Fundamental research on spiking,recovery and understanding seed coat nep counts in AFIS analysis of pre-opened cotton

Abstract

Understanding seed coat fragment (SCF) spiking results in advanced fiber information systems (AFIS) analysis of seed coat neps (SCN) in ginned cottons is confounded by the opening of entangled fibers in the instrument’s fiber individualizer. This may influence seed coat tissue fragmentation and recovery since a high degree of machine-fiber interaction is required to individualize entities for sensing. In this paper, slivers were pre-opened in the AFIS followed by manual cleaning, spiking, and AFIS analysis. A protocol was developed to spike the pre-opened slivers. The percent recovery of the spiked entity was dependent on the species and cultivar of the cotton used to prepare the slivers. The lowest recovery was with Pima fibers. Also, the recovery improved with the increase in length of the fibers biologically attached to the SCF surface. Delinted seed coat fragments produced the lowest recovery. Seed coat fragments carefully removed from ginned lint and added to the processed slivers gave the highest recovery. Averaged SCN recoveries from two AFIS units ranged from 26 to 100% (theoretical). These results helped to explain why the AFIS analysis of SCN counts in processed cotton is lower than by the microscopic analysis.

Keywords

AFIS machine-fiber interactions percent recovery seed coat fragments spiking

Seed coat fragments are one of the most difficult impurities, like plastics, bark and others, to remove from yarn and fabrics.^1–3 A review of seed coat fragment (SCF) formation and removal during the period 1950 to 1987 described how the fragments were created and explained the variation of their concentration in cotton during harvesting, ginning and processing.⁴ Traditional neps are knots of tangled fibers only, and SCF are small pieces of seed coat tissue with or without attached fibers (see Table 1, terms and nomenclature). The possible sources of SCFs are motes, immature seeds and mature seeds. Motes and immature seeds do not develop hard seed coats. These soft seeds and mature seeds are damaged during mechanical operations. Most of the damage occurs when the fibers were separated from the seed at the gin stand.⁵ The seed coat is weakest at the rounded or chalazal end of the seed and peeled more easily than the remainder of the seed coat. Hand ginning of mechanically harvested cotton gave the lowest level of SCFs (eight counts per gram lint); whereas roller ginning produced 14 per gram.⁶ Seed-coat fragments and neps together made up the majority of defects in the yarn, with one study demonstrating SCFs accounting for 27.6% of yarn defects in 384 cottons.⁶

Table 1.

Terms and nomenclature

Neps and Seed Coat Fragments:
1. Seed coat fragments (SCF)	Bits of seed coat tissue with or without attached fibers.
2. AFIS Seed coat neps (SCN)	Seed coat fragments which remain with attached fibers during opening in the AFIS.
3. AFIS Total neps	Includes fiber neps and all other (e.g. SCN).
4. AFIS SCN Selectivity	Distinction between true and false positive SCNs
AFIS Slivers and Machine-Fiber Interactions:
1. Recommended 0.5-g slivers	Raw (ginned lint) or processed lint, 0.5 g, draft to form a 30 cm long uniform cord or sliver.
2. Pre-opened 0.5-g and 0.1-g slivers	Thirty cm sliver (0.5 g) run twice through the AFIS to open the fibers and remove impurities, followed by hand cleaning and redrafting into a 0.5-g sliver (30 cm) or a 0.1-g sliver (6 cm).
3. Maximum machine (AFIS)-fiber interactions	Energy required to open, clean and to transport the raw fiber sliver through the various stages of the fiber individualizer in the AFIS.
4. Minimum machine (AFIS)-fiber interactions	Energy required to open and to transport the pre-opened sliver through the various stages of the fiber individualizer in the AFIS.
Fibers Attached to Seed Coat Fragment (SCF):
1. SCF without fibers	Seed coat fragments cut from mechanically delinted seed.
2. SCF with fuzz fibers	Seed coat fragments cut seed from ginned cotton. The fuzz fibers (about 3 mm long) come from aborted fibers and fibers removed in ginning.
3. SCF with fuzz and normal length fibers	Seed coat fragments manually pulled from ginned lint with attached fiber length distribution, from about 3 mm to normal length.
AFIS Recovery of Spiked Entity:	Terms apply to pre-opened/pre-cleaned slivers only:
1. AFIS result before spiking (BS)	The number of SCN sensed by the AFIS before spiking an X-g sliver and expressed in the AFIS output as counts/2X-g.
2. Number of added SCF (NS)	The number of added SCF per sliver.
3. AFIS result after spiking (AS)	The number of SCN sensed by the AFIS after spiking an X-g sliver and expressed in the AFIS output as counts/2X-g.
4. Percent recovery (%R)	The percent of spiked SCF or fiber neps and sensed by the AFIS as SCN.

AFIS: advanced fiber information systems.

Table 2.

Seed coat limiting conditions for spiking experiments

Property	Limiting conditions examined
Seed coat source	Fragile seed, unknown seed and unknown seed with mechanical delinting
Cut position on seed	Mid, micropylar and chalazal ends
Length of attached fibers	No attached fibers (mechanically delinted seed), fuzz fibers (few mm long) and distribution in ginned lint (few mm to normal length)
Carrier fiber medium	Pima, upland and random selection
Extent of interaction between sliver fiber medium and added SCF	Minimum interaction—sliver previously run through AFIS to open fiber medium and allow for manual removal of residual impurities before spiking

The advanced fiber information system (AFIS) nep module was introduced almost two decades ago and is used throughout the industry.⁷ This is a fully automated instrument, based on single fiber testing, and can count the number and size of neps and seed coat neps (SCN, Table 1). Seed coat neps are SCFs which remain with attached fibers during opening in the AFIS. A sample in the sliver form is fed into the fiber individualizer, which separates the matrix into individual entities⁸ composed of fibers, neps, trash, and dust. Trash and dust particles are suctioned off. The fibers and neps pass through an acceleration channel to improve individualization and pass through the same light sensor. Based on waveforms, light sensed versus time, an algorithm categorizes the neps as either fiber neps or SCN, which were both sized as well.

