See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/43277118

The Impact of Short Fiber Content on the Quality of Cotton Ring Spun Yarn.

Article · January 2008

Source: OAI

CITATIONS READS
21 1,021

5 authors, including:

Devron Thibodeaux David D. Mcalister

87 PUBLICATIONS   833 CITATIONS   
United States Department of Agriculture
46 PUBLICATIONS   443 CITATIONS   
SEE PROFILE
SEE PROFILE

All content following this page was uploaded by Devron Thibodeaux on 01 June 2014.

The user has requested enhancement of the downloaded file.

The Journal of Cotton Science 12:368–377 (2008) 368
http://journal.cotton.org, © The Cotton Foundation 2008

TEXTILE TECHNOLOGY
The Impact of Short Fiber Content on the Quality of Cotton Ring Spun Yarn
D. Thibodeaux,* H. Senter, J. L. Knowlton, D. McAlister, and X. Cui
ABSTRACT in length). Because machine-picked U.S. cotton is
essentially “dirtier,” it requires excessive lint cleaning
This study was designed to quantitatively as- that can lead to fiber damage and the creation of
sess the effects of short fiber (< 12.7 mm long) in broken, short fiber fragments. For the spinner, the
raw cotton on the quality of 20s ring spun yarn. presence of excessive amounts of short fiber in the
Twenty-eight bales of cotton with a wide range of input mix can result in production inefficiencies and
fiber properties and an especially diverse popula- losses in textile quality. Behery (1993) described
tion of short fiber content were utilized. Properties how short fibers behave during textile processing:
of the raw cotton were measured by High Volume during cotton spinning, the fiber strands are thinned
Instrumentation (HVI), Advanced Fiber Informa- or drafted by passing between pairs of drafting rolls
tion System (AFIS) instruments, and the manual that are spaced at distances that allow most fibers
Suter-Webb (SW) Array method. Ring spun to pass through without bridging the gap between
yarns produced from these cottons were tested the rollers, which would result in breaking of the
for process and product quality. Results indicate fibers. Short fibers are allowed to float between the
that most of the yarn properties—especially yarn drafting rollers where they can bunch up or thin out
strength, irregularity, and frequency of thick and causing thick and/or thin imperfections in the yarn
thin defects—are strongly correlated with each of with accompanying diminished strength.
the three measures of short fiber content. A pool of Over the years, several researchers have studied
23 potential predictors from the AFIS, HVI, and problems arising in cotton spinning resulting from the
SW fiber properties was utilized to develop “best” presence of short fibers. Tallant et al. (1959) prepared
regression models for seven yarn properties. In five cottons having four different levels of short fiber by
of the seven models, the short fiber content variable cutting sliver into ¼- and ½-in segments that were
was the most important predictor, exceptions being then added to the parent cotton in varying amounts.
the models for yarn strength and elongation. For Defining short fiber as fiber < 3/8 in (9.5 mm) in
five of the yarn properties, models developed using length, they found that yarn strength and elongation
the four basic HVI properties alone were nearly both diminished with increased short fiber whereas
as good in predicting yarn quality as those using Uster yarn uniformity (% CV) increased. Similarly,
all 23 fiber properties. Exceptions were models for Bargeron (1986) studied the effects of short fiber on
elongation and for ends down. spinning performance by adding varying percentages
of comber noils to raw cottons from well-blended
bales. His study included additions of 5%, 10%, 15%,
O ne of the biggest concerns relative to the global
marketing of U.S. cotton is the perception by
international spinners that it contains excessive
and 20% comber noils. Over the range of increase
(represented by increasing amounts of short fiber),
amounts of short fiber (i.e., fiber < ½ in (12.7mm) ends down per 1000 spindle hours increased exponen-
tially from approximately 50 to more than 200. Yarn
strength diminished by more than 10%, appearance
grade decreased by more than 30%, and irregularity
D. Thibodeaux* , USDA-ARS, Cotton Quality Research
%CV increased by more than 20%.
Station, P.O. Box 792, Clemson, SC 29633; H. Senter,
Clemson University, Department of Mathematical Sciences,
The impact of short fiber in commercial cotton
Clemson, SC 29634; J.L. Knowlton, USDA-AMS-Cotton spinning was reported by Backe (1986). He conducted
Program, 3275 Appling Road, Memphis, TN 38133; D. a large-scale plant trial (400 bales) to examine the
McAlister, Uster Technologies, Inc., 456 Troy Circle, PO Box effects of low (8.6%) and high (11.6%) short fiber
51270, Knoxville, TN 37919; X. Cui, USDA, ARS, SRRC, content as measured by the Peyer AL-101 instrument
1100 Robert E. Lee Blvd., New Orleans, LA 70124 (manufactured by Peyer Texlab, Zurich Switzerland
*Corresponding author: devron.thibodeaux@ars.usda.gov (out of business)). Short fiber showed a statistically
THIBODEAUX ET AL.: IMPACT of Short Fiber Content oN RING SPUN YARN QUALITY 369

significant influence on ends-down in spinning. There Regional Research Center in New Orleans, and from
was also a significant increase in Uster %CV and Clas- SW array analysis performed at the Cotton Quality
simat (manufactured by Zellweger Uster, Charlotte, Research Station in Clemson, NC. This paper focuses
NC) long thin places with short fiber content. Like- on the relationship of fiber properties—especially short
wise, yarn strength, appearance index, and Uster IPI fiber content—to selected characteristics of 20s ring
thick and thin places were all affected by an increase spun yarn made from those 29 cottons. Ring spinning
in short fiber at the 95% confidence level. was carried out at approximately 10,000 rpm spindle
Chanselme et al. (1997) reported on the use of speed with a twist multiplier (T.M.) of 3.85. Testing
Advanced Fiber Information System (AFIS) fiber data involved about 240 spindle hours. Yarn strength was
to study the effects of fiber properties (including short measured on a Statimat M (manufactured by Textechno,
fiber) on the quality of ring spun yarns. They utilized Mönchengladbach, Germany) tester at 254 mm gauge
a micro-spinning test on 23 Upland cottons producing length. The rate of elongation used was 5000 mm/
both ring and rotor yarns (20 tex) and studied yarn min and a total of 400 breaks per determination were
strength, elongation, evenness, thick and thin places, utilized. An Industrial Laboratory Equipment (ILE 65,
neps, and hairiness. Fiber length and diameter param- Charlotte, NC) was used for evenness testing on 20
eters showed the highest correlations with %CV, neps, bobbins using 1000 yd per bobbin at a speed of 400
hairiness, and the square roots of thick and thin places. yd/min. Two independent replications of yarn measure-
Short fiber content (by weight) correlated significantly ments were made for each of the cottons. Yarn data for
with hairiness and the square root of thin places. Hequet one cotton was inadvertently lost reducing the number
(1999) also utilized AFIS fiber data to study the effects of measurements per yarn property to 56. The yarn
of fiber properties on ring spun yarns (both 36s and 50s spinning and measurements of the yarn properties were
English count NE) for both Upland and San Joaquin done at the Cotton Quality Research Station.
Valley (SJV) cottons. Significant correlations were
found between AFIS short fiber by weight and yarn RESULTS
evenness, thin places, thick places, neps, and hairiness.
Hequet and Ethridge (2000) reported further progress Abbreviated names of the important fiber prop-
on the previous study where they considered the ef- erties and yarn characteristics considered, and the
fects of other fiber length distribution parameters as notations for them are listed in Table 1. Lower case
measured by AFIS. These results essentially supported abbreviations are given to the yarn characteristics,
their previous findings on the same set of cottons. reserving upper case abbreviations for fiber proper-
At the 2004 Beltwide Cotton Conference, ties. The fiber variable names are the same ones used
Knowlton (2004) reported on a collaborative effort in the earlier companion report (Thibodeaux et al.,
“to evaluate available short fiber measurements rela- 2007) for these cottons.
tive to each other and relative to textile processing Differences in fiber characteristics among the 28
performance” (Knowlton, 2004). The study encom- bales are indicated by the summary statistics for the
passed several laboratories conducting a wide range HVI variables given in Table 2. As was observed in
of tests on several different instruments. Twenty-nine Thibodeaux et al. (2007, Table 1), the bales show a
commercial bales that had a wide range of fiber wide range of fiber properties: (2.33 < HVIUHM <
properties were chosen for the study. 3.07), (77.8 < HVIUNIF < 84.4), (23.4 < HVISTR <
33.2), (2.92 < HVIMIC < 5.52), and (6.86 < HVISFI
MATERIALS AND METHODS < 17.13). As seen in Table 3, the ranges of short fiber
contents from the three methods are: HVI, 6.86–
For the same set of cottons, relationships among 17.13%; AFIS, 5.61–19.76%; and SW, 6.40–26.60%.
High Volume Instrumentation (HVI), AFIS, and The three measures of short fiber content—HVISFI,
Suter-Webb (SW) fiber properties were discussed in AFISSFC, and SWSFC—are strongly pairwise cor-
Thibodeaux et al. (2007), with emphasis on measures related with Pearson’s correlations between 0.88 and
of fiber length, especially short fiber content. The fiber 0.95 (Table 4). Regression models for predicting
data reported in that paper, which are also used in this SWSFC (the more difficult measurement to make
study, were based on HVI measurements from the US- and generally considered the “gold standard”) from
DA-Agricultural Marketing Service (AMS) Memphis HVI and AFIS fiber properties are also discussed in
Classing Office, the AFIS instrument at the Southern Thibodeaux et al. (2007).
JOURNAL OF COTTON SCIENCE, Volume 12, Issue 4, 2008 370

