Estimating The Quantity of Usable Fibers in Baled Cotton

Download as pdf or txt
Download as pdf or txt
You are on page 1of 7

705

Estimating the Quantity of Usable Fibers in Baled Cotton


ROBERT A. TAYLOR
Cotton Quality Research Station, ARS, USDA, Clemson, South Carolina 29631, U.S.A.
ABSTRACT
The accuracy of estimating weight loss in processing cotton is improved with in-
strument based cotton classification measurements when compared to conventional
cotton classification. Two new methods of rapidly estimating trash content were in-
vestigated as candidates for use in instrument classification. Nonlint losses were also
measured with a Shirley analyzer and a conventional carding machine. Agreement
with carding loss ranged from an 2of 36 for conventional grading to an r
r 2of 77 for
Shirley analyzer measurements. When nonlint content data were combined with other
fiber data, the Shirley analyzer method explained 90% of the variation in carding loss.
When comparing the two new rapid methods, the trash monitor was a better estimator
of card loss than the scanning trashmeter, but when combined with fiber data, the
two were equal. Both of the rapid trash measuring instruments are commercially
available, and both can be used on a routine basis to improve mill performance.

The amount of foreign matter remaining in cotton (based on fibrogram), 3.2 mm gage strength and don-
after ginning largely depends on the condition of the gation, and micronaire reading of fineness. Until a
cotton at harvest time and the kind and amount of suitable trashmeter or nonlint oon mentt
cleaning and drying equipment used during ginning. is developed for these systems, HVI operators arc pro-
Traditionally, cotton has been classed for market pur- vided a set of black and white photographs to use as
poses by describing quality in terms of classer’s grade, reference for coding leaf content in cotton samples.
staple length, and micronaire measurement of fineness. Two approaches are being pursued for potential use
Within this system, grade is determined by considering with HVI systems to mechanize leaf or trash grading:
lint color, foreign matter content such as leaf particles, One approach uses video scanning trashmeters, which
and preparation or condition of the sample. The make a two-dimensional surface scan or image analysis
USDA-AMS Cotton Division prepares and distributes of trash particles by counting dark spots observed with
sets of physical grade standards to be used as references a black and white television camera [5]. Methods for
when determining grade by this method. Adjustments calibrating and operating video scanning trashmeter
to grade classification are made when excessive have been developed [6] but suitable grade reference
amounts of foreign matter other than leaf are observed, cottons have not been developed. A second approach
or when special conditions exist that may cause di~- is to use a high speed lint cleaning-type microdust and
culties in processing [8]. A major purpose for grade trash monitor (MTM) for a potential measurement of
classification is to provide information to cotton buyers nonlint content. Initially, the trash monitor was in-
concerning the quality and quantity of usable fibers in tended to rapidly clean a small sample of cotton with
’ each bale. As a result, classer’s grade is the most im- an efficiency of at least 80% of the cleaning provided

portant factor in setting market discounts or premiums. by a Shirley analyzer. Eventually, the MTM is expected
Numerous advances in instrumentation now provide to provide measurements that are directly comparable
better tools to assist in estimating cotton quality. The to Shirley analyzer results. A prototype version of the
cotton colorimeter [3] used for many years by the MTM trash monitor system has shown promise by
USDA-AMS Cotton Division to select cottons for grade processing 20 gram samples in 25 to 30 seconds with
standards has been accepted throughout the textile in- better repeatability than normally observed using a
dustry for measuring cotton lint color. Recently, cotton Shirley analyzer [4, 7].
colorimeters were included in high volume instrument Our objectives here were to determine relative con-
(HVI) systems developed to mechanize grading of cot- fidence in predicting cotton processing loss by various
ton for marketing purposes. These new instrument grading methods and to develop procedures to measure- °

systems also include measurements of fiber length processing waste.


