PREDICTING YARN QUALITY:

AN INDISPENSABLE TOOL FOR COTTON BREEDERS

E. F. Hequet and B. Kelly

Fiber and Biopolymer Research Institute
Plant and Soil Science Department
Texas Tech University, Lubbock, Texas, U.S.A.

Cotton breeders face the task of developing cultivars that will perform well in the
field, at the gin, and in textile processing. Predicting the process ability of the raw
material is probably the most challenging task. Indeed, producing yarn from each
entry in a breeding program is not possible because of the limited quantity of lint
available. Hence, how could we predict the industrial yarn quality of a breeding line
without spinning the lint into yarn? To answer this question, two sets of commercial
cotton bales were selected based on their distinct physical properties. The results
obtained show that the combination of HVI and AFIS data allows us to predict quite
accurately yarn quality (ring spun yarn) for commercial bales. Such models, if
confirmed on an independent set of samples could be invaluable for the cotton
breeding industry.

Cotton breeders face the task of developing cultivars that will perform well in the
field, at the gin, and in textile processing. Predicting the process ability of the raw
material is probably the most challenging task. Indeed, producing yarn from each
entry in a breeding program is not possible because of the limited quantity of lint
available. In addition, the cost would be prohibitive, hindering the cotton breeding
efforts of both private and public sectors. Hence, how could we predict the industrial
yarn quality of a breeding line without spinning the lint into yarn?

In order to make cultivar selections, breeders must first know which fiber quality
attributes produce high quality yarns for their targeted market (May and Jividen,
1999, Meredith et al., 1991, Meredith, 2005). They must strive to deliver fibers that
perform better in textile manufacturing. This is critical for effective competition with
man-made fibers and with international growths of cotton. Knowing what fiber quality
attributes have the greatest potential to produce high quality ring spun yarns is also
critical for spinning mills.

Modern fiber quality evaluation instruments provide repeatable measurements of a

full spectrum of fiber quality that can potentially be related to yarn quality. Spinning
and weaving are performed at high speeds. This could result in excessive end
downs for cottons with poor tensile properties. Because breakages may cause work
stoppages and slowed production, there has been and emphasis on predicting ring-
spun yarn tensile properties. It has been reported that fiber quality attributes
measured by High Volume Instrument (HVI) have a strong relationship with ring spun
yarn tensile strength. For example, El Mogahzy et al. (1990) showed that cottons
with improved Upper Half Mean Length (UHML), tenacity, micronaire, and length
uniformity can be used to produce yarns with a higher break skein factor. Using the
Advanced Fiber Information System (AFIS), Krifa et al. (2001) reported that yarn
tenacity was affected by the amount of seed coat fragments in the raw material.

Predicting yarn tensile properties is important but offers an incomplete picture of how
fiber quality impacts yarn quality. High quality yarns should also have a low number
of imperfections such as thin places, thick places, and neps. The number of thin and
thick places in the yarn has been shown to be related to HVI Micronaire and
Uniformity. Relationships of these two yarn parameters with AFIS Fineness and
Sutter-Web array Short Fiber Content (SFC) have also been documented
(Thibodeaux et al., 2008). The literature shows that while HVI measurements are
adequate for predicting yarn tensile properties, they are inadequate for yarn
evenness related parameters.

Therefore, one of the basic objectives of our research is to study the spinability of
cottons (carded and combed ring spun yarns), and to derive from this research
recommendations as to which fiber parameters need to be improved through
breeding, agronomic practices, and/or material handling (harvesting and ginning).
The first step for this type of research is to evaluate the advantages and limitations of
the two main cotton fiber testing equipments, i.e., High Volume Instrument (HVI) and
Advanced Fiber Information System (AFIS) for predicting yarn quality. As mentioned
earlier, producing yarn from each entry in a breeding program is not possible
because of the limited quantity of lint available and the cost. To alleviate this
technical difficulty, a solution could be to predict yarn quality from a selected number
of fiber properties.

