index (UI), provided by the High Volume Instrument (HVI), do not characterize the total within-sample variation in fiber length. HVI fiber length measurements are based on the fibrogram principle where a fiber beard is scanned over a beam of red light from 3.81 mm away of its base to the tip and a detector is used to measure the amount of light attenuated across the fiber beard which generates a curve called the fibrogram. Fibrogram curves are not reported in regular HVI output. However, it can be exported from native HVI software as vector graphic images when samples are measured in module testing mode. A method was developed to extract the data that were used to generate the fibrogram curve. This allows statistical analysis to be performed on the whole fibrogram curve, without loss of information. Initial investigation of UHML and UI was conducted using 19,628 commercial bales. Obtained results reveal that the typical HVI length measurements are not characterizing unique types of length variation in the fiber beard. Fibrogram measurements taken from a subset of 538 of commercial samples were used to identify independent types of fiber length variation characterized by the fibrogram that are not currently being characterized by UHML and UI. The results obtained suggest that the HVI fibrogram does capture additional within-sample variation in fiber length that is not being currently reported. Two additional sets of samples were then used to evaluate the importance of this currently unused information about length variation. Partial Least Square Regression (PLSR) models were used to determine the importance of this new information as a tool for explaining variation in yarn quality. The four PLSR models were designed as follows: the first model contains only non-length HVI parameters, the second model contains the current HVI length parameters along with all the non-length HVI parameters, the third model contains fiber length variation captured by the fibrogram along with non-length HVI parameters and the fourth model contains fiber length variation captured by the Advanced Fiber Information System (AFIS, length by number) along with non-length HVI parameters. The results presented here suggest that the additional variation captured by the fibrogram provides better yarn quality prediction (Model 3) than current HVI length parameters (Model 2) and are comparable to the results obtained using the AFIS length distribution by number (Model 4). The PLSR models were then validated using a leave-one-out cross-validation and they show that the models built with information extracted from the fibrogram are better at predicting yarn quality than models with the most commonly used HVI length parameters. These results suggest that the information from the fibrogram is at least as good as the AFIS length distribution by number when characterizing variation in yarn quality. While the additional information provided by the whole fibrogram could provide a new tool to breeders for selecting their breeding lines and spinners for purchasing cotton bales, fibrogram measurements are not calibrated to be used across cotton industry. In this study, a method of correcting the whole fibrogram curve across HVIs using a set of 529 commercial samples was developed. Validation of the correction procedure was also investigated by using an independent set of 932 commercial samples. The results obtained show that it is possible to bring the fibrogram curves measured by multiple instruments into an agreement through multivariate correction. It indicates that it may be possible to use the whole fibrogram to improve HVI fiber length measurements across the cotton industry.