Table 1. Fiber variable names and yarn characteristic variable names

Fiber Variable Names

AFISFINE AFIS fineness (millitex)
AFISIFC AFIS immature fiber content (%)
AFISLW AFIS mean length by weight (cm)
AFISLWCV Coefficient of variation of AFIS mean length by weight (%)
AFISMAT AFIS maturity ratio (dimensionless)
AFISNEPS AFIS nep count (/g)
AFISSFC AFIS short fiber content (%)
AFISUQL AFIS upper-quartile length (cm)
HVILE HVI mean length (cm)
HVIMIC HVI micronaire (dimensionless)
HVISFI HVI short fiber index (%)
HVISTR HVI strength (g/tex)
HVIUHM HVI upper-half mean length (cm)
HVIUNIF HVI uniformity (%)
SWLE SW mean length (cm)
SWSFC SW short fiber content (%)
SWUQL SW upper-quartile length (cm)
Yarn Characteristic Variables
elong Elongation to break (%)
endsdown Number of spinning breaks per 1000 spindle hours
irrcv Coefficient of variation of yarn mass (%)
neps Number of yarn neps/1000 m
thicks Number of yarn thick spots/1000 m
thins Number of yarn thin spots/1000 m
yarnstr Yarn breaking strength (cN/tex)

Table 2. Descriptive statistics for the HVI properties

Variable N Mean StDev Minimum Median Maximum

HVISFI 28 10.183 2.863 6.860 8.915 17.130
HVILE 28 2.262 0.222 1.817 2.291 2.564
HVIUHM 28 2.769 0.220 2.327 2.790 3.066
HVIUNIF 28 81.571 1.886 77.800 81.950 84.400
HVISTR 28 28.338 2.982 23.393 28.278 33.173
HVIMIC 28 4.332 0.591 2.920 4.270 5.520

Table 3. Descriptive statistics for short fiber properties by the three methods.

Variable N Mean StDev Minimum Median Maximum

HVISFI 28 10.18 2.86 6.86 8.92 17.13
AFISSFC 28 12.15 4.16 5.61 12.17 19.76
SWSFC 28 13.21 5.28 6.40 11.20 26.60
THIBODEAUX ET AL.: IMPACT of Short Fiber Content oN RING SPUN YARN QUALITY 371

Table 4. Correlations between the three measures of short fiber The relationships of individual fiber properties to
(Pearson’s correlation coefficient with p value below) . yarn characteristics are examined in the correlation
HVISFI AFISSFC matrix in Table 6. HVISFI is significantly correlated
AFISSFC 0.885 (all r-values with p < 0.05 are shown in bold) to each
0.000 0.000 of the yarn variables. The HVIUNIF correlations are
SWSFC 0.947 0.898 close in magnitude to those of HVISFI but opposite
0.000 0.000 in sign. Correlations between AFISSFC and the
yarn variables are similar to those for HVISFI, but
Here we are concerned with the relationship of are slightly weaker. Correlations between SWSFC
SFC, in conjunction with other fiber properties, to and the yarn variables are close to those for HVISFI.
corresponding yarn properties (for 20s ring spun yarn). Correlations between the yarn variables and the staple
Descriptive statistics for the distributions of the yarn length measures—HVILE, AFISLW, and SWLE—are
properties are given in Table 5. The wide range of close in value, with SWLE showing slightly stronger
values of the fiber properties are reflected in the spans correlations. Correlations between the yarn variables
of the yarn characteristics. Some of the variables— and upper-half lengths—HVIUHM, AFISUQL, and
ends down, thicks, and thins—have strongly (right) SWUQL—track the corresponding correlations for
skewed distributions. In subsequent analyses, square staple lengths, but are generally weaker. These six
root transformations of these variables are used to get length measures are strongly correlated with all the
better predictive models (Chanselme et al., 1997). yarn variables except elong and neps.

Table 5. Descriptive statistics for yarn properties

Variable N Mean StDev CoefVar Minimum Median Maximum
yarnstr 56 14.21 2.69 18.95 9.84 13.90 18.76
elong 56 6.06 0.56 9.27 4.73 6.09 7.04
endsdown 56 21.89 42.10 192.30 0.00 8.00 246.00
neps 56 97.80 40.94 41.86 32.81 88.58 188.10
thicks 56 803.50 490.70 61.07 193.60 738.20 2260.50
thins 56 234.70 279.20 118.97 6.60 120.30 1217.20
irrcv 56 18.41 2.24 12.17 14.70 18.25 24.20

Table 6. Correlations between fiber and yarn properties for short fiber and length variation (Pearson’s correlation coefficient
with p value < 0.05 shown in italics).
yarnstr elong endsdown neps thicks thins irrcv
AFISFINE -0.396 0.057 -0.137 -0.167 0.094 0.135 0.162
AFISIFC 0.019 0.647 -0.345 -0.049 -0.181 -0.271 -0.122
AFISLW 0.885 0.195 -0.404 -0.290 -0.794 -0.801 -0.861
AFISLWCV -0.355 -0.065 0.273 0.526 0.615 0.534 0.583
AFISMAT 0.517 -0.542 0.070 -0.073 -0.306 -0.256 -0.403
AFISNEPS -0.175 0.164 0.206 0.317 0.381 0.316 0.369
AFISSFC -0.725 -0.164 0.472 0.485 0.810 0.777 0.830
AFISUQL 0.856 0.196 -0.372 -0.163 -0.708 -0.768 -0.777
HVILE 0.889 0.231 -0.391 -0.223 -0.761 -0.795 -0.825
HVIMIC -0.077 -0.063 -0.268 -0.171 -0.158 -0.118 -0.132
HVISFI -0.807 -0.406 0.609 0.407 0.881 0.906 0.888
HVISTR 0.959 0.014 -0.326 -0.276 -0.761 -0.747 -0.842
HVIUHM 0.864 0.216 -0.371 -0.147 -0.693 -0.749 -0.765
HVIUNIF 0.824 0.274 -0.428 -0.456 -0.879 -0.844 -0.901
SWLE 0.890 0.291 -0.413 -0.307 -0.828 -0.829 -0.874
SWSFC -0.794 -0.366 0.529 0.484 0.906 0.879 0.897
SWUQL 0.858 0.294 -0.389 -0.181 -0.733 -0.781 -0.791
JOURNAL OF COTTON SCIENCE, Volume 12, Issue 4, 2008 372