MATERIAL AND METHODS

To determine the relationships between fiber properties and yarn quality, two sets of
commercial cotton bales were selected based on their distinct physical properties. A
representative sample of approximately 50 kg was taken from each bale for
processing. Raw cotton, chute, card sliver, DI sliver, and DII sliver lint samples were
collected. The cotton samples were tested on High Volume Instruments (HVI 900A,
Uster, Knoxville, TN), with 10 length and tenacity measurements and 4 micronaire
measurements. They were also tested on the Advanced Fiber Information System
(AFIS, Uster, Knoxville, TN), with 5 replications of 3,000 fibers. Tables I through III
summarize maximum, minimum, and average values of the physical properties of
these samples.

First experiment: A total of 8 bales were selected for this experiment. The results
obtained on 4 of them will be presented in this paper. A complete HVI and AFIS fiber
quality profile was done for each bale. Then, carded and combed ring spun yarns
(Suessen Elite 1000) were produced: 12Ne through 30Ne with a step of 2Ne were
produced for the carded yarns, and 18Ne through 36Ne with a step of 2 for the
combed yarns with 3 combing noils levels (15, 20, and 25%).
Table I. HVI data on the 4 bales selected.

Code Mic. Length Unif. Strength Elon. Rd +b

inch % g/tex % %
09BP 3.5 1.18 82.7 29.2 9.8 81.2 8.6
09BS 3.2 1.17 81.9 28.4 9.8 78.8 8.8
11VCP 4.3 1.19 81.8 29.6 7.4 83.1 8.1
11VCS 4.2 1.17 80.6 29.8 7.3 81.5 8.8

Table II. Main AFIS data on the 4 bales selected

Code Neps L(n) SFC(n) VFM Fine. IFC MR

Count/g inch % % mtex %
09BP 333 0.76 30.6 1.71 152 8.8 0.81
09BS 566 0.74 32.0 3.38 148 9.9 0.81
11VCP 332 0.79 25.9 0.97 161 7.6 0.87
11VCS 432 0.79 25.9 0.94 157 8.4 0.87

Second experiment: Seventy four commercial cotton bales were selected. For each
bale a complete fiber quality profile was done (HVI, AFIS). Then, carded and
combed ring spun yarns (Suessen Elite 1000) were produced (30Ne, knitting twist).
The summary statistics are presented Table III. All HVI fiber quality parameters
represent a large range of fiber properties.
Table III. HVI and AFIS (5 reps) fiber properties obtained on samples from 74
commercial bales. The shaded region shows the opening and card waste
as well as the noils percentages (constant settings of the comber).

Average Minimum Maximum

Micronaire 3.61 2.61 4.50
UHML, inch 1.19 1.11 1.26
Uniformity, % 82.6 79.7 84.4
Strength, g/tex 30.0 24.5 34.8
Elongation, % 9.0 5.4 11.5
Reflectance, % 80.3 73.0 85.3
Yellowness 8.0 6.4 9.7
Neps, count per Gm 447 239 1013
Mean length by weight, inch 0.99 0.89 1.07
Length by weight CV, % 37.3 32.6 42.6
UQL by weight, inch 1.23 1.15 1.32
Short Fiber Content by weight, % 10.3 6.2 16.6
Mean length by number, inch 0.75 0.62 0.86
Length by number CV, % 56.4 48.5 65.7
Short Fiber Content by number, % 31.1 21.7 43.0
Visible Foreign Matter, % 2.24 0.73 4.88
Seed Coat Neps, count per gram 21 8 38
Fineness, mtex 150 131 172
Immature Fiber Content, % 9.2 6.3 12.6
Maturity Ratio 0.82 0.74 0.91
Standard Fineness, mtex 183 168 203
Opening waste, % 2.86 1.55 5.04
Card waste, % 3.91 2.51 5.77
Noils, % 18.0 12.6 26.8