Entities is the term used in the AFIS patent⁸ to emphasize that even though the output data may be regrouped into simple generic classes such as neps, SCN, trash and dust, each is broken down into further categories so that the algorithms minimize misclassification. For example, neps may be broken down into mechanically generated nep, immature seed coat nep, mature seed coat nep, shiny or immature nep, mature nep, etc. but reported straightforward as SCN and neps (all other). In the context of this research, as in the patent, the term entities is a reminder that the individualizer elements (components) in the AFIS, sensors and algorithms interact with many different types of impurities in cotton rather than just a few categories.

According to Uster, typical values for SCN in American upland bale cotton are (ct/g fiber): 11–20, low concentration; 21–30, medium; and 31–40 high.⁹ It should be emphasized that the manufacturer provided no information to the industry on the selectivity (see Table 1 definition) for SCN over the other entities or recovery from ginned lint samples. The possible source lights⁸ for the electro optical sensors in the AFIS for nep classification is NIR at ≈880 nm and UV (blue light). A nep-like entity passing through the sensor volume generates a wave form which is not specific to SCN. Thus, classification algorithms are needed to determine if the entity was a nep and further classifies by type (possible polyester, fiber or seed coat nep). The waveform characteristic signals used in the algorithms include peak area and peak amplitude. Additionally, the algorithms make use of the speed of the entity in the sensor volume. A particle with a larger mass, such as a SCN, will not accelerate as rapidly in the flow stream of the tapered section of the flow nozzle and, therefore, will have a lower velocity while in the measurement volume.⁸

After reviewing the literature on the published AFIS SCN data, no studies have determined if the results are influenced by crop year, cultivar and area grown. Thus, the underlying assumption is that the sensor wave form and therefore results, is independent of the three variables.

The standard test method to determine SCF is ASTM D2496.¹⁰ This method uses a low magnification of pre-weighed fibrous web; SCF and other impurities are manually counted, removed and weighed. The standard is operator dependent, both with skills of detection and removal, and is extremely time consuming.¹¹ However, even when the web was displayed on a monitor, over time the operator may suffer eye fatigue to the point where the probability of not counting SCF in the images increased with the number of impurities to be counted.¹² Apparently, the eye fatigue inaccuracy experienced will vary with the operator and the between operator component of error was avoided by one operator counting each image twice.

Although still in use, D2496 was withdrawn as a standard in 1987. Also note that the two methods, D2496 and AFIS, are inherently different; the ASTM is manual and thus operator dependent, whereas AFIS categorizes a wave form. AFIS classifies SCN as such only, if in the size range of 0.5 to 3 mm and meet the certain criteria regarding morphology and light absorption and reflection. Larger and smaller SCF are directed into the trash and dust stream, whereas manually, those may be counted.¹¹

Several researchers have reported a difference between AFIS SCN and SCF results. One such study involved four different ginning and lint-cleaning combinations used to study the effect of mechanical processing procedures.¹³ A stereo-dissecting microscope was used to determine the number of neps, seed or mote coat fragments and non-seed impurities per the fiber sample weight. The numbers of SCF counted microscopically (105.3 to 187.2 per g) were significantly higher than the corresponding AFIS counts (18.4 to 23.6 SCN per g) in all four gin-type/lint-cleaning combinations. The authors suggested the low instrument counts were due to either the AFIS not detecting the same impurities counted microscopically or, more likely, the particles are categorized differently (e.g. trash instead of nep or SCF).¹³ In another study, the mean counts/g lint averaged across 10 cultivars were: SCF, 75 and SCN, 20 (private communication). Still other studies reported differences between SCF and SCN counts.^11,14,15 A more recent study demonstrated the addition of isolated SCF to hand-cleaned cotton affected the AFIS levels of both SCN and neps.¹⁶ The authors suggested that the opener in the AFIS breaks up the SCF into smaller particles. These smaller particles may become tangled with fibers and classified as neps.¹⁴

The effect of machine–fiber interaction on the cotton fiber quality and foreign-matter particle attachment to the fiber has been studied.¹⁷ The treatments varied from the hand harvested and hand ginned to the machine harvested, seed-cotton cleaned, to the machine ginned, and one-stage lint cleaning. As the number of mechanical interactions increased, the number of SCN measured by AFIS increased along with the particle attachment force.

Machine-fiber interaction in the AFIS may be great, since in one pass, the ginned lint is transformed into individualized entities. The aim of the present study was to investigate SCF spiking and recovery in the AFIS without the confounding effect of machine-fiber friction influencing the breakup of SCF. This may be achieved by pre-opening ginned lint samples to minimize machine interaction. Two passes through the AFIS were used to open the fibers. These pre-opened fibers can then be drafted into the sliver, the sliver spiked and analyzed by the AFIS. The mechanically treated sliver would physically transport the added entity into the AFIS. The expected recovery is no greater than the theoretical 100%. This fundamental research would be useful to help unravel results from AFIS-ginned lint interactions. For example, suppose that the ginned lint is spiked with SCF, analyzed by AFIS and recovery is 160% compared to 100% when the pre-opened lint is spiked. The additional 60% recovery must be due to self-generation in the AFIS from the particle crushing or misclassification. To understand recovery further, another study introduced the concept of AFIS SCN selectivity. Selectivity values were calculated from the published literature and data generated in the study.¹⁸ Diverse fiber mediums were identified that are suitable for selectivity studies involving dilution and spiking with various entities. A technique was developed to blend fiber mediums in the AFIS, and a case study presented of a medium where SCN count by AFIS >> SCF count by D2496.¹⁸

The specific objectives of this fundamental research were to: (1) develop methodology to spike pre-cleaned and pre-opened ginned lint samples with SCF and (2) demonstrate that the percent recovery is less than or equal to 100%. The reason for pre-cleaning was to avoid bias in the results due to false positive SCN counts from the trash impurities in ginned lint. The purpose of pre-opening was to minimize the energy required to open lint samples in the AFIS (machine interaction) and suppress that process from influencing recoveries.