Fiber properties that correlate strongly with yarn Because the four basic HVI properties—HVI-
strength, thicks, thins, and irrcv are HVISTR and all of UHM, HVISTR, HVIUNIF, and HVIMIC—are
the measures of fiber length and short fiber content. By almost always available, whereas some of the other
contrast, only AFISIFC and AFISMAT have a correlation fiber properties (such as the SW ones) may not be,
stronger than 50% with elongation, whereas just one of models built from the basic HVI properties are de-
the fiber properties in Table 6 has a correlation stronger scribed and compared to the best models built from
than 50% with neps. One fiber property that shows little all the fiber properties.
association with the yarn variables is HVIMIC. The re- For model building purposes, the list of can-
lated measure AFISFINE is marginally significantly cor- didate predictors included all 17 fiber properties
related with only yarnstr. The strong pairwise correlations listed in Table 1 plus six AFIS variables not listed:
between the three measures of short fiber content and dust particle count, length by number, mean neps
all of the yarn variables suggest that short fiber content size, total nep count, trash particle count, and vis-
is an important determinant (or predictor) of spinning ible foreign matter. These AFIS variables did not
performance. Despite a lack of simple association with exhibit significant correlation with more than one
the yarn variables, HVIMIC and AFISFINE, in combina- yarn property and they were not significant predic-
tion with other fiber properties, are useful predictors of tors in models for yarn properties.
some yarn properties as will be seen below. Models for yarn strength (yarnstr). As seen in
Regression models for yarn properties. In this Table 7, a good model for estimating yarn strength
section, regression models for estimating the seven using HVI, AFIS, and SW fiber properties as poten-
yarn characteristics from fiber properties are presented. tial predictors is one with three predictors and R2 =
As observed in the previous section, most of the fiber 0.9488. The regression equation is:
properties are correlated with each of the yarn proper-
yarnstr = − 0.543 + 0.696 ∗ HVISTR
ties. The fiber properties are highly correlated among
− 0.746 HVIMIC − 0.132 ∗ SWSFC [1]
themselves (Thibodeaux et al., 2007). The models
discussed here were selected from many candidate Several
yarnstr = −33alternative
.72 + 0.669 models
∗ HVISTRthat include HVIS-
models as parsimonious ones with relatively good TR, HVIMIC,− 0.802plus some measure of fiber length
∗ HVIMIC
explanatory power (high R2), low colinearity, small (AFISLW,+AFISLN, HVILE, SWLE, AFISUQL,
0.398 ∗ HVIUNIF [2]
variance inflation factors (VIFs), and good residuals HVIUHM, SWUQL, AFISSFC, or HVISFI) are
plots. For some yarn properties, two or even several yarnstrin=terms
similar −4.42of− R02.024 ∗ STDFINE
(values from 0.9285 to 0.9511),
models were about equally good because one or more .814 * HVISTR
but exhibit+ 0greater colinearity among the predic- [3]
fiber properties such as HVIUNIF or SWSFC could be tors. The model above has the least colinearity;
elong = 17.90 + 0.014 ∗ AFISFINE
substituted for another such as HVISFI. The models standard errors of the common predictors HVISTR
− 14.17 ∗ AFISMAT − 0.150 ∗ HVISFI [4]
reported are the simplest ones with relatively high R2 and HVIMIC are smallest for this model. Thus the
among the good candidate models. Ideal values for regression
elong = −18coefficients
.27 − 0.192for this model are estimated
∗ HVISTR
the VIFs are 1, indicating no colinearity among the with the −
greatest
14.17 ∗ HVIMIC +Evidently
precision. fiber length[5]
0.391 ∗ HVIUNIF is
predictors. VIF values in excess of 10 suggest severe an important predictor of yarn strength (in conjunc-
colinearity and an unstable model (Kutner et al., 2004). tion with HVISTR and
neps = −871 .8 + 186 .8 ∗HVIMIC)
HVIUHMand any of several
In the models presented, most of the VIFs are less than measures+ 59.73 ∗ HVIMIC can
of fiber length + 15serve nearly as well.
.94 ∗ AFISSFC [6]
5.0, and all are less than 6.0. Because we are interested SWSFC may have an edge in this analysis because
in short fiber content, models with a measure of short the cottons
neps = 1945used
+ 192were deliberately selected to have
.7 ∗ HVIUHM
fiber content were preferred. a wide range
+ 28.84of∗short
HVIMICfiber −content.
30.73 ∗ HVIUNIF [7]
Table 7. Best unweighted least squares linear regression of yarnstr
Predictor Variables Coefficient Std Error ends
T = −34.95P + 1.471 ∗ HVISFI
VIF STB
Constant -0.5433 2.095 -0.260 37 ∗ HVILE
+ 10.0.796 0.000 [8]
HVISTR 0.6960 0.049 14.250 0.000 2.900 0.764
HVIMIC -0.7462 0.163 ends = 78.42
-4.580 − 0.919 ∗ HVIUNIF
0.000 1.300 -0.162 [9]
SWSFC -0.1318 0.029 -4.540 0.000 3.200 -0.256
thicks = −14.53 + 0.134 ∗ AFISFINE
R2 0.9488 Resid. Mean Square (MSE) + 1.47 ∗ SWSFC 0.39259 [10]
Adj. R2 0.9458 0.62657
Standard Deviation
thicks = 360.4 + 1.77 ∗ HVIMIC
− 4.18 ∗ HVIUNIF [11]

thins = −46.07 + 0.240 ∗ AFISFINE

+ 1.455 ∗ SWSFC [12]
THIBODEAUX ET AL.: IMPACT of Short Fiber Content oN RING SPUN YARN QUALITY 373

