Merits and Limitations of Cotton Fiber Length Measuring Instruments (Part3)
N.Balasubramanian
Retd Jt. Director (BTRA) and Consultant
I, Rajeswari, 36, 17th Road, Chembur, Mumbai 400071, 9869716298
Email: balajamuna@gmail.com
Retd Jt. Director (BTRA) and Consultant
I, Rajeswari, 36, 17th Road, Chembur, Mumbai 400071, 9869716298
Email: balajamuna@gmail.com
Previous Part
AFIS (Advanced Fibre Information System)
Fibre samples are opened into individual fibres by feed plate/feed roller and the opened material is aerodynamically presented to electrooptical sensors for measuring length apart from other properties. Apart from mean, distribution is also determined. There are no USDA calibration cottons for AFIS but Uster gives calibration cottons.
Merits
 Numerical sample is used as against length biased sample in HVI
 No need to prepare an array as in Baer or Suter web sorter and Almeter. Inaccuracy in fibre alignment and density in array preparation is avoided.
 As both length and diameter are measured, both numerical and weight based distribution can be determined.
 Fairly quick as 10000 fibres can be measured in a few minutes
 Since fibre breakages occur with longer fibres, length measured is lower.
 Further fibres are not straight which leads to under estimation of length.
Fibres manually placed on glass slide are photographed using CCD camera and suitable software is used to make image processing and measurements. Ikiz15 et al study the effect of lighting, resolution, pre processing and processing algorithms on the accuracy of results. Accuracy of the result is improved with high resolution images. Image processing is more accurate and has higher precision than hand measurement, HVI and AFIS.Y. Xu16 et al developed a method for determining fibre length distribution by preparing an aligned combed tuft and cutting it into number of segments of a known length. The snippets in different segments were scanned by image analysis to get the number of fibres.
Almeter and AFIS give results that are better correlated with sorter results.
Length Biased Mean
The sample from which Baer sorter diagram is prepared is a numerical sample where all fibres have the same probability of occurrence. But the sample in the beard of Fibrograph is a length biased sample (as per Hertel), as probability of fibre being caught is assumed to be proportional to fibre length. If a sliver is clamped across a cross section and loose fibres not held by clamp are combed out, tuft held under the clamp is a length biased sample. Length biased sample frequency can be obtained from the numerical sample by multiplying frequency in each class by the length of fibre. If the fibres in different length groups are weighed as in balls sorter and Suter web sorter, length biased sample will be obtained. Frequency for length in a length biased sample is given by l f(l).
Where,
lm = Length biased mean
ī = Normal mean
CV = Coefficient of variation of fibre length.
In Table 6 Mean, CV and short fibre content of normal and length biased sample of cotton of an assumed frequency distribution are compared.
Table 6 : Comparison of length properties of normal and length biased distributions
ī = Normal mean
CV = Coefficient of variation of fibre length.
In Table 6 Mean, CV and short fibre content of normal and length biased sample of cotton of an assumed frequency distribution are compared.
Table 6 : Comparison of length properties of normal and length biased distributions