Fundamentals

The specific parts/elements of the AFIS individualizer that interact with the entities in cotton are the perforated cylinder and solid cylinder in combination with carding flats.⁸ This process releases the various types of entities, such as neps, fibers and trash, one from the others, and individualizes the entities within each category so that the single separate objects are delivered one at a time at the output of the individualizer.

After the entities are individualized in the AFIS, neps and fibers enter an air stream; waste (trash, dust, large SCF) enter another air stream. Cross-contamination between streams may occur. Neps and fibers pass through the same light sensor. Based on the waveforms (light sensed versus time), an algorithm categorizes the neps into SCN and total neps (fiber neps and all other including SCN) (see Table 1). SCN and total neps are counted and sized. The Uster hardware and software that senses SCN and the subsequent algorithm that classified results are proprietary, and thus the waveforms produced by the AFIS sensor were unavailable for examination.

Spiked slivers and percent recovery

Recoveries (R) for spiked slivers, initially processed (pre-opened and pre-cleaned) through the AFIS and then hand cleaned, are calculated in the following manner

% R = 100 (\frac{AS - BS}{NS}) = 50 (\frac{AS}{NS})

(1)

where AS is the spiked sliver result by the AFIS, BS the sliver result before spiking, and NS – the number of SCF added per sliver (see Table 1). Since the slivers had been run beforehand in the AFIS and residual impurities manually removed, there are no SCF in the sliver before the actual spiking so that BS = 0.

Expected percent recovery

A brief description of the actual “spiking mechanism” of pre-opened/pre-cleaned sliver samples with SCF is presented here to familiarize the reader with this concept. To begin, longitudinal and transversal positioning of added SCFs in the fiber matrix is by means of moving elements (the hands). Handling the spiked sliver before it enters the AFIS fiber individualizer may cause migration of the seed coat bit to the sliver surface and result in it falling out of the sliver. This relocation process is highly influenced by the fiber flow properties and SCF geometry, size and surface characteristics. In contrast, holding together the SCF in the fiber matrix is influenced by entanglement. To increase the effectiveness of entanglement, the spiked sliver is hand rolled. Collectively, the spiking mechanism consists of the sequence of processes that convert two solids to the final product, “spiked sliver.”

After pre-opening the ginned lint twice in the AFIS, the randomly opened and individualized fibers fall into the fibers/neps waste box, are manually removed, hand-cleaned and formed into a sliver. The pre-opened/pre-cleaned sliver should not be identified as a traditional carded, drawn or combed sliver.

We should note that spiking the ginned lint sliver resulted in recoveries greater than 100% due to interaction effects, such as crushing or fragmentation in the AFIS.¹⁸ Misclassification of fiber neps as SCN also caused inflated outcomes. This is a more complicated problem to resolve and justifies the need for research using pre-opened and pre-cleaned cotton.

The spiking protocol described in this paper should produce a carrier medium without SCN suitable for spiking with low levels of SCF and in the instrumental recovery, sensed as SCN, ≤ 100%. The underlying assumptions are: (1) analyzing pre-opened lint in the AFIS—compared to ginned lint—prevents machine-fiber interaction in opening tangled fibers from influencing fragmentation of SCF and recovery of multiple SCNs and (2) no false positive counts from the double-cleaned and manually purified sliver. In the rank order, the expected % recovery of SCN versus the length of fiber attached to the spiked SCF is: no attached fibers < fuzz length fibers < length distribution in the ginned lint. The predicted range of values is: 0% to 100%.

To understand how the machine-interaction issue raised in the introduction has some bearing on the predicted recovery of SCN, consider three possible types of machine interactions in the AFIS. These are AFIS-fiber, AFIS-SCF and fiber-SCF. Using fiber media in the spiked runs that had been pre-opened will result in the reduced friction in opening, and therefore, during opening allow entities to slide by each other with less mechanical damage. Also, in pre-opened fibers the specific interaction, AFIS-SCF, is less dependent on the side reactions in opening tangled ginned lint that may result in inflated recoveries.

The AFIS neps sensor and its algorithm are electronic components whose function is to classify neps in the ginned and processed lint, but the recovery and selectivity documentation was not provided by the manufacturer. Thus, neither the manufacturer nor the authors of this paper can claim that all variables that control recovery of spiked SCF have been identified. The expected range of recoveries in the present study applies only to fibers pre-opened in the AFIS and hand-cleaned prior to spiking (Table 2).

Materials and methods

Production of materials

An overview of the cotton materials generated for this research is presented in Figure 1. Pre-opened and pre-cleaned slivers for AFIS analysis of SCN were spiked with either fiber neps (without SCF) or SCF. The sliver weights used are the recommended 0.5 g (30 cm long) and also 0.1 g (6 cm long). The primary reason for spiking one SCF at a time was that it simplifies interpretation of results. (For example, spiking two SCF at one time, one dubbed SCF_A and the other SCF_B, may yield the following possible SCN counts—actual value before the software multiplies x2—zero or one or two. This is explained as follows due to the SCF may undergo various processes in the AFIS opener: SCF_A and SCF_B yield no SCN for a total count of zero SCN; SCF_A gives one SCN and SCF_B none for a total count of one SCN; SCF_A is not sensed and SCF_B gives one SCN for a total count of one SCN; SCF_A and SCF_B each give one SCN for a total count of 2 SCN; SCF_A is not sensed and SCF_B gives two SCN due to fragmentation for a total count of 2 SCN; and finally SCF_A is sensed as two SCN and SCF_B is not sensed for a total count of 2 SCN. This can only be unraveled by spiking SCF_A or SCF_B one SCF at a time.)

Figure 1.

Flow chart for cotton materials production in this research. Cotton sliver with minimum machine-fiber interactions is coded as CS.