The regression coefficients in Eq. 1 indicate the no correlation with yarnstr (r = –0.077) yet it is a
effect on yarnstr of a unit increase in a fiber property highly significant predictor of yarnstr (p = 0.0001)
holding the other properties fixed. However, the in linear combination with the other two properties.
magnitudes of the coefficients do not necessarily We observe this role of HVIMIC in other models.
indicate the relative importance of the predictors in Hequet (1999) found that the ratio AFISFINE/
their impact on yarnstr because they are measured AFISMAT, which he called “standard fineness,” is
in different units. A unit increase in a length variable a good predictor of yarn strength for ring spun yarn.
(1 cm) is astronomical, whereas a unit increase in yarnstr
For = − 0.543
the cottons 0.696
and+yarns HVISTR
in∗this study, the correlation
short fiber content (1%) is modest. between standard fineness and
− 0.746 HVIMIC yarnstr
− 0.132 is r = –0.70.
∗ SWSFC [1]
Standardized regression coefficients, denoted We, too, found that standard fineness, denoted STD-
STB in Table 7, are unitless and permit a direct com- yarnstrtogether
FINE, = −33.72with .669 ∗ HVISTR
+ 0HVISTR is a good predictor of
parison of the impact of the predictors on the response − 0.with
yarn strength 802 ∗anHVIMIC
R2 = 0.9259. The regression
(Kutner et al., 2004). A standardized coefficient is the equation is+ 0.398 ∗ HVIUNIF [2]
effect on the response in standard deviations of a one
yarnstr = −4.42 − 0.024 ∗ STDFINE
standard deviation increase in a predictor, holding the
+ 0.814 * HVISTR [3]
other predictors fixed. We see in Table 7 that a one
standard deviation increase in HVISTR will increase elongReplacing
= 17.90 +HVISTR with HVISFI in the above
0.014 ∗ AFISFINE
yarnstr by 0.764 standard deviations (holding HVI- model gives
− 14.an 17 ∗equally
AFISMAT good− model
0.150 ∗(R 2 = 0.9241),
HVISFI [4]
MIC and SWSFC fixed). Because the standardized illustrating that different combinations of fiber
yarnstr= =−18
elong − 0.27 .543 0.696
− 0+.192 ∗ HVISTR
∗ HVISTR
coefficient of HVISTR is the largest in absolute value, properties can yield comparable models for yarn
0 . 746 HVIMIC 0
− 14.17 ∗ HVIMIC + 0.391 ∗∗HVIUNIF
− − . 132 SWSFC [1] [5]
we can say that HVISTR has the most effect on yarnstr properties.
of the three predictors. Moreover, it has about three Models
yarnstr = −33 for.72 elongation
+ 0.669 ∗(elong).
HVISTRMost of the fiber
times the effect of SWSFC. properties
neps = −871 show. 8 + little
186
− 0.802 ∗ HVIMIC . 8 correlation
∗ HVIUHM with elong (Table
A model that is nearly as good is one with three 6). The +strongest
59
+ .073 .398 association
∗ HVIMIC
∗ HVIUNIF is between
+ 15.94 ∗ AFISSFCelong and
[6]
[2]
of the four basic HVI properties (see Table 8), with yarnstr = − 0 .543 + 0 .696 ∗ HVISTR
AFISIFC (r = 0.64). As seen in Table 9, a good model
HVIUNIF replacing SWSFC in the first model. neps = 1945
yarnstr
for elong −40one
= is− +..42
746
192−.HVIMIC
07.∗024
with HVIUHM− 0.132
∗ STDFINE
three fiber ∗ SWSFChaving
properties, [1]
The Pearson’s correlation between HVIUNIF and an R =28
2 =+0.81.
yarnstr +
−33.084 ∗ HVIMIC
.The
814
.72 +regression 30
.669 ∗−HVISTR
* 0HVISTR .73 ∗ HVIUNIF
equation is [3]
[7]
SWSFC is –0.95, and HVIUNIF acts as a surrogate −.90 0.802 ∗ HVIMIC
elong
ends ==17 340..95 + 0+.1014 ∗ ∗AFISFINE
HVISFI
for SWSFC. A model relating SWSFC and HVIUNIF −+ 398 ∗.HVIUNIF
471 [2]
yarnstr
is = − 0in.543
discussed + 0.696 ∗etHVISTR
Thibodeaux al. (2007). The regres- 10.37 ∗ HVILE − 0.150 ∗ HVISFI [4]
−+14 . 17 ∗ AFISMAT [8]
sion equation 746 HVIMIC − 0.132 ∗ SWSFC [1]
− 0.is yarnstr
In =
this− 4 .
model, 42 −
elong = −18.27 − 0.192 ∗ HVISTR 0
if . 024 ∗
AFISFINESTDFINE and AFISMAT are
ends = 78
replaced +.42 − 0.*919 ∗ HVIUNIF [9]
yarnstr = −33.72 + 0.669 ∗ HVISTR −by
14 0.standard
.814
17 ∗ HVIMIC fineness
HVISTR = AFISFINE/AFIS-
+ 0.391 ∗ HVIUNIF [3] [5]
MAT (Hequet, 1999), the R 2 value drops substantially
− 0.802 ∗ HVIMIC elong 17−.90 .+530model
.+014 AFISFINE
0.44.=The
tothicks = 14best 0.134 ∗for AFISFINE
∗elong with just basic HVI
+ 0.398 ∗ HVIUNIF [2] neps = −−871 14 .
. 8
17 +∗
1.47 ∗in 186 .
AFISMAT
SWSFC 8 ∗ HVIUHM ∗ HVISFI [10]
− 0.150equation [4]
properties +results the regression
+ 59.73 ∗ HVIMIC + 15.94 ∗ AFISSFC [6]
It is =
yarnstr not−4surprising
.42 − 0.024that HVISTR and HVIUNIF
∗ STDFINE elong
thicks= =−18
360 .27 .4 −+ 10.77 HVISTR
192∗ ∗HVIMIC
are significant
+ 0.814 * HVISTRbecause both are strong-
in the model [3] neps = 1945 + 192 . 7 ∗ HVIUHM
− 14 . 17 ∗
− 4.18 ∗ HVIUNIFHVIMIC + 0.391 ∗ HVIUNIF [11] [5]
ly correlated with yarn strength (r = 0.96 and 0.82 + 28.84 ∗ HVIMIC − 30.73 ∗ HVIUNIF [7]
