Dupont
ABSTRACT
Color represents an important element in the visual aspect of a fabric. However, during the different stages in producing a fabric (spinning, weaving, knitting . ), color evolves due to different surface textures. This paper presents a preliminary approach to modeling color evolution during the spinning stage when the roving is transformed into yarn. After spinning, the yarn is not dyed, and the color depends uniquely on the initial color of the roving.
This project was undertaken in the framework of the colorimetry of precolored fiber blends [4]. Blending precolored fibers involves obtaining a color by mixing different colored fibers. This technology is used today to produce different articles, such as spinning mottled cotton, technical textiles and/or needle felt and needle loom carpeting, ribbon dyeing, and textile recycling (fibers, wovens, or knits after shredding, . . .). A number of theories have been elaborated to describe the colorimetric behavior of fiber blends [2, 3]: Kubelka-Munk's theory [1], Friele's model [5], and Stearns and Noechel's model [9].
Since the fibers are already colored, there is no dyeing or printing stage in the production cycle for products made from these fibers. The desired color is achieved by intimate blending of different colored fibers. However, when the roving is transformed into yam, nonwovens, felt, or carpeting, and when the yam is transformed into fabric (woven or knitted), a change in color appears [6, 8]. It is therefore essential to model the influence of these transformations on the color aspect in order to predict the color of the finished article in relation to the initial fibers. This problem is the same for all companies that use colored fibers at the beginning of the production process with no dyeing stage.
This paper presents the results of our study limited to the influence on color aspect of the transformation of roving into yam. Certain characteristics of the material and/or the spinning process influence this color change, notably the parallelization degree of the fibers, fiber fineness, yam count and its twist, and paraffining.
Parallelization degree of the fibers: Each space between the fibers is a light trap, which casts a shadow on its environment and so darkens the color [10]. The brighter the sample, the more important the darkening. Thus open-ended spinning, which produces fibers with a weaker degree of parallelization than conventional spinning, implies a greater degree of darkening.
Fiber fineness, yam count and its twist: The average number of fibers in a yarn cross section depends on their fineness and the yam count. So since each space between the fibers represents a light trap, the more light traps there are, the more the color of the yam will be darkened in relation to the color of the roving. Fineness and count thus represent a parameter that influences this phenomenon.
Yarn twist also modifies fiber parallelization and so the surface aspect of the yam. In addition, the twist is linked to the yam count by the Koechlin relation:
Experimental Conditions
In order to limit the influence of these characteristics, we first made a study of plain cotton fibers with the same fineness (1.6 dtex) and an array of nineteen pre-colored roving coverings, using the CIELAB L*a*b* color space. The different yams produced (using open-end spinning) included paraffined and nonparaffined yarns with counts of 20, 25, and 33 tex (single yarn). Thus, for each roving there were six series of yarn produced. Each yarn was flat-wound on a cardboard support.
The colorimetric measurements involved a Chromasensor CS3 spectrophotometer by Datacolor International with a geometry of D/8, the specular component excluded, and using the large area view (18 mm diameter). Each sample was measured ten times (D65 illuminant, 100 observer) in order to allow a statistical analysis and increase the reliability of the measurements. Measurement by tolerance was chosen as the method, with a tolerance of 0.03 CMC (2:1) color difference [11] and a minimum of four measurements per value.
Interpreting the Results
Table I gives the results of the colorimetric measurements for the 33 tex nonparaffined yam sample. All of the measurements with the different yam counts are represented in the CIELAB L*a*b* color space. In order to clarify the representation, the points are noted in the three bidimensional spaces: a*b* (Figure 1), L*a* (Figure 2), and L*b* (Figure 3).
An analysis of these representations shows that during the transformation from roving to yam, all of the samples tend to move towards the origin. In fact, in relation to the L * a * b * color space, the color tends to be attracted to the weak values of L and to drab colors during the transformation from roving into yam. The coordinates of the attraction point are L* = 0, a* = 0, b* = 0.
This interpretation is made even clearer by the representation in L*C* (Figure 4) and the representation of the graph of the yam hue in relation to the roving hue (the latter curve is represented solely as a reference) (Figure 5). In fact, we noticed that the angle of the hue varies little (Figure 5), so we have h yam equal to h roving. In contrast, the move toward the origin is found again in L*C* (Figure 4). We can therefore state that when roving is transformed into yam, the color tends toward darker and duller without modifying the hue. Certain samples do not quite follow this rule, notably the bleached ones, which show a fluorescence phenomenon (presence of fluorescent brightening agent).
Numerical Solution of Model: L*C*
Since the angle of the hue varies little during the transformation from roving into yarn, we think that the luminosity of a given yarn Lf and its chroma Cf depend only on the luminosity Lb and the chroma Cb of the roving used to produce the yarn: where a11 is the contribution of roving luminosity to yam luminosity, a22 is the contribution of roving chroma to yarn chroma, a21, is the contribution of roving chroma to yam luminosity, and a12 is the contribution of roving luminosity to yam chroma.
Resolving this over-determined system (nineteen equations with four unknowns) for the six series of yam, we obtain the following pseudo-results [71 (Table II). Because of the significance of each coefficient, the coefficients in the upper diagonal are more important than those in the lower diagonal. The angle of the yam hue is identical to the angle of the roving hue (see, for example, Figure 5).
COMPUTATIONAL RESULTS OF THE L*C* MODEL
Applying this model to the nineteen samples gives the following results (Table III), expressed in cMC (2:1). Note that for certain samples, the cmc (2:1) difference is less than 1, which is perfectly acceptable. However, for some other samples, the CMC (2:1) differences are over 2 (or even 3), which is far too high.