Other practical reasons for using the 0.1-g slivers rather than the 0.5-g slivers are to minimize the eyestrain associated with complete manual removal of residual impurities in the pre-opened fibers and a smaller number of SCF need to be prepared for spiking. The residual particles are smaller in the pre-cleaned cotton than in the raw material, are more difficult to visualize and extraction from the fiber causes eyestrain from intense use.

Table 3 details the material sources used in the investigation. There were four categories of cotton material: ginned lint (three sources), fiber neps (one source), special ginned lint (two sources) and cottonseed (three sources). Ginned lint provided the raw cotton to convert into AFIS slivers with minimum machine-fiber interactions. The sources of ginned lint represented domestic species of commercial importance: Gossypium barbadense (Pima) and G. hirsutum (Upland). Additionally, the remaining ginned lint sample was selected at random—from a pool of unspecified cottons—of unknown species and cultivar within the species.

Table 3.

Material sources in detail

Cotton material	Use	Source/species/cutivar
Ginned lint	AFIS pre-opened and pre-cleaned particle	1. Pima^a
(three sources)	carrier slivers.	2. Upland^a
		3. Unknown^a
Neps	Spike AFIS carrier slivers.	1. Cotton balls—scoured and bleached cotton; sold in retail stores by that name.
(one source)
Special ginned lint	Extract SCF from the lint with length	1. Phytogen cotton^a
(two sources)	distribution of attached fibers and spike into pre-treated AFIS carrier slivers.	2. Gossypium hirsutum L.—New cotton strain that contains a fragile seed coat that breaks easily.
Cotton seed	Cut SCF from seed with and without	1. Unknown seed^a
(three sources)	attached fuzz length fibers and spike pre-treated AFIS carrier fibers.	2. Fragile seed—from new strain of cotton mentioned above.
		3. Mechanically delinted lint—before and after delintering, one pass, grown in MS delta, 2014 crop, from an oil mill, cultivar not known.

Selected at random from pool of similar kinds of sample.

The stock of fiber neps (Figure 1) were pulled from commercial cotton balls that had been mechanically cleaned, scoured and bleached. The two special samples of ginned lint provided the supply of SCF pulled from the lint with the length distribution of attached fibers (a Phytogen cotton and Gossypium hirstum L, a new cotton strain that contains a fragile seed coat). Three cottonseed sources provided the resource of SCF cut from the seed, with and without attached fuzz fibers.

Steps 1 to 3 of the spiking protocol (Table 4) present the details to prepare the pre-treated AFIS slivers, 0.1 g and 0.5 g. Only SCF with a diameter of 0.5 to 3 mm were used in spiking. Note the protocol included quality control steps (QC1 and QC2) to ensure that the spiked particle was encased in the fiber matrix.

Table 4.

Spiking protocol

Step #	Description
0.1 g sliver
1	Pre-open ginned lint by passing 0.5 g sliver (30 cm long) twice through the AFIS.
2	After twice through the AFIS, pre-clean the open fibers by manual removal of all residual trash impurities including seed coat fragments and large neps.
3	Draft a 0.1 g sliver that is 6 cm long.
4	Manually open the sliver. Insert one SCF (diameter range: 0.5 to 3 mm) into the midsection (3 cm from one end) of the sliver so as to encase it in the fiber matrix.
5	Place the spiked sliver on a clean white surface and roll back and forth.
Quality Control (QC)-
QC1	Holding one end of the sliver vertically, wiggle it over a white surface and observe if the spiked particle has fallen out of the fiber matrix. If yes, repeat steps 4 and 5. If no, squeeze the sliver with the fingers to confirm that the added particle is enclosed in the matrix.
QC2	Reshape the sliver as need be and repeat step 5.
0.5 g sliver (protocol changes only)
3	Prepare a 0.5 g sliver that is 30 cm long.
4	Manually open the sliver. Using a ruler, insert five SCFs (diameter range: 0.5 to 3 mm) into the sliver, one each at a distance from one end (at 3, 9, 15, 21 and 27 cm) so as to encase all the bits in the fiber matrix.
5	Place the spiked sliver on a clean white surface and using both hands roll back and forth.
QC1	Holding one end of the sliver vertically, wiggle it over a white surface and observe if any spiked particle has fallen out of the fiber matrix. If yes, squeeze the sliver with the fingers to locate the position of the missing SCF; insert the SCF back into the sliver and roll back and forth. Squeeze the sliver with the fingers to confirm that all five SCFs are enclosed in the matrix.

AFIS analysis

An overview of all methods in this exploratory study is presented in Figure 2. Spiked slivers were analyzed for SCN on two AFIS units: AFIS 2, 0.5-g (Uster recommended sliver weight) and 0.1-g slivers; AFISPRO, only 0.5-g slivers. The acceptable weight range for the 0.5-g slivers is 0.49 g to 0.51 g. By entering in exactly 0.5 g and analyzing 0.5 g, the AFIS software automatically converts the SCN counts/0.5 g to counts/g by multiplying the counts × 2. By entering in exactly 0.5 g and analyzing 0.1 g, the internal multiplier is still × 2, equivalent to counts/0.2 g.

Figure 2.

Flow chart for experimental methods and statistical tests in this research.

To enable the system to test 0.1-g slivers (not the standard procedure) required the operator intervention each time that amount of sliver was run. This slowed down the process and prevented the AFIS from automatically analyzing multiple slivers. Thus, the primary AFIS in our laboratory for this research was the AFIS 2.

Mean and standard deviation of the spiked sliver results were calculated and expressed as the number of observed SCN and the calculated % recovery. One-tailed t-tests were used to test for the significant difference in % recovery between the 0.1-g and 0.5-g slivers, and between AFIS units. Traditional fiber properties were also analyzed on both AFIS units and examined for trends with the recovery.

Results and discussion

Fundamental research

In the context of this study, no studies have been reported that make use of cotton pre-opened in the AFIS, followed by manual cleaning to remove residual impurities, spiking with SCF, and then processing in the AFIS to measure SCN count. This paper focuses on curiosity driven research to understand the percent recovery of the spiked material in the ideal fiber matrix (minimum energy to open in the AFIS and no impurities) using three cottons and three sources of seed coats. Our curiosity progresses as we gather more data. Unrelated facts may be combined and produce new outcomes. These fundamental studies will tell us whether the recoveries are quantitative in the simplified fiber matrix, are related to composition and if a detailed study is feasible in the future.