elong = 17.90Table
respectively, + 0.014 AFISFINE
6).∗But HVIMIC alone shows These
thins = − results
46 . 07 + are
0 . inferior
240 ∗ with R2 = 0.37.
AFISFINE
− 14.17 ∗ AFISMAT − 0.150 ∗ HVISFI [4] neps = −871.8 + 186.8 ∗ HVIUHM
ends =+ −+ 1..73
34
59 455 1SWSFC
.95∗+∗HVIMIC + 15.94 ∗ AFISSFC [12]
.471 ∗ HVISFI [6]
+ 10.37 ∗ HVILE variables
Table 8. Best unweighted least squares linear regression of yarnstr based solely upon basic HVI [8]
elong = −18.27 − 0.192 ∗ HVISTR thins = 349 + 3 . 22 ∗ HVIMIC
Predictor 14.17 ∗ HVIMIC +
− Variables 0.391 ∗ HVIUNIF
Coefficient [5]
Std Error nepsT = 1945 + 192P.7 ∗ HVIUHM VIF STB
ends =+ 78−
28 4..42
.84 HVIUNIF
29−∗∗0HVIMIC
.919 ∗ HVIUNIF − 30 .73 ∗ HVIUNIF [13]
[9]
Constant -33.7190 6.412 -5.260 0.000 0.000 [7]
neps = −871.8 + 186.8 ∗ HVIUHM
0.6694 0.060 irrcv = 4.2014
11.090 − 0.53
.054 AFISFINE
∗.134
HVISTR
+ 59.73 ∗ HVIMIC + 15.94 ∗ AFISSFC [6]
thicks
ends ==−−34 .950.000
++1.0471 4.100
AFISFINE
∗ ∗HVISFI 0.735
+ 0 .398 ∗ SWSFC -0.174[14]
[10]
HVIMIC -0.8017 0.181 -4.430 ++ 101..37 ∗∗ SWSFC
470.000HVILE 1.500 [8]
0.3976 0.101 3.930
neps = 1945 + 192.7 ∗ HVIUHM
HVIUNIF
irrcv = 109
thicks .6 +.400.000
= 360 .+673 4.700
∗ HVIMIC
1.77 ∗ HVIMIC
0.276
+ 28.84 ∗ HVIMIC − 30.73 ∗ HVIUNIF [7] ends = 78.42 − 0.919 ∗ HVIUNIF [9]
[15]
HVIUNIF
− 1.−154∗.18 ∗ HVIUNIF [11]
R2 0.9448 Resid. Mean Square (MSE) 0.42287
endsR2=
Adj. −34.95 + 1.471 ∗ HVISFI
0.9416 Standard Deviationthicks 14.07
thins ==−−46 .53++00.240 AFISFINE
.134∗∗AFISFINE
0.65029
+ 10.37 ∗ HVILE [8] .47 ∗∗ SWSFC
++11.455 SWSFC [10]
[12]
thicks 360+.43+.22
thins ==349 1.77 ∗ HVIMIC
∗ HVIMIC
ends = 78.42 − 0.919 ∗ HVIUNIF [9]
[11]
−−44.29 HVIUNIF
.18∗∗HVIUNIF [13]
 − 0.746 HVIMIC − 0.132 ∗ SWSFC [1]
yarnstr = −33.72 + 0.669 ∗ HVISTR
− 0.802 ∗ HVIMIC
JOURNAL OF COTTON SCIENCE, Volume 12, Issue 4, 2008 + 0.398 ∗ HVIUNIF [2]
374
yarnstr = −4.42 − 0.024 ∗ STDFINE
Table 9. Best unweighted least squares linear regression of elongyarnstr =+ − 00..814543 *+ HVISTR 0.696 ∗ HVISTR [3]
Predictor Variables Coefficient Std Error T − 0 .746 HVIMIC
P VIF ∗ SWSFC
− 0.132 STB [1]
yarnstr
Constant
= − 0.543 + 0.696 ∗ HVISTR 17.8979 1.0884 elong
16.44= 17.90 + 00.000 .014 ∗ AFISFINE 0.000
− 0.746 HVIMIC − 0.132 ∗ SWSFC [1] yarnstr =− 14−33 . 17 + 0.669 ∗ HVISTR
.72∗ AFISMAT − 01.1.150 ∗ HVISFI [4]
AFISFINE 0.01399 0.0040 3.50 0.001
− 0.802 ∗ HVIMIC 0.217
yarnstr
AFISMAT = −33.72 + 0.669 ∗ -14.1691 HVISTR 1.0768 -13.16
elong = −18 + 0.27 .398 −0.000 HVIUNIF
0∗.192 ∗ HVISTR 1.2 -0.894 [2]
HVISFI − 0.802 ∗ HVIMIC-0.1498 0.0131 -11.42 − 14.17 ∗0.000 1.2
HVIMIC + 0.391 ∗ HVIUNIF -0.757 [5]
+ 0.398 ∗ HVIUNIF [2] yarnstr = −4.42 − 0.024 ∗ STDFINE
R2 0.8079 Resid. Mean Square (MSE)+ 0.814 * HVISTR 0.064 [3]
yarnstr neps = −871.8 + 186.8 ∗ HVIUHM
Adj. R2 = −4.42 − 0.024 ∗ STDFINE 0.7969 Standard Deviation 0.2529
elong + 59.90 .73+∗0HVIMIC + 15.94 ∗ AFISSFC [6]
∗ AFISFINE
+ 0.814 * HVISTR [3] yarnstr= 17 = − 0.543 .014
+ 0.696 ∗ HVISTR
yarnstr
Models= − 0for .543 neps .696 ∗ HVISTR
+ 0(neps). Yarn neps range some- − 14− 0.17
The positive .746 AFISMAT
∗coefficient
HVIMIC − −00.HVILE
for HVISFI
.150∗∗SWSFC
132 [4]
is surprising.
[1]
elong = 17 −.90 0 . +
746 0 . 014
HVIMIC ∗ AFISFINE
− 0 . 132 ∗ SWSFC [1] neps = 1945 + 192.7 ∗ HVIUHM
what uniformly from 30 to 172. A strong predictive The simple correlation between √ends and HVILE
− 14.17could ∗ AFISMAT − 0.150and ∗ HVISFI [4] elong
yarnstr 28
=+=−The18
−33..84
27 ∗−HVIMIC
.72 0model
+ .192
0.669 30.73 ∗ HVIUNIF [7]
∗−HVISTR
HVISTR
∗ using
model
yarnstrfor = neps
−33.72 + 0.not 669be found,
∗ HVISTR using a loga- is –0.46. best just basic HVI fiber
rithmic or square
− 0.27 .802 rootHVIMIC transformation of the response properties is a simple linear model∗ HVIUNIF
− 14
− 0. 17
. 802 ∗ HVIMIC
∗ HVIMIC + 0 .391 involving HVI- [5]
elong = −18 − 0∗.192 ∗ HVISTR ends = − 34 .
+ 0which95 +
.398 ∗has 1 . 471
HVIUNIF∗ HVISFI [2]
did not help. + 0.The .398 best
∗ HVIUNIFmodel (Table 10) has an R[2] 2=
UNIF alone, an R2 = 0.30. The regression
− 14 17 ∗ HVIMIC + 0.391 ∗ HVIUNIF [5] neps = − + 10..837
871 + ∗ HVILE
186 . 8 ∗ HVIUHM [8]
0.48. The regression equation is equation is
yarnstr+=59 −4.73 .42∗ −HVIMIC 0.024 ∗ STDFINE
+ 15.94 ∗ AFISSFC [6]
yarnstr = −4.42 − 0.024 ∗ STDFINE 0.814 HVISTR [3]
neps = −871.8 + 186.8 ∗ HVIUHM ends = 78+.42 − 0.*919 ∗ HVIUNIF [9]
+ 0.814 * HVISTR [3]
+ 59 .73 ∗ HVIMIC
yarnstr = − 0.543 + 0.696 ∗ HVISTR + 15 . 94 ∗ AFISSFC [6] nepsModels
= 1945 +for .7 ∗ HVIUHM
192thicks
elong 17−.90 0.+014 (√thicks). A square-root
∗ AFISFINE
elong = 17 −.90 0 . +
746 0 . 014
HVIMIC ∗ AFISFINE
− 0 . 132 ∗ SWSFC [1] thicks=+= 28 14