Normal

Length biased

Mean mm

22.5

24.7

CV %

30.1 %

24.7 %

SFC

10.8 %

4.7 %

Length biased sample has a higher mean but lower SFC and CV.
Audivert and Casteller18 found varying degrees of correlation between span length by Digital fibrograph (SL) and 1. upper half mean length UHML, upper quartile length (UQL) and mean length from comb sorter and 2. Staple length by Shriley photoelectric stapler. Correlations approach optimum at 30 % SL for comb sorter and 2.5 to 10 % SL for Shirley photoelectric stapler. Jai Prakash19 found close agreement between 2.5 % span length from Digital fibrograph and mean length by Balls sorter with short staple cottons. With increase in length, 2.5 % length was progressively more than mean length by Balls sorter. SFL (< ½ inch) by digital fibrograph was highly correlated with SFL by ( < 8/16 inch)byBalls sorter. Ramsey and Beaton20 found a highly significant correlation between HVI Uniformity index (UI) and Comb sorter SFC with US upland standard cottons. But correlation between HVI UI and comb sorter is poor with commercial crops. Likewise HVI UI has a poor correlation with digital fibrograph UI. UI from HVI is equal to SFC in comb sorter in predicting yarn quality and process performance. UI from HVI has a good correlation with SFC by Almeter and SFC by Almeter gives a good prediction of processing performance and yarn strength. Bargeron21 found mean length by AFIS Almeter to be shorter than that of comb sorter by1.37 mm. SFC by number by Almeter was higher than that of comb sorter by 1.7 %. This may be because Almeter measures length with fibres in relaxed state.2.5 % and mean length of cotton are lower but SFC is higher in Almeter than Suter web sorter. This is attributed to the fact that fibres are straight with crimp removed in Suter while fibres are in relaxed state in Almeter. However high correlation is found between results of Almeter and comb sorter.
Short fibre content from HVI and AFIS not only differ considerably from that of Baer sorter but also bear no correlation with it. Bragg and Shofner22 incorporated a fibre speed sensor in AFIS to improve the accuracy. With this SFC levels by AFIS come closer to Suter web sorter, though they are still higher by about 8 – 10 units. On the other hand,Cui23 et al found significant differences in SFC between HVI, AFIS and Suter web sorter. Suter sorter gave highest value and AFIS lowest and HVI in between. This is in contradiction to the results of Bragg and Shofner22. Correlation between the instruments for SFC was lower than that for mean length. Calibration and sample non uniformity are mainly responsible for high variation in SFC. Thibodeauk24 et al also found SFC to be highest by Suter web sorter compared to HVI and AFIS. Further Suter web sorter is more accurate to detect difference in SFC between cottons. Higher correlation is found between SFC by HVI and Suter web sorter than that between SFC by AFIS and Suter web sorter. Zeidman et al review mathematical fundamentals on derivation of SFC by number and weight. SFC is related to range and shape of fibre length distribution25. Krifa26 found bimodal frequency length distribution in many cottons, as measured by AFIS and found that the extent of bimodality is correlated to fibre strength and maturity. Apart from the normal mode a second peak of lower amplitude is found at 34 mm. Stronger cottons exhibit bimodal distribution while weaker cottons exhibit unimodal distribution. Aggressive opening of the cotton makes it unimodal particularly with weaker cotton because of fibre breakages. Frequency distribution from Ball sorter does not, however, show bimodal distribution. This may be because it is weight length distribution as against number length distribution by AFIS. In a subsequent paper Krifa27 found that the length distribution is close to unimodel at both low and high breakage levels in ginning and processing in spinning i.e. ginned lint and card sliver. Mature cotton exhibits an extended intermediate stage of bimodal distribution compared to immature cottons because of lower breakage level.
Xu16 et al, who used image analysis, however did not find a peak at 3 mm. Landstreet28 derived the fibrogram from number length frequency distribution by double summation, Based on a reversal of this technique, Krowicki29 et al compared the fibrogram obtained from frequency length distribution by differential and algorithm methods. Algorithm method gives results close to differential method while taking less time. Equations for fibrogram from the fibre length distribution were determined by Azzouz30 et al assuming fibre length distribution to be normal. Good correlation is found between actual weight length frequency distribution and adjusted to normal distribution. However number length distribution deviates considerably from normal and the equation is not applicable. Cai31 et al found fibre length distribution plays an important role in prediction of yarn strength and irregularity and it is therefore useful to include fibre length distribution as an important quality parameter of cotton.