706

Experimental Procedures 4. Shirley analyzer (using a two-pass, 100 gram pro-


cedure [lb]): (a) visible weight loss (percent) and (b)
An experimental set of cottons was selected by the total weight loss (percent).
USDA-AMS Cotton Division Standards Section, 5. Peyer length analysis (Fibroliner AL k0 [2]): (a)
Memphis, Tennessee, during the 1982 cotton harvest- short fiber content (percent).
ing season. Except for grade 21, the distribution of the 6. Processing loss (USDA-Clemson, 50 gram sample
grades selected was similar to cotton quality in the [9]): (a) total weight loss in carding with working flats
market place (Table I). A special effort was made to
(percent) and (b) total weight loss in carding with sta-
include several high grade cottons with other fiber
tionary flats (percent).
properties. Irregular and special condition cottons were
deliberately excluded. Additionally, there were no cot- TRASHMETER METHODS
tons that had been downgraded for excessive grass,
bark, or other conditions. Each of the 68 cotton samples The video scan trashmeter automatically compresses
was given a leaf grade by a committee of classing spe- each sample against an observation window. When the
cialists. sample has reached the desired 24.5 Newton compres-
Six different tests related to fiber properties and mill sion, there are simultaneous readings of trash area and
processing performance were performed on each cotton trash count..
sample. Prior to each test, the samples were exposed A schematic of the trashmeter’s optical configuration
for at least 12 hours to the recommended laboratory is shown in Figure 1. Each trash area measurement is
environment [I a]. The testing included subsampling an indication of the total number of pixels (small spots)
the proper weight of cotton from the larger supply to that are darker than a predetermined threshold level.
insure fresh cotton for subsequent experiments. Data
were recorded as follows for each of the six tests:
1. High volume instrument (motion control sys-
tem) : (a) Micronaire fineness, (b) length (upper half
mean and uniformity index), (c) strength (3.2 mm
gage), and (d) color (reflectance and yellowness).
2. Video scan trashmeter (motion control system):
(a) trash area in (pixelsr-’, (b) trash count in ( l Ox
number of particlesp-’, and (c) scan code and predicted
trash in percent were calculated by .

and

~ 3. MTM (ppm Inc. system using a single pass, 20


gram procedure): (a) visible weight loss (percent) and I .

(b) total weight loss (percent). FyGURE 1. Sketch of optical scanning trashmeter geometry.

TABLE I. Leaf grade distribution of 68 test cottons.

’ Average leaf = 3.71 (range =


158%).
707

For this experiment, the threshold level was 30% lower


in reflectance than the average reflectance of the sample
being tested. Trash count measurements are reported
as the square root of 10 times the total number of
par-
ticles encountered while scanning the sample. The
coefficients shown in Equation 1 were previously es-
tablished to provide a good agreement (r~ 0.94) be-
=

tween scan code and classer’s leaf grade when the .


trashmeter was used to observe a grade standard biscuit
set [6]. Similar statistical methods were also used pre-
viously to establish the coeflicients in Equation 2 for
good agreement (r2 = 0.91 ) between predicted trash
and Shirley analyzer visable nonlint content [5].

MTM METHODS
The MTM produces high speed physical separation
of lint from nonlint material found in each cotton
sample. Weights of uncleaned and cleaned lint are
taken to determine the total percent loss. The weight
of visible trash catch is recorded to determine visible
percent loss. Differences between visible and total loss
are considered to be microdust, residual moisture, and
other volatiles lost to the air handling system. The FIGURE 2. Sketch of microdust and trash
MTM requires a 20 gram sample and completes pro- monitor geometry (modei H).

cessing in 25 to 30 seconds. Since MTM was initially


designed to remove trash with minimum fiber damage
and to meet the high sample processing rate of HVI in predicting processing loss from short fiber measure-
systems, its lint cleaning efficiency has not yet reached ments.
that of a conventional carding machine or Shirley an- .

alyzer. A small quantity of trash remains in the lint Experimental Results


and some lint is found in the visible catch.
The trash monitor uses counter air flow in vertically Except for the MTM experiments, only one repli-
cation of each measurement was recorded, but within
oriented chambers to return the light weight cotton
each replication, a sufficient number of observations
fibers from the dust and trash outlet stream (Figure 2).
were recorded to establish the desired measuring con-
This approach causes some light trash particles to re-
turn with lint and some lint attached to heavy trash
fidence. Results show that the cottons selected provided
a wide range of physical properties (Table II). The
particles to be lost to the trash hopper. The MTM sys- range
in fiber properties (excluding varied from a low
tem was included in this study to compare the precision trash)
of using an aerodynamic classification method to es- B of 8% for fiber length uniformity
index to a high of
’ 79% for yellowness (+b units). Observed trash or non-
timate processing loss.
lint indicators varied from a low 133% for scan cvde
with the video scanning trashmeter (Equation 1 ) to a
SHORT FIBER METHOD ,

high of 307% for predicted trash content using the same


The Peyer Fibroliner and AL 101 length analysis sys- instruments (Equation 2). The average count of short
tem produces length distribution measurements by fibers ranged from 0.59% when fibers shorter than 5
aligning all fibers in the specimen with one end even. mm were selected to 13.35% when the limiting length
Mass measurement scans are made using capacitance was increased to 12.5 mm. Total carding weight loss
methods with the aligned fibers suspended between two recorded by the manufacturing waste experiments [9]
layers of plastic. The method provides observations ranged from 3.77% for the replication using a card with
over the entire specimen length. A more complete de- nonloading stationary flats to 9.75% for the replication
scription of operating procedures and accuracy was de- using working flats.
scribed by Bargeron [2]. Measurements with this system Both instrument methods considered for coding
were included to determine the degree of improvement trash or nonlint content in cottons produced a certain
708

TABLE II. Test results for cottons studied.

