360
This document was uploaded by user and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this DMCA report form. Report DMCA
Overview
Download & View 360 as PDF for free.
More details
- Words: 3,987
- Pages: 7
Textile Research Journal
Article
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties Abstract
The main aim of the present study was to predict the most important yarn quality characteristics derived from cotton fiber properties that were measured by means of an HVI system. With this aim 15 different cotton blends were selected from different spinning mills in Turkey. The cotton fibers were processed in the short staple spinning line at Ege University Textile and Apparel Research–Application Center and were spun into ring yarns (20s, 25s, 30s and 35s). Each count was spun at three different coefficients of twist (αe 3.8, αe 4.2 and αe 4.6). Linear multiple regression methods were used for the estimation of yarn quality characteristics. Yarn count, twist and roving properties all had considerable effects on the yarn properties and therefore these parameters were also selected as predictors. After the goodness of fit statistics very large R2 (coefficient of multiple determination) and adjusted R2 values were observed. Furthermore, analysis of variance tables showed that our equations were significant at the α = 0.01 significance level.
Mustafa E. Üreyen1 and Hüseyin Kadoglu2 Textile Engineering Department, Ege University, 35100 Izmir, Turkey
Key words cotton yarn, HVI, cotton fiber, ring spinning, regression analysis, yarn quality estimation
The physical characteristics of a fiber determine its processing behavior, production efficiency and finally yarn and fabric quality. Therefore, predicting the quality characteristics of yarns, such as the tensile properties from the raw material properties, was the main purpose of many studies in the last century. In addition to raw material processing conditions, preparation stages, machine parameters and the spinning method also have considerable effects on the yarn properties. Generally two approaches were used in these studies for predicting yarn quality from fiber and yarn characteristics: • an empirical and statistical approach; and • a theoretical or analytical approach.1
Textile Research Journal Vol 76(5): 360–366 DOI: 10.1177/0040517506062262 Figures 1–4 appear in color online: http://trj.sagepub.com
The empirical and statistical approach to establishing a relationship between fiber and yarn quality characteristics has been the most popular method during the second half of the twentieth century. Fast and accurate measurement of fiber properties by means of high volume instruments (HVI) and more powerful computers are the two main reasons for this tendency. With this method, important fiber and yarn properties can be measured for a range of samples and by using these results empirical relationships have been established by means of statistical analysis. One of the
1 2
E-mail: [email protected] Corresponding author: e-mail: [email protected]
www.trj.sagepub.com © 2006 SAGE Publications
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu most common statistical approaches is the multiple regression method (e.g. linear, log–log, etc.). Such an approach is used to investigate the interdependence of the different fiber properties and to estimate the relative contribution of each fiber property to the overall yarn properties. Several researchers [1–5] have established various regression equations using this method. The theoretical approach is based on physical and mechanical principles. These models usually yield good information about interactions between different fiber properties and yarn characteristics. However, practical applications are almost impossible because of the complexities of the models. They are usually based on certain assumptions and their success is determined by the feasibility of these assumptions [4, 6]. In recent years some researchers [2, 4, 7–9] have shown an interest in the use of artificial neural networks (ANN) to predict yarn characteristics. This analytical system is also useful for discovering relationships between variables [9]. Chattopadhyay and Guha [10] have reviewed textile applications of artificial neural networks in detail. The tensile properties of a spun yarn have always been very important in determining the quality of the yarn, since they directly affect the winding and knitting efficiency as well as warp and weft breakages during weaving. It is, therefore, important to establish which fiber and yarn parameters influence the yarn tensile properties and if possible, to derive functional relationship between them. So far, numerous mathematical and empirical models have been established for the estimation of single yarn tenacity [3, 5, 11–13] and count strength product (CSP) [1, 3, 14, 15] using fiber properties and some yarn parameters. Hearle [16] reviewed various mathematical and empirical studies concerning yarn strength, which were published between 1926 and 1965. Hunter [6] reported on more than 200 published papers about the prediction of yarn quality parameters, particularly tensile properties, up to 2004. Another important yarn parameter is breaking elongation. It influences the performance of spun yarns during winding, warping, and weaving. Yarn elongation is chiefly influenced by fiber properties, yarn twist, and yarn count
361
[17]. However, prediction models dealing with the breaking elongation of cotton yarns are few in number. Mathematical models have been proposed by Aggarwal [18, 19], Frydrych [11], and Zurek et al. [20]. Statistical models have been developed by Hunter [3] and ANN models produced by Majumdar [4]. Unevenness is a very important factor for the yarn and fabric quality. Cross-sectional fiber variation is the basic reason for unevenness. In addition to machine parameters, spinning method, yarn count and some fiber parameters have decisive influence on the unevenness of the yarn. Hunter [3] and Ethridge et al. [2] have developed some models to determine yarn irregularity by using fiber parameters. Hairiness, another measurable yarn characteristic, is usually an undesirable property. Acceptable measuring devices for the determination of hairiness, such as the Uster Tester and Zweigle Hairiness Tester are relatively recent and therefore fewer research articles have been published on the estimation of hairiness by using fiber parameters. In this study, we developed statistical models to estimate ring yarn tensile properties, unevenness and hairiness values by using multiple linear regressions. In addition to HVI fiber properties, we also used roving properties, yarn count, and yarn twist.
Experimental In this work, a total of fifteen different cotton samples were collected in roving form from various spinning mills. The spinning operations can affect the fiber properties in different ways, depending on the machinery line and adjustments, etc. For the elimination of these effects, fiber properties were measured from finisher drawing slivers by using the Uster HVI testing system. Table 1 shows the fiber properties measured by HVI. All samples were spun into yams on ring spinning machine (Rieter Model G30) at a yarn count of Ne 20 (29.53 tex), Ne 25 (23.63 tex), Ne 30 (19.69 tex), and Ne 35 (16.88 tex). Each yarn count was spun at three different twist multipli-
Table 1 Main fiber properties of the tested samples. Property Fineness (microner) Strength (g/tex) Length (UHML) Uniformity (%) SFI (%) Elongation (%) Reflectance (Rd) Yellowness (+b)
Sample no. 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
4.7 31.6 30.30 86.6 3.5 6.1 82.2 9.4
4.0 50.6 32.60 86.8 3.5 7.4 77.7 9.2
4.3 33.2 28.55 86.5 3.5 6.8 79.9 8.9
4.8 32.0 29.35 85.9 3.5 6.2 81.9 8.9
4.1 34.2 27.70 83.5 6.7 5.8 78.2 9.8
4.7 33.7 29.95 86.9 3.5 6.4 82.4 9.3
4.2 47.8 35.10 90.8 3.5 7.7 80.6 9.4
3.9 30.3 29.90 84.8 4.1 6.8 68.1 11.0
4.5 33.8 29.25 85.5 3.6 6.7 80.9 9.4
4.9 34.3 29.50 86.2 3.5 6.6 81.8 9.0
4.9 31.7 27.60 83.4 6.8 5.6 80.2 9.1
4.3 33.7 29.45 86.2 3.5 6.7 80.9 9.5
4.9 33.6 29.90 87.4 3.5 6.4 82.9 9.1
4.5 38.8 30.55 87.7 3.5 7.6 81.4 9.4
4.5 33.1 27.45 82.4 8.1 5.7 77.8 8.9
TRJ
TRJ
362
Textile Research Journal 76(5)
Table 2 Main properties of rovings. Property Roving cnt. (Ne) Um (%) CVm (%) CVm (1m) (%)