Krifa32 emphasized the need for taking fibre length distribution in selection of cotton. Group of bales selected on common HVI properties produces mixing with uncontrolled length distribution variability, caused by fibre breakages, from bales with same but low maturity. Shapiro33 et al discuss mathematical models to examine cotton fibre length distribution under various breakage models. Breakages on unclamped and clamped fibres were analysed. A procedure for testing the validity of model determining effect of breakage on length distribution is proposed.
Krowicki34 examined the effect of lens width on length in Digital fibrograph. Kelly35 et al found that selection in breeding program using HVI and AFIS data gave nearly the same order of improvement in fibre quality.
Hequet36 et al showed that yarn elongation can be predicted from HVI elongation and upper half mean length while yarn strength can be predicted from AFIS mean (by weight), fineness and maturity ratio.
Ureyan and Kadoglu37 found highly significant correlations between fibre properties measured by AFIS and yarn properties.
Cui14 et al showed that a mix of Weibull distributions gives a good fit to the actual fibre length distribution of original sample as well as of the sample picked by clamp of fibro sampler. Mean length and upper half mean from theoretical distribution shows good agreement with actual results.. Belmasrour38 et al proposed a method for estimating fibre length distribution of a sample from length distribution of fibres projecting from the comb of fibrograph. Predicted curve gives good agreement with experimental except in short fibre region.
References
Comparison of results from different instruments
Nair17 et al found that though 2.5 %length by Baer sorter, HVI and AFIS are correlated, the values by HVI are much lower. 2.5% span length by HVI is about 612 mm lower than that of baer sorter and the difference increases with fibre length. AFIS gives results close to Baer sorter with shorter cottons but gives about 24 lower values with longer cottons possibly because of fibre breakages. Audivert and Casteller18 found varying degrees of correlation between span length by Digital fibrograph (SL) and 1. upper half mean length UHML, upper quartile length (UQL) and mean length from comb sorter and 2. Staple length by Shriley photoelectric stapler. Correlations approach optimum at 30 % SL for comb sorter and 2.5 to 10 % SL for Shirley photoelectric stapler. Jai Prakash19 found close agreement between 2.5 % span length from Digital fibrograph and mean length by Balls sorter with short staple cottons. With increase in length, 2.5 % length was progressively more than mean length by Balls sorter. SFL (< ½ inch) by digital fibrograph was highly correlated with SFL by ( < 8/16 inch)byBalls sorter. Ramsey and Beaton20 found a highly significant correlation between HVI Uniformity index (UI) and Comb sorter SFC with US upland standard cottons. But correlation between HVI UI and comb sorter is poor with commercial crops. Likewise HVI UI has a poor correlation with digital fibrograph UI. UI from HVI is equal to SFC in comb sorter in predicting yarn quality and process performance. UI from HVI has a good correlation with SFC by Almeter and SFC by Almeter gives a good prediction of processing performance and yarn strength. Bargeron21 found mean length by AFIS Almeter to be shorter than that of comb sorter by1.37 mm. SFC by number by Almeter was higher than that of comb sorter by 1.7 %. This may be because Almeter measures length with fibres in relaxed state.2.5 % and mean length of cotton are lower but SFC is higher in Almeter than Suter web sorter. This is attributed to the fact that fibres are straight with crimp removed in Suter while fibres are in relaxed state in Almeter. However high correlation is found between results of Almeter and comb sorter.
Short fibre content from HVI and AFIS not only differ considerably from that of Baer sorter but also bear no correlation with it. Bragg and Shofner22 incorporated a fibre speed sensor in AFIS to improve the accuracy. With this SFC levels by AFIS come closer to Suter web sorter, though they are still higher by about 8 – 10 units. On the other hand,Cui23 et al found significant differences in SFC between HVI, AFIS and Suter web sorter. Suter sorter gave highest value and AFIS lowest and HVI in between. This is in contradiction to the results of Bragg and Shofner22. Correlation between the instruments for SFC was lower than that for mean length. Calibration and sample non uniformity are mainly responsible for high variation in SFC. Thibodeauk24 et al also found SFC to be highest by Suter web sorter compared to HVI and AFIS. Further Suter web