0
Number of observations/cotton/replication. &dquo;The range is the difference between the maximum and minimum expressed as a percent
of the average..

amount of quantity error or level bias with the exper- on the amount of weight loss to expect during cleaning
imental cottons. Comparing the average measurement and preparation for spinning. The largest losses usually
result (Table II) proved interesting. Total nonlint losses occur during carding. Carding loss data reported here
during two replications with the MTM system were for working flats (9.75%; Table II) were unusually high
4% and 3196 lower than Shirley analyzer measurements. because the small sample facility operates with un-
Similar measurements for visible loss showed the MTM loaded flat strips. Simple correlations with carding loss
system produced only 58% of the trash removed during data ranged between a low value of 0.58 for the visually
Shirley analysis. Visable loss estimates with the video coded leaf grade for carding with stationary flats to a
scanning system (predicted trash) produced an average high value of 0.90 for Shirley analyzer total nonlint
value 1196 higher than actually measured during the content when carding with working flats (Table III).
same Shirley analysis. When the video scanning system In general, the lint cleaning methods of estimatirg
was calibrated against official grade standard samples carding losses were better than the surface observation
(scan code) it produced a trash estimate 1.31 grades methods. All instrument methods of coding cottons
(35%) below the average grade observed when classing were better than the visual method.
the experimental cottons. For these experiments, we Some investigators have suggested that lint lost in
see the MTM system nonlint estimates were consis- processing consists mainly of short fibers, so cottons
tently lower, whereas the video scanning system esti- with higher short fiber content should sustain larger
mates were consistently higher. than previously ac- weight losses. Correlation coefficients between short fi-
cepted methods. ,
ber content and processing weight loss appear to agree
A major purpose for using trash or leaf grade to code with the hypothesis, since all correlation coefficient
cottons is to provide buyers with accurate information were positive (see Table IV). The level of agreement
709

TABLE HI. Simple correlation coefficients among trash grading methods and carding km observed for 68 expaimental ooaoot.’

I
Based on processing 3 replications using 50 g samples each. &dquo;The numbers in p~nthesia repraent Mpbomon. .

TABLE IV. Simple correlation coefficients among fiber measure- with short fiber data was quite poor, however. The
ments and observed carding loss for 68 experimental cottons.
highest correlation coefficient was 0.50 for 5 mm fiber
with losses by a card using working flats.
The single measurement having the highest precision
to predict carding loss was Shirley analyzer total nonhnt
(rl 0.81), while the poorest method was leaj grade
=

(r2 0.36, Table V). The best performance by the


=

MTM system was (r2 0.69) while carding with work-


=

ing flats, and the best performance by the video scan-


ning trashmeter system was (r2 0.53) when carding
=

with stationary flats. All regression coefficients im-


proved as HVI fiber and Peyer short fiber data were
I
Based on 4 replications using the Peyer Fibroliner AL 101 with added. Video scanning system predictions of carding
card blended samples. loss with HVI fiber data were approximately equal

TABLE V. Multiple regression coefficients (r’) for predicting cotton weight loss in carding
with different grading methods in combination with other 6ber measurements.

’ HVI data added included upper half mean length, length uniformity index, Va gage strength, micronaut fineness, and corimeter
reflectance..
710

in precision to the same combination with the did not appear as a significant parameter. The use of
MTM system. short fiber content data improved the correlation coef-
Standard error estimates for the MTM and video ficients for all cases, as much as 0.14 partial r2 im-
scanning predictions were compared with the leaf provement to predict card loss (with working flats) us-
grading system and Shirley analyzer nonlint content ing a video scanning trashmeter. This finding suggested
predictions of carding weight loss (Table VI). Except that short fiber content is an important factor in esti-
for one case, the two instrument methods were better mating carding loss. Also, the length uniformity index
than the visual leaf grading system but not as precise provides similar information but with less precision.
as estimations with Shirley analyzer nonlint content. Regression analyses using different trash measure-
The exception occurs when other HVI data are added ments in combination with HVI data suggest the fol-
to the video scanning measurements for estimating lowing equations for estimating percent weight loss
weight loss with stationary flats. At present, we do not (loss) in carding with working flats:
have a technical understanding of improved precision
With Shirley analyzer total nonlint content data (SAT):
for video scanning trashmeters in combination with
.