. 84 .+53
∗ HVIMIC0.134 ∗−AFISFINE HVIUNIF
30.73 ∗√thicks, [7]
Although one would expect neps to increase transformation− 14 . 17 of
∗ thicks, denoted
AFISMAT − 0.150 ∗ HVISFI is used
[4]
neps = 1945− 14 +
. 17 192 ∗ . 7
AFISMAT∗ HVIUHM − 0 .150 ∗ HVISFI [4] + 1 . 47 ∗ SWSFC [10]
with short
yarnstr+=28 fiber
−33 .84.72 content, it is
+ 0.669 ∗−HVISTR
∗ HVIMIC surprising that
30.73 ∗ HVIUNIF [7]increased as the response because the distribution of thicks is
ends
elong
highly == −
− 34
18 . .
2795 −+ 01 .
. 471
192 ∗
∗ HVISFI
HVISTR
HVIUHM −is0associated
elong = −18 .27.802 − 0∗.192 HVIMIC with more neps. Relatively,
∗ HVISTR thicks(positively)
=+360 10..17 37
skewed
.4 +∗1HVILE.77 ∗ HVIMIC (Chanselme et al., 1997).
[8]
short fiber content is the most important of the three A good − 14
model with HVIMIC
− 4.18 ∗ HVIUNIF0.391involves
∗ no +
colinearity ∗ HVIUNIF [5]
fineness
[11]
ends = −−14 +340..17 .95
398 1∗.HVIUNIF
∗+HVIMIC 471 ∗ HVISFI+ 0.391 ∗ HVIUNIF [2] [5]
predictors. + 10The .37best ∗ HVILE model for neps using basic HVI [8] and short fiber content (Table 12) yielding an R2 =
properties substitutes HVIUNIF for AFISSFC in the ends=The
thins
neps == 78
−871 .42
46 ..807
−regression 0.0919
+− 186
+ ..240 HVIUNIF
∗∗ HVIUHM
∗ AFISFINE
8equation [9]
yarnstr = − 4.42 − 0.024 ∗ STDFINE 0.86. is
neps = − 871 . 8 + 186 . 8 ∗ HVIUHM ++ 591..73455 SWSFC + 15.94 ∗ AFISSFC [12]
∗ ∗HVIMIC [6]
previous +.42
ends =+ model
78 0.814 .*919
−∗yielding
0HVIMIC HVISTR∗an R2 = 0.41. The regression
HVIUNIF [3]
[9] thicks = −14.53 + 0.134 ∗ AFISFINE
equation is 59 .73 + 15 . 94 ∗ AFISSFC [6]
elong = 17.90 + 0.014 ∗ AFISFINE thins= 1945
neps = 349 + 1+.+47 3.∗22
192 .SWSFC
7∗∗ HVIMIC
HVIUHM [10]
thicks
neps = −14
= 1945 + .192
53 +.70∗.134 ∗ AFISFINE
HVIUHM
− 14.17 ∗ AFISMAT − 0.150 ∗ HVISFI [4] −284..84
The+=importance 29∗∗HVIMICHVIUNIF − 30.73 ∗ HVIUNIF [13]
of ∗fineness [7]
+ 28+ .184 .47∗ ∗HVIMICSWSFC − 30.73 ∗ HVIUNIF [10] [7] thicks 360.4 + 1.77 HVIMICand short fiber con-
tent on thick −− 4.018places∗ HVIUNIF concurs with results reported [11]
elong = −18.for
Models 27 −ends 0.192 ∗ HVISTR
down (√ends). Ends down is irrcv
byends = =4.−20
Hequet 34 .95.054
(1999). + 1.He ∗471 AFISFINE
∗ HVISFI
used standard fineness, the
thicks =
ends = −−14 360
34..17 . 4 +
95∗+HVIMIC1 . 77 ∗ HVIMIC
1.471 ∗ HVISFI + 0.391 ∗ HVIUNIF [5] + +0−.46
398
10 ∗ SWSFC
∗+ HVILE [14]
[8]
HVIUNIF properties [11] thins = ..07
37 0.240 ∗ AFISFINE
+−104..37 18 ∗∗ of
the most variable the seven yarn con- ratio AFISFINE/AFISMAT. Replacing AFISFINE by
HVILE [8]
sidered (CV = 192%, Table 5) and its distribution of standard fineness
+ 1.455 ∗inSWSFC the above model has little effect [12]
neps −871
thins=is= −46..807 + 186 8 ∗ HVIUHM
..240
+ 0skewed. ∗ AFISFINE irrcv =109
78.6.42
2 = =0.85). + 0.673 ∗ HVIMIC
values positively To mitigate skewness, a (Rends The− 0best .919 ∗ HVIUNIF
model for √thicks using basic [9]
ends =+ 78 59
+ 1 .42
.73
.455 −∗ 0
∗ . 919
HVIMIC
SWSFC ∗ HVIUNIF
+ 15 . 94 ∗ AFISSFC [9]
[6]
[12] thins −
= 1 .
349 15 ∗
+ HVIUNIF
3 . 22 ∗ HVIMIC [15]
square-root transformation of ends down, denoted HVI properties involves HVIMIC and HVIUNIF and
√ends, is used as the response (Chanselme et al., thicks
yields 414
R2==−−0.82, .29 .53 + 0.134
∗ HVIUNIF
where ∗ AFISFINE
HVIMIC substitutes in the[13]
first
thicks
neps
thins= 1945
= −14
=349 ++.192
533 . +.70∗∗.134
22 HVIMIC∗ AFISFINE
HVIUHM + 1.47 ∗ SWSFC [10]
1997). The+best model (Table 11) has
SWSFC − 30.73 ∗ HVIUNIF [10] an R 2 = 0.55. model for AFISFINE and HVIUNIF for SWSFC:
+− 284.1.84 .47∗∗∗HVIMIC
29 HVIUNIF [7]
[13] irrcv = 4.20 − 0.054 ∗ AFISFINE
The regression equation is
thicks+=0360 .398.4∗+SWSFC 1.77 ∗ HVIMIC [14]
thicks
irrcv =
ends= =4.−20360
34−.95 . 4
0.054 + 1 .
+ 1.∗471 77 ∗ HVIMIC
AFISFINE
∗ HVISFI − 4.18 ∗ HVIUNIF [11]
+ +0−.10
398 4..37
18 HVIUNIF
∗∗ HVILE
∗ SWSFC [11]
[8]
[14] irrcv = 109.6 + 0.673 ∗ HVIMIC
thins −=1−.15 46.∗07 + 0.240 ∗ AFISFINE
HVIUNIF [15]
thins10.=Best
Table −46unweighted
.07 + 0.240least ∗ AFISFINE
squares linear regression of neps
+ 1 .455 ∗ SWSFC [12]
irrcv
ends= =10978 .6.42 + 0−.673 0 . 919 ∗ HVIMIC
∗ HVIUNIF [9]
Predictor 1.455 ∗ SWSFC Coefficient
+Variables [12]
− 1.15 ∗ HVIUNIF Std [15]
Error T P VIF STB
thicks −14+.53 + 0∗.134 -871.834
∗ AFISFINE 177.290 thins
-4.92 = 349 + 3.22 0.000 ∗ HVIMIC 0.000
thins ==349
Constant
3.22 HVIMIC − 4 . 29 ∗ HVIUNIF [13]
HVIUHM + 1.47 ∗ SWSFC 186.7880 37.642
[10] 4.96 0.000 4.0 0.997
− 4.29 ∗ HVIUNIF [13]
HVIMIC 59.7332 12.780 4.67 0.000 3.3 0.854
thicks = 360 . 4 + 1 . 77 ∗ HVIMIC 15.9426 2.412
irrcv = 4.20 − 0.054
6.61 0.000
∗ AFISFINE5.9 1.607
irrcv = 4.20 − 0.054 ∗ AFISFINE