Sample no. 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1.09 4.11 5.06 2.96
1.08 4.95 6.29 2.82
0.66 3.23 4.06 2.17
0.93 4.00 5.07 2.48
1.22 4.68 5.92 3.01
1.22 3.90 4.91 1.96
1.11 5.24 6.69 4.06
0.89 5.35 6.77 3.04
1.05 4.90 6.16 2.31
1.00 4.42 5.56 1.61
1.01 3.91 4.93 2.68
1.02 5.72 7.24 2.82
1.00 5.06 6.42 4.05
0.90 3.61 4.56 1.46
1.00 4.98 6.29 2.02
ers (αe 3.8, αe 4.2, αe 4.6). A total of 180 spinning trials were done. The appropriate drafting ratios were adjusted on the ring spinning machine for each sample. Other spinning conditions were kept constant. Orbit rings (42 mm diameter) and travellers (suitable weights were selected for each yarn count) were used. For each yarn sample ten cops were produced and tested. The tensile properties of the yarns were evaluated on an Uster Tensorapid tensile testing machine. Unevenness and hairiness tests were performed on an Uster Tester 3. Rovings were tested on the Uster Tester 3. The measurements of the main properties are shown in Table 2. The multiple regression method has the advantage of simplicity in describing the quantitative relationship between textile material properties. Therefore this method was selected for establishing the relationships between fiber and yarn properties. At the beginning the types of relationship between selected properties (independent variables) and yarn properties (dependent variable) were checked individually. The statistical tests indicated that there was a nearly linear relationship between fiber properties and yarn properties. Therefore a linear multiple regression analysis method was chosen for our study in order to establish a quantitative relationship of yarn properties with respect to fiber properties, roving properties, yarn count, and yarn twist. Forward stepwise (for estimation of tenacity, unevenness and hairiness) and backward elimination (for estimation of elongation) procedures were selected for the linear regression analysis. Statistical analyses were performed using the SPSS 11.0.1 and Minitab 11.12 programs.
Results and Discussion Influence of Fiber Properties on Yarn Tenacity As might be expected, a high positive correlation was found between fiber and yarn strength (0.905). After regression analysis it was seen that fiber strength was the most important parameter for yarn tenacity. Fiber elongation, fiber length (UHML), uniformity index, fiber fineness, yarn count, yarn twist, roving count and unevenness of roving are other parameters that can have a significant influence on yarn tenacity. Table 3 shows the regression coefficients of variables, the t-values and significance level of each variable. The estimated coefficients of variables indicate the direction of influence. As the strength of a yarn depends on the weakest place, an increase in irregularity should lead to a lower strength [16]. Therefore any unevenness of the rovings has a negative effect on the yarn tenacity. Roving count determines the draft on the ring spinning machine. Our model shows that the drafting ratio on the spinning machine has a positive influence on tenacity, but it should be noted that in our study the drafting ratio ranged from 18.80 to 54.20. We found a negative effect between fiber elongation and yarn tenacity. A stepwise method was used for the regression analysis. The significance values obtained were based on fitting a single model and in this situation the significance values can sometimes be invalid when a stepwise method is used. Furthermore, autocorrelation between fiber properties is another basic problem for linear regression analysis. How-
Table 3 Regression coefficients, t-values and significance level of t-values of our linear regression model for yarn tenacity. Independent variables (Constant) Strength b* –19.883 t –3.632 Sig. 0.000
0.701 29.900 0.000
*Partial regression coefficient.
Twist co. (αe) 1.880 10.749 0.000
Elongation Uniformity –3.504 –12.527 0.000
0.287 2.741 0.007
Yarn cnt. (Ne)
Roving CVm %
UHML
Roving cnt. (Ne)
Mic.
–0.089 –8.063 0.000
–0.687 –7.559 0.000
0.556 4.611 0.000
–2.639 –3.981 0.000
–1.145 –3.853 0.000
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu
363
Figure 1 Experimental and predicted values of yarn tenacity.