RESULTS AND DISCUSSION

First experiment: The examination of Table I shows that the bales 09BP and 09BS
are quite similar. Both have a low micronaire and a very good UHML. Based on HVI
data only, the differences observed between these two bales should not translate
into large differences in ring spun yarn quality. The AFIS data (Table II) show that
bale 09BS has significantly more neps than 09BP, a shorter mean length by number,
and a higher trash content. The bales 11VCP and 11VCS have a better micronaire
than the two previous bales, and a very good UHML. Again, based on HVI only, the
differences observed between these two bales should not translate into large
differences in ring spun yarn quality. The AFIS data show that 11VCS has more
neps than 11VCP. As discussed earlier, yarn tensile properties can be reasonably
well predicted from HVI data. Figures 1 and 2 show that, as expected, the carded
yarn work-to-break for the two sets of two bales are very similar across the full range
of yarn counts produced. Similar results were obtained on combed yarns (data not
shown).

Carded
1400
Work-tobeak, cN.cm

1200
1000
800
600
400
200
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
09BP 09BS

Figure 1. Work-to-break vs. Yarn count (carded ring spun yarn, knitting twist) for
the bales 09BP and 09BS

Carded
1400
Work-tobeak, cN.cm

1200
1000
800
600
400
200
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
11VCP 11VCS

Figure 2. Work-to-break vs. Yarn count (carded ring spun yarn, knitting twist) for
the bales 11VCP and 11VCS
The situation is quite different for yarn evenness related parameters as shown in
Figures 3 through 6. For the two sets of two bales, there are clear differences
between the bales for each couple in terms of both Uster CVm and Uster IPI (total
number of imperfections). Figures 3 and 4 show parallel lines, the Uster CVm of
09BS is always higher than the one of 09BP. The difference between the two cottons
is nearly constant. The same is true for the second set of bales and for the combed
yarns (data not shown). For the total imperfections, the differences between the two
bales within a set increase as the yarn count increases. Finer is the yarn larger is the
difference. The same was observed on combed yarns (data not shown). This
illustrates the limitations of HVI quite well. Indeed, yarn evenness is not predictable
based on HVI data alone.

Carded
18
17
16
CVm, %

15
14
13
12
11
10
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
09BP 09BS

Figure 3. Uster CVm vs. Yarn count (carded ring spun yarn, knitting twist) for the
bales 09BP and 09BS
 Carded
18
17
16

CVm, %
15
14
13
12
11
10
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
11VCP 11VCS

Figure 4. Uster CVm vs. Yarn count (carded ring spun yarn, knitting twist) for the
bales 11VCP and 11VCS

Carded
800
700
IPI, count/km

600
500
400
300
200
100
0
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
09BP 09BS

Figure 5. Uster IPI (total imperfections) vs. Yarn count (carded ring spun yarn,
knitting twist) for the bales 09BP and 09BS
 Carded
800
700

IPI, count/km
600
500
400
300
200
100
0
10 12 14 16 18 20 22 24 26 28 30 32
Count, Ne
11VCP 11VCS

Figure 6. Uster IPI (total imperfections) vs. Yarn count (carded ring spun yarn,
knitting twist) for the bales 11VCP and 11VCS

Second experiment: Our goal with this second experiment was to investigate the
potential of the AFIS as a tool to predict yarn quality. For this, a set of 74 commercial
bales representing a large range of fiber properties especially for micronaire and
AFIS maturity was selected (Table III). Table IV and Figures 8 through 11 show the
summary of the multiple regression analyses (backward stepwise regression) for ring
spun carded yarns (30Ne with a knitting twist). The main yarn quality parameters
were predicted from the HVI and AFIS fiber properties. We choose a knitting twist
because we hypothesize that the weaving twist produces yarns that are less
sensitive to fiber quality than the knitting twist. In addition, we also included in some
models the Noils percentages, and the opening and carding wastes. The reason
being that the AFIS samples are quite small (15,000 fibers for the length
distributions, and 2.5 grams for the trash content determination) making it difficult of
obtain a fully representative sample of the bale being tested. Hence, we substituted
the noils percentages for the AFIS short fiber content in some models and the
opening and card waste to the AFIS visible foreign matter content. Also, as regular
classing office data does not report bundle elongation because of the lack of
calibration cottons for elongation, models were built both with and without HVI
bundle elongation.
Table IV. Seventy four commercial bales – Ring spun yarn carded 30Ne – Knitting
Twist - Multiple regressions summary (backward stepwise) – Adjusted R2