Spiking protocol

Because this is fundamental research, protocols were established that were presumed to have provided the opened and cleaned sliver that could be spiked with entities that remained housed in the sliver until opened by the AFIS. Ideally, the hand-rolling steps in the SCF spiking protocol (Table 4) resulted in the entanglement between the fibers and added particles, and produced the fiber-SCF attachment force in proportion to the length of the fibers biologically attached to the SCF surface. The nature of this attachment force is ideally such that when the spiked sliver is passed through the AFIS individualizer, the added particles become individual entities and enter the air stream of the nep sensor.

Percent recovery of spiked entities

Tables 5 and 6 data

Table 5 documented the need to hand clean the AFIS pre-opened fibers prior to the spiking experiments. The grand mean SCN levels before hand cleaning (1.3 ct/g, 0.1-g sliver and 1.2 ct/g, 0.5-g sliver) were significant compared to the low spiked levels (NS of 1 and 5). After hand cleaning, no SCN were detected. The results suggested that with AFIS pre-opened fibers followed by hand cleaning to remove trash impurities and large neps, the reduced machine (AFIS)-fiber interactions do not result in generation of false SCN, which would positively bias the spiking results. Apparently the small neps, when subjected to minimum machine-fiber interactions, are not altered in the structure to the point where software classification as a SCN occurs.

Table 5.

Comparability of AFIS 2 SCN baseline levels (BS^a in equation (1)) for three cultivars: twice pre-opened in the AFIS, and before and after hand cleaning

	0.1-g slivers, mean cts/g (5 reps)
Cultivar	Before hand picking	After hand picking
Unknown (random selection)	1.6	0
Upland	0.4	0
Pima	1.6	0
Grand mean	1.3/1.2^b	0/0^b
NS^c	1/5^b

BS is the sliver result before spiking.

0.5-g slivers.

NS is the number of SCF added/sliver.

Table 6.

Percent recovery of spiked entity—at fixed sliver weight of 0.1 g—as a function of length of attached fibers, position on seed and carrier fibers (cultivar). AFIS 2 system

Carrier fibers: previously run in AFIS twice followed by hand cleaned before spiking; % R is percent recovery
Spiking: add one particle, inserted into 0.1 g sliver, 6 cm long
SCF cut from seed: pt, cut micropylar or pointed end of seed; rd, chalazal or rounded end of seed; frgl, fragile seed
SCF scources: cut seed—mechanically delinted seed, unknown seed and fragile seed; pulled from ginned lint—fragile seed cultivar
Fiber nep: CB nep, nep pulled from control cotton, scoured and bleached (cotton balls).
Attached fiber length (see Table 1): N, no fibers, mech. delinted; F, fuzz, cut the seed after ginning; and D, length distribution, pulled from ginned lint
		SCF Carrier (0.1 g sliver, spiked with one fiber nep or SCF; AFIS SCN counts/% R)^a
		Pima		Upland		Unknown
SCF source	Description	Means (cts /0.2 g/% R)	Std dev (cts)	Means (0.2 g/% R)	Std dev (cts)	Means (0.2 g/% R)	Std dev (cts)
1. fiber nep^b	CB nep	0.80/40	1.1	2.0/100	2.0	2.4/120	2.6
2. mech. delinted seed N	pt	0/0	0/0	0.4/20	0.9	0.4/20	0.9
3. N	rd	0/0	0/0	2.8/140	1.8	1.6/80	1.7
4.before delinting F	pt	0/0	0/0	2.0/100	2.0	0.8/40	1.1
5. F	rd	0/0	0/0	0/0	0	1.6/80	2.6
6. unknown seed F	pt	0.4/20	0.9	0/0	0.0	2.4/120	1.7
7. F	rd	0.4/20	0.9	2.0/100	3.5	4.0/200	1.4
8. fragile seed (FS) F	pt	0/0	0.0	2.0/100	2.8	0/0	0.0
9. F	rd	0.4/20	0.9	2.4/120	3.3	0.8/40	1.8
10. ginned lint from FS D	frgl	2.8/140	2.3	3.2/160	2.3	3.2/160	2.7

Means based on 15 reps. ^b% R as a SCN, not a fiber nep.

Table 6 recorded the recovery of spiked entities at a fixed sliver weight of 0.1 g. The effects of SCF source, the length of attached fibers, the position on the seed surface for cut fragments, and carrier fibers (cotton species and cultivars) are probed. For convenience and ease in understanding the spiking results, the rows and columns in Tables 6 and 7 were sorted in order of increasing percent recovery. Generally, the first row and Pima column holds the smallest recoveries within a SCF source, and the last row and the unknown cultivar include the highest recovery.

Table 7.

Percent recovery of spiked entity—comparing 0.1-g and 0.5-g slivers—as a function of length of attached fibers, position on seed and carrier fibers (cultivar). AFIS 2 system

Carrier fibers: previously run in AFIS twice followed by hand cleaning before spiking; % R is percent recovery
Spiking: 0.1-g sample: add 1 SCF, inserted into 6 cm long sliver; 0.5-g sample: add 5 SCFs, inserted into 30 cm long sliver
SCF source:
Unknown seed: cut from midsection of seed
Phytogen and Fragile seeds: gently pulled from ginned lint from Phytogen and fragile seed cultivars
Attached fiber length: F, fuzz, cut the seed after ginning; and D, length distribution, pulled from ginned lint
		SCF Carrier; Means (0.1-g sliver, cts/0.2-g; 0.5-g sliver, cts/g)
		Pima		Upland		Unknown
SCF source	Description	Means	Std dev	Means	Std dev	Means	Std dev
		cts/% R	cts	cts/% R	cts	cts/% R	cts
Unknown seed F	0.1-g sliver	1.2/60	2.7	1.6/80	2.6	0/0	0
F	0.5-g sliver	2.8/28	3.3	4.8/48	4.6	8.0/80	4.0
Phytogen seed D	0.1-g sliver	0.8/40	1.1	0.4/20	0.9	1.2/60	1.1
D	0.5-g sliver	4.4/44	1.7	6.0/60	3.2	12.8/128	3.3
fragile seed D	0.1-g sliver	2.8/140	2.3	3.2/160	2.3	3.2/160	2.7
D	0.5-g sliver	4.4/44	5.0	5.6/56	5.4	6.8/68	1.8