These results show that this model is not applicable as it stands, and leads us to investigate a model with reflectance curves. Moreover, knowing only the L*C*h coordinates of an object, it is very difficult to trace back to the reflectance curve, although this is often necessary.
Numerical Solution of Model: DeltaR%
Since a given color (represented in the L*a*b* color space) can have several corresponding spectra, we decided to look at the reflectance curves of the different samples. The model AR% is the result of representing the variations in reflectance values between the roving and the yarn, independent of the wavelength. It quantifies the difference of roving/yarn (AR%) between the reflectance values of the roving and of the yarn produced. In our representation, this difference is shown in relation to the reflectance values R% of the roving (Figure 6).
In this graph, we see that the dispersion of points is greater when the reflectance value R% is high. However, given that the reflectance variations in the higher values are less sensitive (for calculating L*, a*, and b*) than the same variations in the lower reflectance values, it is possible to link these points using a polynomial approximation of 2 (even if a strong divergence exists in the higher reflectance values). This polynomial approximation is presented in Figure 7 with the uncertainties in reflectance (equal to three times the standard difference). The dispersion of the points observed does not call into question the DeltaR% model; in fact, a given reflectance difference in the lower reflectance values has much more influence on the color aspect than the same difference on the higher reflectance values.
COMPUTATIONAL RESULTS
Table IV presents the coefficients of the six transformation polynomials as well as the R2 correlation coefficients.
Table V presents the cmc (2:1) obtained between the color of the yams measured and that of the yams modeled. Note that certain samples, such as the black, show important differences. In other respects, the model produces some very interesting results, and significantly reduces the differences between the roving and the yam. This allows spinners to predict as precisely as possible the final color of a yam from the color of the roving (before spinning).
General Discussion
The model DeltaR% leads to some very interesting resuits. However, certain samples (such as the black) show an unacceptable margin of error (except for the paraffined yarn with a count of 20 tex). This margin of error can be explained by the fact that, as we already stated, a reflectance variation on the lower reflectance values (which is the case for black over the whole spectrum) has a huge influence on the calculation of L*, a*, and b*. This explains why a small error in reflectance leads to an important cMc (2:1) difference for the blacks. We tested DeltaR% in relation to the wavelength, but it did not lead to any increased accuracy, which shows that the transformation does not depend on wavelength.
Analyzing the different yarns appears to reveal two trends that have to be taken with caution and further explored, since this study only looked at three counts of yarn, paraffined or nonparaffined. First, the finer the yarn, the smaller the difference between the roving and the yarn. We can try to explain this phenomenon by the notion of light traps: the finer the yarn, the smaller the interstices between the fibers, leading to a weaker loss of luminosity between the roving and the yarn. Second, paraffining increases the DeltaR% difference between the roving and the yarn. This influence seems to be the same for the different counts.
Conclusions and Perspectives
The representation in the CIELAB color space has shown that during the roving/yarn transformation, the color seems to be attracted to L* = 0, a* = 0, b* = 0 (black). However, the L*C* model does not give sufficiently satisfactory results to validate this method where all the samples are taken together. On the other hand, the so-called DeltaR% model based on reflectance curves gives very good results and is promising for the future. At present, we are continuing to analyze all the samples and to examine the influence of yarn properties and fiber properties. Furthermore, we are preparing a study to examine yarn/fabric and roving/nonwoven transformations.
ACKNOWLEDGMENTS
We would like to thank the ADEME (Agence De I'Environnement et de la Maitrise de I'Energie) and ITF-Lille (Institut Textile de France) for funding this research.
Literature Cited
1. Amirshashi, S. H., and Pailthorpe, M. T., Applying the Kubelka-Munk Equation to Explain the Color of Blends Prepared from Precolored Fibers, Textile Res. J. 64, 357364 (1993).
2. Burlone, D. A., Formulation of Blends of Precolored Nylon Fibers, Color Res. Appl. 8(2), 114-120 (1983).
3. Burlone, D. A., Theoretical and Practical Aspects of Selected Fiber-Blend Color Formulation Functions, Color Res. Appl. 9(4), 213-219 (1984).
4. Dupont, D., Steen, D., and Caze, C., Colorimetrie de Melange de Fibres Textiles Pr6color6es, in "Proc. Conf. ADER'98," 1998.
5. Friele, L. F. C., The Application of Color Measurement in Relation to Fiber Blending, J. Textile Inst. 43, 604-611 (1952).
6. Mersereau, E. P., and Rainard, L. W., Color Changes in Wool During Processing, Textile Res. J. 21(4), 239-247 (1951).
Nougier, J. P., "Methodes de Calcul Numerique," Masson, France, 1993.
Sirikasemlert, A., and Tao, X., Effects of Fabric Parameters on Specular Reflection of Single-Jersey Knitted Fabrics, Textile Res. J. 69, 663-675 (1999).
Steams, E. I., and Noechel, F., Spectrophotometric Prediction of Color Wool Blends, Am. Dyest. Rep. 33(9), 177180 (1944).
Steen, D., and Dupont, D., The Influence of Relief in the Measurement of Colours and Whites, paper presented to Conf. Ecole de Printemps: Couleur et Apparence Visuelle, March 2000.
Steen, D., and Dupont, D., The Control of Textile Colours-A Practical Method and Comparison of CMC(2: 1) and CIE(KI,Kc,Kh), paper presented to Congres EUROCOAT'99, Lyon, 1999.
Manuscript received May 30, 2000; accepted September 19, 2000.
D. DUPONT, D. STEEN, AND C. CAZE
Laboratoire de Genie des Materiaux et des Textiles, Departement Colorimitrie, ESTIT, BP 209, 59654 Villeneuve d'Ascq, France
