Location: Cotton Structure and Quality Research
2012 Annual Report
2. Calculating the length distribution of the original fibers. We will convert the distributions of the scanned projecting portions of the beards to compute the entire length distributions of the original fibers. These original length distributions are obtained from AFIS tests. The conversion can be achieved by implementing the relationship between the distributions (pdf) of HVI beards and those of the original fibers we developed in earlier stages, such as the application of PLS regression algorithm.
3. Verification After deriving the entire length distributions of original fibers from scanning the beards, we then will verify our results by comparing them to AFIS measured fiber length distributions of the original fibers. Based on the verification we will make proper adjustments to our algorithms and models. For this purpose, we are carrying out a larger set of HVI and AFIS tests. AFIS data have been used as references in this study.
4. Implementation As a result, we can provide the industry a method that enables the HVI to obtain the entire length distributions of the original cotton fibers. Implementation of our results includes providing the industry algorithms and equations for computing fiber length distributions and length parameters such as Lower Half Mean Length and Short Fiber Content.
ARS scintists at SRRC in New Orleans, LA completed more samples to expand the database of High Volume Instrument (HVI) and AFIS length distributions, which include distributions constructed from AFIS data of individual fiber lengths. For each cotton sample, a set of AFIS length data includes the original fibers before HVI testing, the fibers taken from the HVI fiber beards, the fibers cut from the projecting portion, and the fibers of the hidden portion, as well as data from HVI tests. A larger set of data is essential for validating and refining the algorithm for processing experimental data from HVI signals, the staple diagram model, and the Partial Least Squares Regression (PLS) method. Computer programs have been developed to handle these data.
ARS scientists at SRRC in New Orleans, LA used the five-parameter mixed Weibull function to model the probability density functions of cotton length distributions. We developed the PLS method to compute the set of five parameters of the original beard from that of the projecting portion. We applied this technique on the new experimental data and obtained results of length density functions and length parameters computed from them. Both the above mentioned approaches can be very helpful for developing industrial implementations to enhance the rapid beard testing method.
Cotton fiber length parameters represent different characteristics of cotton length distributions. These parameters’ impacts on yarn properties are different. We investigated the effects of a set of cotton fiber length parameters on yarn properties. Linear regression models involving different number of fiber length parameters and their combinations were developed to predict ring and Open End (OE) spun yarns’ properties, including strength, irregularity, thick places, thin places, neps, ends down, and elongation. Statistical analysis results revealed that different combinations of length parameters are needed to produce the best yarn property models and that the variations of length distributions play a very important role in predicting yarn properties. This emphasizes the importance of obtaining and utilizing the entire length distributions of cotton samples, which enables the computation of different length parameters including variation of length distribution.
The methods used to monitor activities for this agreement were annual reports, technical visits/e-mails/interactions, presentations at scientific and industry meetings, and publications.