AFISSFC
+ 0.398 ∗ SWSFC [14]
− 4 .18
+ 0.398 ∗ SWSFC ∗ HVIUNIF [11]
[14]
R2 0.4802 Resid. Mean Square (MSE) 921.359
thinsR2= −46.07 + 0.240 ∗ AFISFINE irrcv = 109.6 + 0.673 ∗ HVIMIC
irrcv
Adj. = 109.6 + 0.673 ∗ HVIMIC0.4502 Standard Deviation
− 1.15 ∗ HVIUNIF
30.3539
[15]
+ 1.455 ∗ SWSFC [12]
− 1.15 ∗ HVIUNIF [15]
thins = 349 + 3.22 ∗ HVIMIC
− 4.29 ∗ HVIUNIF [13]
 yarnstr = − 0.543 + 0.696 ∗ HVISTR
+ 0.398 ∗ HVIUNIF [2]
− 0.746 HVIMIC − 0.132 ∗ SWSFC [1]
yarnstr = − 0.543 + 0.696 ∗ HVISTR
yarnstr =−
yarnstr = 4.42
−33 .669∗∗STDFINE
.72−+0.0024 HVISTR
− 0.746 HVIMIC − 0.132 ∗ SWSFC [1]
+ 0 . 814 * HVISTR
− 0.802 ∗ HVIMIC [3]
yarnstr = − 33 .72 + 0 .669 ∗ HVISTR
THIBODEAUX ET AL.: IMPACT of Short Fiber Content elong 0.398
+.90 ∗ HVIUNIF [2]
oN RING = 17 SPUN 0.014
+YARN ∗ AFISFINE
QUALITY 375
− 0.802 ∗ HVIMIC
− 14.17 ∗ AFISMAT − 0.150 ∗ HVISFI [4]
+ 0.398 ∗ HVIUNIF
Table 11. Best unweighted least squares linear regression of √ends
[2] yarnstr = −4.42 − 0.024 ∗ STDFINE
elong + 0.814 *.192 HVISTR ∗ HVISTR [3]
yarnstr
Predictor STDFINE
−4.42 − 0.024 ∗ Coefficient
= Variables Std Error T = −18.27 − 0 P VIF STB
Constant + 0.814 * HVISTR -34.9505 9. 045 [3] elong − 14.17 ∗0HVIMIC
-3.86 = 17.90 +0.0003
+ 0.391 ∗ HVIUNIF
.014 ∗ AFISFINE 0.000
[5]
HVISFI 1.4705 0.234 6.27 − 14.170.0000∗ AFISMAT − 5.2 0.150 ∗ HVISFI1.320 [4]
elong = 17.90 + 0.014 ∗ AFISFINE neps = − 871 . 8 + 186 . 8 ∗ HVIUHM
HVILE 10.3689 3.02 3.43 0.0012 5.2 0.722
− 14.17 ∗ AFISMAT − 0.150 ∗ HVISFI [4] elong =+−59 27∗−HVIMIC
18..73 + 15.94 ∗ AFISSFC [6]
0.192 ∗ HVISTR
R2 0.5503 Resid. Mean Square (MSE)− 14.17 ∗ HVIMIC + 0.4.6583 391 ∗ HVIUNIF [5]
elong = −18.27 − 0.192 ∗ HVISTR neps = 1945 + 192 . 7 ∗ HVIUHM
Adj. R2 − 14.17 ∗ HVIMIC + 0.391 0.5334∗ HVIUNIF
Standard[5]Deviation 2.1583
neps = + −87128.84.8 +∗186HVIMIC − 30.73 ∗ HVIUNIF [7]
.8 ∗ HVIUHM
Table 12. Best unweighted least squares linear regression of √thicks + 59.73 ∗ HVIMIC + 15.94 ∗ AFISSFC [6]
neps = −871.8 + 186.8 ∗ HVIUHM ends = −34.95 + 1.471 ∗ HVISFI
+ 59
Predictor .73 ∗ HVIMIC Coefficient
Variables + 15.94 ∗ AFISSFC [6]
Std Error nepsT= 1945 + 10+.37 192∗ .HVILE
P7 ∗ HVIUHMVIF STB [8]
Constant -14.5327 8.400 -1.73 + 28.84 ∗0.0894 HVIMIC − 30.73 ∗ HVIUNIF [7]
neps = 1945 + 192.7 ∗ HVIUHM
AFISFINE 0.13395 0.049 ends
2.71 = 78.42 −0.0091 0.919 ∗ HVIUNIF 1.0 0.141 [9]
SWSFC + 28 . 84 ∗ HVIMIC − 30 . 73
1.46808 ∗ HVIUNIF 0.082[7] ends
17.69 = −34.95 0.0000 + 1.471 ∗ HVISFI 1.0 0.925
thicks =+−10 14.37 0.134 ∗ AFISFINE
.53∗+HVILE [8]
ends
R 2
= −34.95 + 1.471 ∗ HVISFI 0.8559 Resid. Mean Square (MSE) + 1.47 ∗ SWSFC 10.2980 [10]
Adj. R2 + 10.37 ∗ HVILE 0.8504 Standard [8]Deviation
ends = 78.42 − 0.919 ∗ HVIUNIF 3.20906 [9]
thicks = 360.4 + 1.77 ∗ HVIMIC
endsModels
= 78.42 for−thins0.919(√thins).
∗ HVIUNIF As for thick places, [9]a HVI
thicks =− −14
properties 4.18 is∗+
.53 HVIUNIF
one in which
0.134 [11]
HVIMIC substitutes
∗ AFISFINE
square-root transformation of thins, denoted √thins, in the first+model for
47 ∗+ SWSFC
1..07 AFISFINE and HVIUNIF for
[10]
isthicks
used as = −14.53 + 0.134 ∗ AFISFINE
the response because the distribution of thins = − 46 0 . 240 ∗2AFISFINE
SWSFC with a resulting R = 0.8711. The regression
thins is highly
+ 1.47 ∗ SWSFC skewed. The best model
(positively) [10] thicks =is
equation 1.455
+360 .4 +∗1SWSFC
.77 ∗ HVIMIC [12]
for √thins (Table 13) has the same predictors as those − 4.18 ∗ HVIUNIF [11]
thicks = 360 ∗ HVIMIC thins = 349 + 3.22 ∗ HVIMIC
for √thicks and.4has + 1.a77similar R2 value (R2 = 0.9023).
∗ HVIUNIF
− 4.18equation [11] thins = − −46 .07∗+HVIUNIF
4.29 0.240 ∗ AFISFINE [13]
The regression is
+ 1.455irr ∗ SWSFC (irrcv). As for √thicks [12]
thins = −46.07 + 0.240 ∗ AFISFINE irrcvModels
= 4.20 −for 0.054 ∗CV AFISFINE and
√thins, the best model for irrcv is one with AFISFINE
+ 1.455 ∗ SWSFC [12] thins
and =+ 0349
SWSFC
.398 + ∗3.SWSFC
(see 22 ∗ HVIMIC
Table 14) with R2 = 0.8785. The
[14]
As was the+case − 4.29 ∗ HVIUNIF [13]
thins = 349 3.22for √thicks, replacing AFISFINE
∗ HVIMIC irrcv
regression 6 + 0.673is∗ HVIMIC
= 109.equation
by standard − 4fineness yields an equivalent model[13] (R2 .15−∗0HVIUNIF [15]
.29 ∗ HVIUNIF irrcv = −4.120 .054 ∗ AFISFINE
= 0.8974). The best model for √thins using basic
+ 0.398 ∗ SWSFC [14]
irrcv = 4.20 − 0.054 ∗ AFISFINE
0.398
Table 13.+Best ∗ SWSFCleast squares linear regression
unweighted [14]of √thins
irrcv = 109.6 + 0.673 ∗ HVIMIC
Predictor Variables Coefficient Std Error − 1.15 ∗ HVIUNIF
T P VIF STB[15]
irrcv =
Constant
109 .6 + 0 .673 ∗ HVIMIC -46.0740 6.787 -6.79 0.0000
− 1.15 ∗ HVIUNIF [15]
AFISFINE 0.2398 0.040 6.00 0.0000 1.0 0.258
SWSFC 1.4550 0.067 21.70 0.0000 1.0 0.935