HVI HVI AFIS + AFIS + HVI + Noils +

without With HVI Opening waste + Card
elongation elongation waste
CSP 0.657 0.657 0.881 0.909
Elongation 0.615 0.741 0.872 0.887
Tenacity 0.642 0.860 0.877 0.899
Work 0.489 0.511 0.717 0.746
CV% 0.498 0.498 0.833 0.909
Thin places 0.475 0.475 0.771 0.814
Thick Places 0.333 0.333 0.767 0.875
Neps 200% 0.398 0.398 0.789 0.806
Hairiness 0.635 0.635 0.694 0.865

The results are quite clear, adding AFIS data to HVI data improves the predictions of
all yarn quality parameters measured, especially yarn evenness related parameters.
Substituting noils and processing wastes for short fiber content and visible foreign
matter also improves the quality of the predictions but, with the exception of
hairiness, this improvement is somewhat modest. This modest improvement
indicates that the AFIS measurements of short fiber content and VFM are quite good
in predicting yarn evenness but that they can probably be improved with a larger
sample size. The relationship between AFIS short fiber content and combing noils is
surprisingly good (Figure 7). Indeed, let us assume a cotton sample with the
following fiber properties: 10% short fiber content by weight, mean length by number
of 0.75 inch, and fineness of 150 mtex. If 15,000 fibers have been measured with the
AFIS, it would correspond to a weight of 42.9 mg. Hence, about 4.3 mg of those are
short fibers. This is to compared to 100 pounds of lint processed through our short
staple spinning facility with an average noils percentage of 18% (about 1.8 kg of
noils for 10 kg of lint being combed).
 28
26 2
R = 0.766
24

Noils, %
22
20
18
16
14
12
4 6 8 10 12 14 16 18
AFIS SFC(w), %

Figure 7. Combing noils vs. AFIS short fiber content by weight

18.0 18.0
Adjusted R2 = 0.498 Adjusted R2 = 0.833
17.5 17.5
Observed values

Observed values

17.0 17.0
16.5 16.5
16.0 16.0
15.5 15.5
15.0 15.0
14.5 14.5
14.0 14.0
14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0 14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0
CVm predicted values, % CVm predicted values, %

18.0
Adjusted R2 = 0.909
17.5
Observed values

17.0
16.5
16.0
15.5
15.0
14.5
14.0
14.0 14.5 15.0 15.5 16.0 16.5 17.0 17.5 18.0
CVm predicted values, %

Figure 8. Uster CVm Observed values vs. Predicted values (top left = predictors
HVI only, top right = predictors HVI + AFIS, bottom = predictors HVI +
AFIS + Noils + Card & Opening wastes). It should be noted that
redundant variables were systematically removed from the models.
 60 60
55 Adjusted R2 = 0.475 55 Adjusted R2 = 0.771
50 50
Observed values

Observed values
45 45
40 40
35 35
30 30
25 25
20 20
15 15
10 10
5 5
0 0
0 5 10 15 20 25 30 35 40 45 50 55 60 0 5 10 15 20 25 30 35 40 45 50 55 60
Thin places predicted values, count/km Thin places predicted values, count/km

60
55 Adjusted R2 = 0.814
50
Observed values

45
40
35
30
25
20
15
10
5
0
0 5 10 15 20 25 30 35 40 45 50 55 60
Thin places predicted values, count/km