All rows of data in Table 6 were the result of spiking the single particle into the pre-treated sliver that weighed one tenth of a gram. In normal testing operations using 0.5-g slivers, the AFIS takes the actual measurements of both fiber and seed coat neps and multiplies the result by two to get the nep count per gram. In analyzing the 0.1-g sliver, the operator inputs 0.5-g sliver weight, which instructs the software to multiply the result by two to get the nep count/0.2-g. Thus, for 100% recovery (R), the AFIS result displayed will be two fiber neps or two seed coat neps.

The first row of results in Table 6 was concerned with spiking the pre-treated sliver with the single fiber nep from cotton balls (mechanically cleaned, scoured and bleached). For the Pima carrier sliver, the mean counts observed as a SCN—not a fiber nep—was 0.8 ct/0.2 g, equivalent to 40% R. Recovery increased to 100% with Upland and 120% with Unknown carrier slivers. Difference in R between the Upland and Unknown slivers was not significant. Direct AFIS analysis of cotton balls revealed the high count of both neps and SCN; some fiber neps were misclassified as SCN.¹⁸ In Figure 3, light micrographs of neps selected at random from cotton balls and ginned lint shows the nep in cotton balls is more tightly tangled.

Figure 3.

Light micrograph of: (a) nep selected at random from ginned lint, (b) nep selected at random from cotton balls, (c) SCF without attached fibers, cut from delinted seed, (d) SCF with attached fuzz fibers, cut from seed, and (e) SCF carefully removed from ginned lint showing length distribution of attached fibers.

Next, in Table 6, rows 2 through 5 introduced mechanically delinted seed as the source of SCN. SCF cut from the seed before and after delinting, from both the micropylar, or picropylar, and the weaker, chalazal end of the seed. No SCN were observed when the bits were spiked into Pima carrier slivers. The Upland and Unknown slivers produced significant recoveries compared to the Pima slivers. However, before and after delinting (attached fuzz fibers and no attached fibers, respectively) produced similar results. In three of the four pairs of data (the exception being Upland carrier: before delinting, micropylar and chalazal) recoveries of the spiked SCF were greater when the added fragment was from the chalazal end of the seed.

Continuing in Table 6, SCF cut from the Unknown seed and the fragile seed are shown in rows 6 to 9. The spiked Pima sliver produced the small recovery and recoveries sometimes 100% or greater in the other two carrier slivers. For all four pairs of data, recoveries of the spiked SCF were greater when the added fragment was from the chalazal end of the seed. In row 10 of Table 6, SCF were carefully removed from the ginned lint of fragile seed cotton and added into the AFIS carrier slivers. The measured SCN recoveries were consistency higher than 100% in all three types of carrier sliver.

Trends in the data

Although there is scatter in the data, the following trends have been assembled from the results under the AFIS conditions of minimum machine-fiber interactions. Fiber neps from cotton balls were again sensed as SCN, confirming an earlier study on the different AFIS exemplar.¹⁸ The question of data repeatability of spiked results between machines is another issue and was beyond the scope of this fundamental research. Repeatability between AFIS units, within and between the various models of the AFIS, may also be a function of the age difference between the systems.

Also, with no fibers attached to the SCF and with fuzz length attached fibers, recovery of the spiked entity in the Pima carrier was less than in the other two carriers. With SCF gently removed from the ginned lint and the SCF structure showing the length distribution of attached fibers, SCN recovery was greater in the Pima carrier. Furthermore, in eight matched pairs of spiked SCF coming from the micropylar and chalazal end of seeds, seven of the pairs produced a higher recovery of SCN when the fragment originated from the chalazal end of the seed.

Table 7 compares the % R of one SCF added to the 0.1-g sliver and five SCF added to the 0.5-g sliver, all measured on the AFIS 2 and collected under the conditions of minimum machine-fiber interactions. To determine if the results from either sliver weight are consistently biased positively or negatively, we used a signed test to rank paired results. There are nine paired mean % R values in the table (example, unknown seed source, Pima carrier: 0.1-g sliver, 60% R and 0.5-g sliver, 28% R). For five of the nine pairs recovery is greater in the 0.5-g slivers and less in the remaining four pairs. This is not a significant difference and there is no reason to believe that either sliver weight produced biased results.

The obvious difference between the sliver weights is that the 0.5-g slivers containing the higher level of spiked fragments produced more consistent results. For all three SCF sources, % R increased in the following order without exception: Pima < Upland < Unknown for the 0.5-g slivers. Within the SCF carrier sliver, SCF with attached fuzz fibers gave the lower recoveries compared to SCF with the length distribution of attached fibers.

Comparison of percent recovery by two AFIS units

Table 8 summarizes the paired SCN results for the AFIS 2 and AFIS PRO. The ginned lint results are given in ct/g. These were the actual levels of SCN in the baled cotton without spiking. The processed lint SCN values are given in terms of ct/g and % R.

Table 8.