R2 0.9023 Resid. Mean Square (MSE) 6.7247

Adj. R2 0.8986 Standard Deviation 2.5930

Table 14. Best unweighted least squares linear regression of irrcc

Predictor Variables Coefficient Std Error T P VIF STB
Constant 4.1972 2.082 2.02 0.049
AFISFINE 0.0538 0.012 4.39 0.001 1.0 0.211
SWSFC 0.3983 0.021 19.37 0.000 1.0 0.930

R2 0.8785 Resid. Mean Square (MSE) 0.6327

Adj. R2 0.8739 Standard Deviation 0.7954
 ends = −34.95 + 1.471 ∗ HVISFI
+ 10.37 ∗ HVILE [8]

ends = 78
JOURNAL OF.42 ∗ HVIUNIF
− 0.919SCIENCE,
COTTON [9]
Volume 12, Issue 4, 2008 376
thicks = −14.53 + 0.134 ∗ AFISFINE
Table 15. Effects of short fiber content on predicted yarn properties
+ 1.47 ∗ SWSFC [10]
HVISFI yarnstr elong neps endsdown thicks thins irrcv
thicks = 360.4 +14.75
8.00 1.77 ∗ HVIMIC
6.39 46.74 0.06 447.71 50.49 16.78
− 4.18 ∗ HVIUNIF [11]
10.00 14.19 6.09 105.91 10.15 751.80 176.94 18.48
thins
12.00 = −46.07 + 0.240 ∗ AFISFINE
13.69 5.79 152.33 37.53 1081.42 350.14 19.96
14.00 + 1 .455 ∗ SWSFC
13.27 5.49 [12]
185.99 82.20 1410.56 544.61 21.23

thins
Replacing 3.22 ∗ HVIMIC
= 349 +AFISFINE by standard fineness gives ables is significantly correlated with elong, whereas
− 4 . 29
essentially the same model (R2 = 0.8813). The [13]
∗ HVIUNIF best AFISIFC and AFISMAT are.
model using basic HVI properties substitutes AF- A summary of regression models for predicting
irrcv
ISFINE= 4with 0.054 ∗ AFISFINE
.20 − HVIMIC and SWSFC with HVISFC yarn properties that utilized all fiber properties is given
+ 0 .398 ∗ SWSFC
giving an R = 0.8438. The resulting equation is[14]
2
in Table 16. From a pool of 23 potential predictors
including 14 AFIS, 6 HVI, and 3 SW fiber properties,
irrcv = 109.6 + 0.673 ∗ HVIMIC
best models for seven yarn properties were obtained.
− 1.15 ∗ HVIUNIF [15]
Each of the models included a measure of short fiber
The pairs of models for √thicks, √thins, and ir- content. In five of the seven models, the short fiber
rcv use the same sets of predictors (AFISFINE and content variable was the most important predictor,
SWSFC or HVIMIC and HVIUNIF) and have simi- exceptions being the models for yarnstr and elong.
lar standardized regression coefficients (STBs) and Table 16. Summary of regression models for predicting yarn
R2 values. In the models with AFISFINE and SWSFC properties that utilized all fiber properties
as predictors, SWSFC is relatively more important.
Analogously HVIUNIF is relatively more important Variable Predictor Variables R2
in the models with HVIMIC and HVIUNIF. yarnstr HVISTR, HVIMIC, SWSFC 0.9488
Each of the best models for the seven yarn elong AFISFINE, AFISMAT, HVISFI 0.8079
properties includes a short fiber content predictor, neps HVIUHM, HVIMIC, AFISSFC 0.4802
either HVISFI, AFISSFC, or SWSFC. To quantify
√ends HVISFI, SWLE 0.6062
the effects of short fiber content on yarn, the mod-
els were used to predict properties of yarn spun √thicks AFISFINE, SWSFC 0.8559
from cottons with short fiber content ranging from √thins AFISFINE, SWSFC 0.9023
8% to 14%. Other fiber properties were assumed irrcv AFISFINE, SWSFC 0.8785
to be at their mean levels: specifically, HVISTR =
28.34 g/tex, HVIMIC = 4.33, HVIUHM = 2.77 cm, Although HVIMIC and AFISFINE show little as-
HVILE = 2.26 cm, AFISFINE = 166.26 millitex, sociation individually with the seven yarn properties,
and AFISMAT = 0.892. Table 15 shows the effects they are important predictors in combination with
on predicted yarn properties of increasing HVISFI other fiber properties. The only model that does not
from 8% to 14%. For models that use AFISSFC or contain one of these as a predictor is the model for
SWSFC instead of HVISFI, corresponding values ends down. When just the four basic HVI variables
were obtained by regression. This clearly illustrates are considered as candidate predictors, HVIMIC is
the degree to which all yarn properties will degener- in every model except the one for ends down.
ate as short fiber content increases. A summary of regression models for predicting
yarn properties that utilized only HVI properties is
SUMMARY AND CONCLUSIONS given in Table 17. For five of the yarn properties,
models built from the basic HVI properties were
Measures of short fiber content are correlated nearly as good (in terms of R2 and colinearity) as
with all of the yarn properties, most strongly with those built from all of the 23 fiber properties. Excep-
yarnstr, thicks, thins, and irrcv. The six long fiber tions were the models for elongation and for ends
measures are also strongly correlated with yarnstr, down. Even the best model for ends down was poor
thicks, thins, and irrcv. Correlations between the (R2 = 0.48). From a practical point of view, the basic
various fiber properties and elong, ends down and HVI properties provide as much information about
neps are much weaker. None of the long fiber vari- yarn characteristics as do all of the fiber properties.
 THIBODEAUX ET AL.: IMPACT of Short Fiber Content oN RING SPUN YARN QUALITY 377

Table 17. Summary of regression models for predicting yarn Knowlton, J.L. 2004. Evaluation of short fiber measurement
properties that utilized only HVI fiber properties methods. p. 2370–2377. In Proc. Beltwide Cotton Conf.
San Antonio, TX. 3–6 Jan. 2004. Natl. Cotton Counc.
Variable Predictor Variables R2
Am., Memphis, TN.
yarnstr HVISTR, HVIMIC, HVIUNIF 0.9448
Kutner, M., C. Nachtsheim, J. Neter, and W. Li. 2004. Ap-
elong HVISTR, HVIMIC, HVIUNIF 0.3700 plied Linear Statistical Models, Fifth Edition. McGraw-
neps HVIUHM, HVIMIC, HVIUNIF 0.4076 Hill/Irwin, New York, NY.
√ends HVISFI, HVILE 0.5503 Tallant, J. D., Fiori, L., and L. Legendre. 1959. The Effect of
√thicks HVIMIC, HVIUNIF 0.8235 Short Fibers in a Cotton on its Processing and Product
Quality, Part I: Affecting the Short Fiber Content by the
√thins HVIMIC, HVIUNIF 0.8711 Addition of Cut Cotton Fibers. Textile Res. J. 29:686–
irrcv HVIMIC, HVIUNIF 0.8438 695.

Thibodeaux, D., H. Senter, J. Knowlton, D. McAlister, and X.

Among the 14 AFIS properties considered, three Cui. 2007. Measuring the short fiber content of cotton. In
of them—AFISIFC, AFISFINE and AFISMAT— Cotton: Nature’s high-tech fiber. Proc. World Cotton Res.
appear to capture qualities of the fibers that other Conf.-4, Lubbock, TX. 10–14 Sept. 2007.
variables do not measure. Although AFISFINE is
correlated with HVIMIC, the two variables evidently
do not measure the same characteristics of the fibers.
The ratio of AFISFINE/AFISMAT, which estimates
standard fineness, did not offer advantages over
AFISFINE is predicting yarn properties.

ACKNOWLEDGMENTS

We gratefully accept the support of Cotton Incor-

porated under Cooperative Agreement No. 07-158.

REFERENCES

Behery, H. 1993. Short fiber content and uniformity index

in cotton. International Cotton Advisory Committee
Review Articles on Cotton Production Research No. 4.
CAB International, Wallingford, Oxon, UK.

Backe, E. 1986. Effect of short fiber content in cotton on plant

performance and quality. Textile Res. J. 56:112–115.

Bargeron, J. D. III. 1986. Relationship cotton length unifor-

mity to yarn quality. Proceedings of the National Cotton
Textile Conference. Myrtle Beach, SC. 6-8 November.

Chanselme, J., E. Hequet, and R. Frydrych. 1997. Relation-

ship between AFIS fiber characteristics and yarn even-
ness and imperfections. p. 512–516. In Proc. Beltwide
Cotton Conf., New Orleans, LA. 7–10 Jan. 1997. Natl.
Cotton Counc. Am., Memphis, TN.

Hequet, E. 1999. Application of the AFIS multidata. p.

666–670. In Proc. Beltwide Cotton Conf., Orlando, FL.
3–7 Jan. 2000. Natl. Cotton Counc. Am., Memphis, TN.

Hequet, E. and D. Ethridge. 2000. Effect of cotton fiber

length distribution on yarn quality. p. 1507–1514. In
Proc. Beltwide Cotton Conf., San Antonio, TX. 4–8 Jan.
2000. Natl. Cotton Counc. Am., Memphis, TN.

View publication stats