Figure 9. Uster Thin places (-50%) Observed values vs. Predicted values (top left
= predictors HVI only, top right = predictors HVI + AFIS, bottom =
predictors HVI + AFIS + Noils + Card & Opening wastes). It should be
noted that redundant variables were systematically removed from the
models.
 420 Adjusted R2 = 0.333 420 Adjusted R2 = 0.767
380 380
Observed values

Observed values
340 340
300 300
260 260
220 220
180 180
140 140
140 180 220 260 300 340 380 420 140 180 220 260 300 340 380 420
Thick places predicted values, count/km Thick places predicted values, count/km

420 Adjusted R2 = 0.875

380
Observed values

340
300
260
220
180
140
140 180 220 260 300 340 380 420
Thick places predicted values, count/km

Figure 10. Uster Thick places (+50%) Observed values vs. Predicted values (top left
= predictors HVI only, top right = predictors HVI + AFIS, bottom =
predictors HVI + AFIS + Noils + Card & Opening wastes). It should be
noted that redundant variables were systematically removed from the
models.
 700 700
Adjusted R2 = 0.398 Adjusted R2 = 0.789
600 600
Observed values

Observed values
500 500
400 400
300 300
200 200
100 100
100 200 300 400 500 600 700 100 200 300 400 500 600 700
Neps 200% predicted values, count/km Neps 200% predicted values, count/km

700
Adjusted R2 = 0.806
600
Observed values

500
400
300
200
100
100 200 300 400 500 600 700
Neps 200% predicted values, count/km

Figure 11. Uster Neps (+200%) Observed values vs. Predicted values (top left =
predictors HVI only, top right = predictors HVI + AFIS, bottom =
predictors HVI + AFIS + Noils + Card & Opening wastes). It should be
noted that redundant variables were systematically removed from the
models.

CONCLUSIONS

In conclusion, the combination of HVI and AFIS data allows us to predict yarn quality
(ring spun yarn) for commercial bales quite accurately. A better measurement of both
short fiber content and trash content should allow us to predict nearly perfectly yarn
quality of ring spun yarns (we used noils percent instead of AFIS SFC and opening
plus carding wastes instead of AFIS VFM in this study). Such models, if confirmed
on an independent set of commercial bales could be invaluable for cotton breeders.

Breeder samples often contain more trash particles than commercial bales because
they are ginned on small scale ginning equipment. The presence of excessive trash
content in the lint could have a very negative effect on the quality of predictions for
all yarn evenness related parameters. This is due to both less accurate AFIS
readings with trashy cottons and more trash remaining in the lint after carding
(translating in poorer yarn evenness). It is therefore essential to expand this research
to include breeder type samples if we want to breed for improved yarn quality and
remain competitive on the international market.
REFERENCES

El Mogahzy, Y. E., Broughton, R., and Lynch, W. K. 1990. A Statistical Approach for
Determining the Technological Value of Cotton Using HVI Fiber Properties. Textile
Research Journal, 60(9), 495-500. doi:10.1177/004051759006000901

Krifa, M., Gourlot, J.-P., and Drean, J.-Y. 2001. Effect of Seed Coat Fragments on
Cotton Yarn Strength: Dependence on Fiber Quality. Textile Research Journal,
71(11), 981-986. doi:10.1177/004051750107101108

May, O.L. and Jividen, G.M. 1999. Genetic Modification of Cotton Fiber Properties as
Measured by Single and High Volume Instruments, Crop Sci. 39, 328-333

Meredith, W. R., Culp, T. W., Robert, K. Q., Ruppenicker, G. F., Anthony, W. S., and
Williford, J. R. 1991. Determining Future Cotton Variety Fiber Quality Objectives.
Textile Research Journal, 61(12), 715-720. doi:10.1177/004051759106101203

Thibodeaux, D., Senter, H., Knowlton, J. L., Mcalister, D., and Cui, X. 2008. The
Impact of Short Fiber Content on the Quality of Cotton Ring Spun Yarn. The Journal
of Cotton Science, 12, 368-377.