Seed coat nep (SCN) measurements by two AFIS units.^a Paired results only.^b

	AFIS 2		p values and grand means		AFIS PRO
Cotton	Means	Std Dev	p value^c	AFIS Means (2 + PRO)/2	Means	Std Dev
Ginned lint (raw cotton, cts/g, no spiking)
Pima	4.5	3.1	0.25	1st 4.25	4.0	2.0
Upland	18.7	5.9	7.8 x 10^-5	2nd 14.75	10.8	7.3
Unknown	20.1	10.3	0.074	3rd 22.45	24.8	5.8
Fragile	27.7	9.9	0.43	4th 29.75	31.8	7.2
Processed lint spiked with 10 SCF/g fiber. Recovery: cts/g; % recovery^d,e
Recovery examples: AFIS2, 2.8 cts/g; 28% R. AFISPRO, 2.4 cts/g; 24% R.
[SCF source: unknown seed with attached fuzz fibers]
Pima	2.8; 28	3.3	0.83	1st 26	2.4; 24	2.2
Upland	4.8; 48	4.6	0.63	2nd 42	3.6; 36	2.6
Unknown	8; 80	4	0.023	3rd 54	2.8; 28	1.1
[SCF source: ginned lint from Phytogen seed with length distribution of attached fibers]
Pima	4.4; 44	1.7	0.62	1st 48	5.2; 52	3.0
Upland	6.0; 60	3.2	0.52	2nd 52	4.4; 44	4.3
Unknown	12.8; 128	3.3	0.025	3rd 102	7.6; 76	2.6
[SCF source: ginned lint from fragile seed with length distribution of attached fibers]
Pima	4.4; 44	5.0	1.	1st 44	4.4; 44		2.6
Upland	5.6; 56	5.4	1.	2nd 56	5.6; 56		3.0
Unknown	6.8; 68	1.8	0.43	3rd 78	8.8; 88		5.0

0.5 g sliver analyzed.

Replicates per cultivar are all five except 15 for AFIS 2 ginned lint with no spiking.

p value of mean paired difference between AFIS 2 and AFIS PRO.

Processed lint is ginned lint run twice through the AFIS, trash impurities and large neps removed manually prior to spiking.

Spiking protocol (see Table 4).

Note the p values of the 13 mean paired differences between AFIS 2 and AFIS PRO. Only three significantly differ at the 0.05 level. The first is ginned lint, Upland cotton. Second is processed lint spiked with SCF with attached fuzz, unknown cotton. The third is processed lint spiked from Pytogen seed with SCF with the length distribution of attached fibers, unknown cotton. Variance was reduced further by calculating the grand paired means across the two AFIS units (2 + PRO)/2 (Table 8).

Now supported by grand paired means, the ginned lint results without spiking (ct/g) and the processed lint results with spiking (% R) show the same increasing-rank order, without exception: Pima cotton, Upland cotton and Unknown cotton. This is a significant finding because it suggests that perhaps both the cotton sliver source and the SCF source play major roles in the recovery of SCF from the ginned lint and processed lint. Furthermore, the ginned lint results were collected under the conditions of maximum machine AFIS-fiber interactions and the spiked results were collected under the conditions of minimum machine AFIS-fiber interactions. The interaction effects may not be as great as that for material sources.

To help visualize the rank-order relationships, Table 8 data was used to create Figure 4. The x-axis is the SCN concentration found in the unspiked ginned lint (ct/g): Pima, 4.25; Upland, 14.8; and Unknown, 22.5. The y-axis is the SCN recovery in the spiked lint (% R). The horizontal dashed line represents the theoretical recovery of 100% R; all nine data points were expected to be on the 100% R line. Note the overall range of recoveries, from 26 to 102%. For SCF1 source with attached fuzz fibers, the range of recoveries was 26 to 54%. For SCF2 and SCF3 with the length distribution of attached fibers, the range of recoveries was 44 to 102%. Out of six trials with the added SCF that had been carefully pulled from the ginned lint, only one gave the expected 100% recovery (actually 102% but the difference relative to 100% is not significant).

Figure 4.

Relationships between SCN in ginned lint, ct/g and SCN recovery in spiked lint, % R. SCF# code: SCF1, cut seed from unknown seed with attached fuzz fibers; SCF2, ginned lint from fragile seed with length distribution of attached fibers; and SCF3, ginned lint from Phytogen seed with length distribution of attached fibers.

The AFIS was designed to measure SCN in raw and processed cottons. It is reasonable to assume that SCF cut from the seed with only attached fuzz fibers should have a smaller attachment force to the sliver fibers than SCF with the length distribution of attached fibers. At this early stage of research it is unknown why the reduced recovery of SCF is observed with just attached fuzz fibers. Future work may elucidate, if some SCF with only attached fuzz fibers get directed into the wrong air stream (i.e. the trash air stream rather than the neps and fibers air stream).

Also of interest is the lower recovery of the spiked bits of seed coat tissue in the Pima slivers compared to the other two cottons. Of the three carrier fibers, the Pima was longer, finer and contained more dust and trash (Table 9). It is clear that not only is the recovery of SCF in the pretreated Pima sliver low, but the actual count was very low in the ginned lint. Because this is fundamental research, there is still more work to be done. Determining the fate of the SCN not sensed by the AFIS and recovery of SCF added to the ginned lint without pretreatment of the fibers will be tackled in future studies.

Table 9.

Selected fiber properties by AFIS PRO and AFIS 2^a.

	SCN		Neps		Length	Fineness	Maturity	Dust	Trash
Ginned Lint (AFIS PRO)
Cotton	(ct/g)	(µm)	(ct/g)	(µm)	(in)	(mTex)	(mat. ratio)	(ct/g)	(ct/g)
PIMA	4	1013	230.6	694.8	1.226	166	1.056	1297	111.6
Upland	10.8	1364	218.4	696.4	0.974	199.6	1.018	340.2	52.4
Unknown	24.8	1323	209.2	753	0.906	174.8	0.97	2076	462
Fragile seed	31.8	1232	278.2	723.6	0.984	176.2	1.006	390.4	84
Ratio: After Two Passes/Ginned Lint (AFIS 2)^b
PIMA			1.18		0.979
Upland			1.22		0.948
Unknown			1.66		0.972

Mean of five reps.