HVI measurement. -

Discussion
With MTM total nonlint content data (MTMT):
Normally, the standard error of prediction (SEP) is
considered the true judge of prediction model accuracy.
SEP values are usually determined on operational-type
instruments using a large number of cottons different
from those used to establish the model, so accurate With classer’s leaf grade data (leaf):
evaluations for SEP values should include errors caused
by instrument stability and calibration differences, op-
erator influence, and testing laboratories. Such evalu-
ations were beyond the scope of this report. With video scan code data (code):
To enhance acceptance of a new grading system in
cotton trading, the new trash grade or nonlint mea-
surement must provide a high degree of accuracy in
predicting cotton processing weight loss and spinning
performance. Regression analyses were performed to Acceptance of HVI trash or nonlint measurements
define the relative agreement between the processing in marketing cotton is most likely to be based on con-
. loss and the proposed trash grade indicators. It is fre-
quently helpful to include other fiber measurements ’
Measured variables carding weight loss ), loss Shirley analyzer
(
(such as short fiber content) to estimate the value of total nonlint content (SAT), microdust and trash monitor total nonlint
content (MTMT), HVI length uniformity index (UI), and Colorimeter
considering such new measurements in a bale testing reflectance (Rd) are in percent. Variables : Classer’s leaf grade ),
leaf
(
program. When card loss regressions were made with video scan trashmeter ),code leaf grade and Colorimeter yellowness
(
short fiber data included, length uniformity index (UI) (b) are in coded units.

TABLE VI. Standard error of the mean in estimating weight loss in carding by different methods.’

I
All data are in percent of cotton weight loss during carding.
711

sistency between instruments, which includes not only ciples involved, which closely resemble the mechanics
the precision but the accuracy of individual instru- in carding. The level of MTM conbdena in prodicting
ments. Nonlint testers are not easily adjusted or cali- loss was unacceptably low, however (r2 ranged from
brated and sometimes produce wide level differences 0.61 to 0.84, Table V). Many minor MTM design fea-
between indentical systems (as indicated by the level tures have been changed during its development, but
difference between replications 1 and 2 in this report). none greatly altered its performance. Additional basic
Video scanning trashmeters must be calibrated with changes are needed to increase ckaning efficiency and
a series of bale samples having grades or nonlint content introduce more processing mechanics similar to those
established by specialists. When scanning devices are in carding.
calibrated against grade standard guide box samples, The most likely avenues for improving the precision
as with Equation 1, resulting trash grades are 1 to 2 of video scanning methods are increased numbet (or
grades lower than comparable classer’s values (our data area) of sample observations or revised bale sampling
averaged 1.31 grades lower). procedures to provide more homogenous samples for
Correlation between the video scanning trashmeter testing.
results and the leaf grade values for the 68 cottons de-
scribed here produced the following trashmeter equa-
tion : Literature Cited
scan code =
0.159 + 0.662 (area)°~6 1. American Society for Testing Materials, Committee D-
13 on Textile Materials, ASTM designations (a) 1776-
1
-1.58 (area/count)2 . (7) 79 and (b) 2812-76.
An unbiased calibration for scanning-type trashmeters 2. Bargeron, J. D., Preliminary Investigation of the Length
can be made by selecting more numerous samples of Measurement of Cotton Fibers with the Peyer Texlab
raw cotton from bales representing a complete range System, Textile Res. J. 55, 121-123 (1985).
3. Nickerson, D., Color Measurement of Standards for
of leaf grades.
Grades of Cotton, Textile Res. J. 16,441-449 (1946).
4. Shofner, F. M., Lalor, W. F., and Hanley, J. H., A New
Conclusions and Recommendations Method for Microdust and Trash Measurement in Bale
or Processed Fiber, Textile Res. J. 33, 78-89 (1963).
Instrument based trash or nonlint grading is essential
5. Taylor, R. A., Using High Speed Image Analysis to Es-
for estimating weight loss in carding. The MTM system
timate Trash in Cotton, J. Eng. Ind. 107 (2), 206-219
nonlint content estimates are consistently biased lower
(1985).
than expected (Shirley analyzer data), and the video 6. Taylor, R. A., A Calibration Reference Tile for Optical
scanning trashmeter estimates are consistently biased Image Type Cotton Trashmeters, in "Proc.1984 Cotton
higher than expected (Shirley analyzer and classing Production Research Conference," 1984, pp. 351-358.
grade data). Short fiber content is a very important 7. Taylor, R. A., Foreign Matter Separation and Measure-
measurement for estimating loss in carding when using ment for Cotton Quality Classification and Estimation
working flats, but it is less important for cards with of Spinning Behaviour, Melliand Textilber. 65, 511-514
stationary flats. Use of a short fiber measurement in (1984).
8. USDA Handbook No. 566, The
Claskfication of Cotton,
HVI systems should be seriously considered.
When comparing trash measuring instruments for
use with HVI systems, the MTM was clearly a more

precise estimator of carding loss. Superiority of the


9. USDA Handbook No. 594,
U.S. Department of Agriculture, Washington, DC.
Cotton Testing Service, U.S.
Department of Agriculture, Washington, DC.
6-
MTM over video scanning is due to the cleaning prin- Manuserip remved A ugum 2. J985; aao~tad April 7, 1986.

You might also like