sorter is more accurate to detect difference in SFC between cottons. Higher correlation is found between SFC by HVI and Suter web sorter than that between SFC by AFIS and Suter web sorter. Zeidman et al review mathematical fundamentals on derivation of SFC by number and weight. SFC is related to range and shape of fibre length distribution25. Krifa26 found bimodal frequency length distribution in many cottons, as measured by AFIS and found that the extent of bimodality is correlated to fibre strength and maturity. Apart from the normal mode a second peak of lower amplitude is found at 34 mm. Stronger cottons exhibit bimodal distribution while weaker cottons exhibit unimodal distribution. Aggressive opening of the cotton makes it unimodal particularly with weaker cotton because of fibre breakages. Frequency distribution from Ball sorter does not, however, show bimodal distribution. This may be because it is weight length distribution as against number length distribution by AFIS. In a subsequent paper Krifa27 found that the length distribution is close to unimodel at both low and high breakage levels in ginning and processing in spinning i.e. ginned lint and card sliver. Mature cotton exhibits an extended intermediate stage of bimodal distribution compared to immature cottons because of lower breakage level.
Xu16 et al, who used image analysis, however did not find a peak at 3 mm. Landstreet28 derived the fibrogram from number length frequency distribution by double summation, Based on a reversal of this technique, Krowicki29 et al compared the fibrogram obtained from frequency length distribution by differential and algorithm methods. Algorithm method gives results close to differential method while taking less time. Equations for fibrogram from the fibre length distribution were determined by Azzouz30 et al assuming fibre length distribution to be normal. Good correlation is found between actual weight length frequency distribution and adjusted to normal distribution. However number length distribution deviates considerably from normal and the equation is not applicable. Cai31 et al found fibre length distribution plays an important role in prediction of yarn strength and irregularity and it is therefore useful to include fibre length distribution as an important quality parameter of cotton.
Krifa32 emphasized the need for taking fibre length distribution in selection of cotton. Group of bales selected on common HVI properties produces mixing with uncontrolled length distribution variability, caused by fibre breakages, from bales with same but low maturity. Shapiro33 et al discuss mathematical models to examine cotton fibre length distribution under various breakage models. Breakages on unclamped and clamped fibres were analysed. A procedure for testing the validity of model determining effect of breakage on length distribution is proposed.
Krowicki34 examined the effect of lens width on length in Digital fibrograph. Kelly35 et al found that selection in breeding program using HVI and AFIS data gave nearly the same order of improvement in fibre quality.
Hequet36 et al showed that yarn elongation can be predicted from HVI elongation and upper half mean length while yarn strength can be predicted from AFIS mean (by weight), fineness and maturity ratio.
Ureyan and Kadoglu37 found highly significant correlations between fibre properties measured by AFIS and yarn properties.
Cui14 et al showed that a mix of Weibull distributions gives a good fit to the actual fibre length distribution of original sample as well as of the sample picked by clamp of fibro sampler. Mean length and upper half mean from theoretical distribution shows good agreement with actual results.. Belmasrour38 et al proposed a method for estimating fibre length distribution of a sample from length distribution of fibres projecting from the comb of fibrograph. Predicted curve gives good agreement with experimental except in short fibre region.
References
 K.Q. Robert and L. J. Blanchard, Cotton Cleanability : Part I: ModelingFiber Breakage, Textile Research J, 1997, 67, p 417
 K.L. Hertel, A Method of FibreLength Analysis Using the Fibrograph, Textile Research J, 1940,10, p 510.
 J.D. Tallant, Use of a Servo System for Automatic Operation of the Fibrograph, Textile Research J, 1952, 22, p 617
 J.D. Tallant, Improvement on the Servo Conversion of Manual Fibrographs, Textile Research J, 1958, 28, p 815.
 J.T. Rouse, The Use of Dial Gauges in Calculating the Results of Fibrograph Length Tests, Textile Research J, 1958, 28, p 505.
 P.R. Ewald and S.Worley, JR, Converting the Fibrograph to Automatic Direct Reading Operation, Textile Research J, 1961, 31, p 602
 K.L.Hertel and C.J.Craven, J of Textile Industries, 1960, 124, 7, p 103.
 R.S. Krowicki and D. P. Thibodeaux, Holding Length: Effect on Digital Fibrograph Span Length, Textile Research J, 1990, 60, p 383,