No hand cleaning after pre-opening in the AFIS 2

Conclusions

This fundamental research, with limited number of cottons and seed coat sources, focused on understanding percent recovery of spiked seed coat fragments in pre-opened and pre-cleaned cotton fibers. Under these conditions, cotton slivers spiked with seed coat fragments (SCF) with attached fibers resulted in a range of recoveries, from 26 to 102% when averaged across two AFIS units. Added SCF without attached fibers produced the recovery from 0 to 80% when averaged across the paired data from one AFIS unit. The spiking protocol included the hand rolling process to produce the sliver-SCF attachment force.

The recovery of the spiked entity was dependent on the species and cultivar of the cotton used to prepare the slivers. The lowest recovery was observed with spiked Pima slivers.

Percentage recovery was greater when the SCF came from the chalazal end of the seed. Also, the percentage recovery increased with the increase in length of the fibers biologically attached to the SCF surface. Seed coat fragments with no attached fibers—cut from the mechanically delinted seed—produced the lowest recovery. Seed coat fragments removed from the ginned lint with the length distribution of attached fibers gave the theoretical recovery (100%) for only one of six trials.

Separation of “large” SCFs in the AFIS possibly could further complicate the validity of the results of SCNs and neps. The variability of spiked SCF recovery sensed as SCN seems to be influenced by: the loss of large SCFs, length of attached fibers, fiber species and cultivar, SCF source and position on seed.

This fundamental work suggests that a detailed study is feasible in the future. Under the AFIS conditions of pre-opened/pre-cleaned fibers, future work will determine the fate of the spiked SCFs not sensed as SCN. SCFs will be collected from both fiber and trash boxes. Other experiments will continue to probe seed coat nep levels and mid-IR spectra at limiting conditions. Also, future studies will determine if the length and fineness value of spiked cottons affects the recovery of SCF. A much larger sample size (picked randomly from the population) and many replications of cotton in its ginned version (i.e. not yet pre-processed) are needed to determine if the exploratory results translate into reliable and reproducible bale results.

Footnotes

Acknowledgements

The author acknowledges the cooperation of Zellweger Uster for specifying the size range of seed coat neps sensed by the AFIS, and confirming that with 0.5-g slivers the AFIS software multiplies the actual measurements of both fiber and seed coat neps by two to get the count per gram.

Disclaimer

Use of company or product name is for information only and does not imply approval or recommendation by the United States Department of Agriculture to the exclusion of others.

Declaration of conflicting interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

References

Csiszar

Losonczi

Szakacs

. Enzymes and chelating agent in cotton pretreatment. J Biotechnol 2001; 89: 271–279.

Taveer

. The influence of different pretreatments on the quantity of seed-coat fragments in cotton fibres. Fibres Text East Eur 2008; 16: 19–23.

Yan

Hua

Qian

. Analysis of the chemical composition of cottonseed coat by Fourier-transform infrared (FT-IR) microspectroscopy. Cellulose 2009; 16: 1099–1107.

Mangialardi GJ. A review of seed-coat fragment research. In: Proceedings of the beltwide cotton conferences (eds CP Dugger and DA Richter), New Orleans, LA, 5–8 January 1988, pp.592–594. Cardova, TN: National Cotton Council of America.

Watson H and Helmer JD. Cottonseed quality as affected by the ginning process—a progress report. Publication no. 42–107. 1964. Washington, DC: USDA-ARS.

Pearson NL. Seed coat fragments in cotton—an element of yarn quality. Tech. Bull. no. 1116. 1955. Washington, DC: USDA.

Zellweger

. AFIS detection of seed-coat neps. Text World 1996; 146: 87–88.

Shofner FM, Baldwin JC, Galyon ME et al. Apparatus and methods for measurement and classification of generalized neplike entities in fiber samples. Patent 5430301. USA, 1995.

Uster AFIS PRO. Presented at the Seedbreeders Conference, Raleigh, NC, 11–16 September 2005. Cary, NC: Cotton Incorporated.

10.

ASTM D2496:1985. Standard test methods for seed coat fragments and funiculi in cotton fiber samples. In: Annual Book of ASTM Standards. American Society for Testing and Materials, West Conshohocken, PA, 07.02:570-575. Withdrawn 1987.

11.

Boykin

. Tracking seed coat fragments in cotton ginning. Trans. ASABE 2008; 51: 365–377.

12.

Giner M, Gourlot J-P, Bachelier B, et al. Latest developments and results in automatic SCF counting. Part I: Development of algorithms and preliminary results. In: Proceedings of the beltwide cotton conferences (eds CP Dugger and DA Richter), New Orleans, LA, 6–10 Jan 1997, pp.1633–1636. Cardova, TN: National Cotton Council of America.

13.

Jacobsen

Grossman

Hsieh

Y-L

. Neps, seed-coat fragments, and non-seed impurities in processed cotton. J Cotton Sci 2001; 5: 53–67.

14.

Boykin C. Origin of seed coat fragments in ginned lint. In: Proceedings of the beltwide cotton conferences, New Orleans, LA, 4–7 January 2005, pp.745–755. Cardova, TN: National Cotton Council of America.

15.

Boykin C. Genetic traits associated with seed coat fragments, motes and neps. In: Proceedings of the beltwide cotton conferences, San Antonio, TX, 3–6 January 2006, pp.436–447. Cardova, TN: National Cotton Council of America.

16.

Bel P and Xu B. A close look at cottonseed coat fragments with AFISPRO. In: Proceedings of the beltwide cotton conferences (eds S. Boyd, M. Huffman and B. Robertson), Atlanta, GA, 4–7 January 2011, pp.1226–1235. Cardova, TN: National Cotton Council of America.

17.

Sui

Thomasson

Byler

. Effect of machine-fiber interaction on cotton fiber quality and foreign-matter particle attachment to fiber. J. Cotton Sci. 2010; 14: 145–153.

18.

Montalvo JG, Von Hoven T, Hinchliffe D, et al. AFIS seed coat neps baseline response from diverse mediums. In: Proceedings of the beltwide cotton conferences (eds S. Boyd, M. Huffman and B. Robertson), New Orleans, LA, 6–8 Jan 2014, pp.945–952. Cardova, TN: National Cotton Council of America.