 R. S. Krowicki and H. H. Ramey, An Examination of the Digital Fibrograph Length Uniformity Index, Crop Science, 1984, 24, p 378.F
 Carpenter, Evaluation of the fibro sampler and the Digital Fibrograph for sampling cotton fibres, and measuring length characteristics. http://ia601601.us.archive.org/5/items/evaluationoffibr775carp/evaluationoffibr775carp.pdf.
 Yoakum.Roger L. Preliminary evaluation of the Digital fibrograph and fibro sampler, U.S. Dept. Agr.[Unpublished report.] 1959.
 Y. Cai, X. Cui, J. Rodgers, V. Martin andM. Watson, An investigation of the sampling bias of the beard method as used in HVI. J Textile Inst. 2010, 101, p 958.
 Y.T.Chu and C. R. Riley, Jr.New Interpretation of the Fibrogram, Textile Research j, 1997, 67, p897
 X.L.Cui, J.Rodgers, Y.Cai, L. Li, R.Belmasrour and d S. Pang, Obtaining Cotton Fiber Length Distributions from the Beard Test Method Part 1  Theoretical Distributions Related to the Beard Method, J of cotton Science, 2009, 13, p265
 Ikiz, J. P. Rust, W. J. Jasper and H. J. Trussell, Fiber Length Measurement by Image Processing, Textile Research J, 2001, 71, p 905
 W.Xu, B.Xu, W. Li and W. Cui, Snippet Counting for Cotton Length Distribution Measurement Using Image Analysis, Textile Research J, 2008, 78, p 336
 A.U.Nair, R.P.Nachane and B.A. Patawardan, Comparative study of different test methods for the measurements of physical properties of cotton, Indian J of fibre and textile research, 2009, 34, p 352.
 R. Audivert and Ma. D. de Casteller, The relationsbetweenthefibrelengthparameters obtained fromthe Digital Fibrograph, the comb sorter, and the Shirley Photoelectric staplerJ, Textile Inst., 1972, 63, p 356.
 Jai Prakash, Evaluation of Length Parameters Obtained with the Digital Fibrograph with SpecialReference to Fiber Length Nonuniformity, Textile Reaearch J, 1064, 34, p 857.
 H.H. Ramey, JR and P.G. Beaton, Relationships Between Short Fiber Content and HVI Fiber Length Uniformity, Textile Research J, 1989, 59, p 101.
 J.D. Bargeron III, Preliminary Investigation of the Length Measurement of Cotton Fibers with the PeyerTexlab System: Comparability and Repeatability, Textile Research J, 1986, 56, p121
 C. K. Bragg and F. M. Shofner, A Rapid, Direct Measurement of Short Fiber Content, Textile Research J, 1993, 63, p 171
 X.Cui, T. A. Calamari, K. Q. Robert, and J. B. Price, Measuring the Short Fiber Content of Cotton, Textile Research J, 2003, 73, p891
 D. Thibodeaux, H. Senter, J. L. Knowlton, D. McAlister, and X. Cui, A comparison of Methods for Measuring the Short Fiber content of cotton, J of Cotton science, 2008, 12, p 298.
 M. I. Zeidman, S. K. Batra and P. E. Sasser,Determining Short Fiber Content in Cotton : Part I: Some Theoretical Fundamentals, Textile Research J, 1991, 61, p 21.
 M.Krifa, Fiber Length Distribution in Cotton Processing: Dominant Features and Interaction Effects, Textile Research J, 2006, p 426.
 M.Krifa, Cotton fiber length distribution modality alteration in ginning and mill processing, J Textile Institute, 2013, 104, p 731.
 Landstreet, C. B., The Fibrogram : Its Concept and usein Measuring Fiber Length, Textile Bull. 1961, 87, p54.
 R.S. Krowicki, D.P. Thibodeaux and K.E. Duckett, Generating Fiber Length Distribution from the Fibrogram, Textile Research J, 1996, 66, p 306.
 B. Azzouz, , M. B. Hassen,., F Sakli,Adjustment of Cotton Fiber Length by the Statistical Normal Distribution: Application to Binary Blends, J of Engineering Fibres and Fabrics, 2008, 3, 3, p 35
 Y.Cai, X. Cui, J. Rodgers, D. Thibodeaux, Vi. Martin, M. Watson and S. Pang, A comparative study of the effects of cotton fiber length parameters on modeling yarn properties, Textile Research J, 201383, p 961
 M.Krifa, Fiber length distribution variability in cotton bale classification: Interactions among length, maturity, Textile Research J, 2012, 62, p 1244.
 H.N. Shapiro, G.Sparer, H.E. Gaffney, R. H. Armitage and J.D. Tallant, Mathematical Aspects of Cotton Fiber Length Distribution under Various Breakage Models, Textile Research J, 1964, 34, p 303.
 R. S. KrowickiThe Effect of Lens Width on Length Measurements by the Digital Fibrograph, J, Textile Institute, 1986, 77, p 223.
 C.M. Kelly, E.F. Hequet, and J.K. Dever, Interpretation of AFIS and HVI Fiber Property Measurements in Breeding for Cotton Fiber Quality Improvement, J of cotton science, 2012, 16, p 1.
 E.Hequet, N.Abidi, and J. R. Gannaway, Relationships between HVI, AFIS, and yarn tensile properties,http://wcrc.confex.com/wcrc/2007/techprogram/P1794.HTM.
 M. E. Üreyen, and H.Kadoğlu,The Prediction of Cotton Ring Yarn Properties from AFIS Fibre Properties by Using Linear Regression Models, Fibres and Textiles in Eastern Europe, 2007, 15, 4, p 63
 R.Belmasrour, L. Li, X. L. Cui, Y.Cai, and J.Rodgers, Obtaining Cotton Fiber Length Distributions from the Beard Test Method Part 2 – A New Approach through PLS Regression, J of cotton Science, 2011, 15, p 73