Figure 2 Experimental and predicted values of yarn elongation.
ever, Ethridge and Zhu [2], Hunter [3], Ramey [5] and some other researchers have also found a similar negative trend. The model’s coefficient of multiple determination R2 (= 0.954), adjusted R2 (= 0.952) and standard error of estimate (= 0.838) show that the predictive power of the model is very high. Figure 1 show the scatter plot of predicted values versus experimental values and regression line of our model. Due to extremely high strength values of the second and seventh cotton samples in comparison with the values of the other samples, there are two clusters in the figure.
roving count (i.e. draft ratio on the ring spinning machine). The elongation increased with higher twist and decreased with finer rovings. Coarser yarns have more elongation than finer yarns. Among fiber properties fiber elongation and fiber fineness have more influence on the yarn elongation and both parameters had a negative effect. Other important fiber parameters are reflectance, fiber strength, short fiber index and length, respectively. Table 4 shows the linear regression analysis results. We had worse R2 (0.720), and adjusted R2 (0.705) values in comparison with other yarn properties. The analysis shows that a lot of other factors have effects on yarn elongation in addition to the fiber parameters, yarn count and twist. The standard error of estimates value for our model was 0.454. Figure 2 shows the wide spread of values around the regression line. The presence of nonlinearity between the independent variables and the dependent variables reduces the success of a linear regression model. Our statistical curve estimation analysis showed that the relationship between fiber
Influence of Fiber Properties on Yarn Elongation The backward elimination method was used for the prediction of elongation. According to our model, the breaking elongation is highly influenced by yarn count, twist and
Table 4 Regression coefficients, t-values and significance level of t-values of our linear regression model for yarn elongation. Independent variables
b* t Sig.
(Constant)
Mic.
Strength
UHML
SFI
Elongation
Rd
10.237 5.421 0.000
–1.119 –5.296 0.000
0.051 3.269 0.001
0.132 2.731 0.007
–0.142 –3.008 0.003
–0.948 –5.973 0.000
0.067 3.596 0.000
*Partial regression coefficient.
Yarn cnt. (Ne) –0.142 –16.041 0.000
tpi 0.136 7.168 0.000
Roving cnt. (Ne) –2.819 –7.875 0.000
TRJ
TRJ
364
Textile Research Journal 76(5)
Table 5 Regression coefficients, t-values and significance level of variables of our linear regression model for yarn unevenness. Independent variables (Constant) b* 110.491 t 20.808 Sig. 0.000
Yarn cnt. (Ne)
Strength
0.204 29.942 0.000
–0.669 –23.972 0.000
Roving CVm% 0.989 17.710 0.000
Rd –0.482 –11.076 0.000
Elongation Uniformity 5.485 22.462 0.000
–0.513 –9.100 0.000
+b –4.504 –13.817 0.000
Roving cnt. (Ne) 9.164 11.627 0.000
Mic.
SFI
–1.049 –4.132 0.000
0.170 3.119 0.002
*Partial regression coefficient.
properties and yarn elongation was not very linear and so the prediction power of our regression model was not very high. Some other studies [2–5, 21] have shown that there is a contrary relationship between fiber strength and fiber elongation. One possible explanation of the negative sign for fiber elongation, similar to the prediction of yarn tenacity, is that autocorrelation (between fiber strength and fiber elongation) could cause instability in the estimation coefficients. The positioning of fibers in the yarn cross-section, gripping of fibers by other neighboring fibers, and slippage of fibers during yarn rupture are all decisive factors for both tenacity and elongation of yarns. These factors increase the non-linear impacts of the fiber properties on yarn tensile properties. Therefore it is thought that further investigations are needed to provide better explanations of the effects of fiber strength and fiber elongation on yarn tensile properties.
Influence of Fiber Properties on Yarn Unevenness Evenness is one of the most important properties. Our analyses show that yarn unevenness is mainly affected by yarn count and roving unevenness. Among fiber properties, strength has the greatest effect on the yarn unevenness and higher fiber strength leads to a better yarn evenness value and furthermore, fibers may be prevented from rupture due to higher strength. Reflectance, elongation, uniformity index, yellowness, fineness and short fiber content are other important factors that influence yarn evenness. The roving count also has an important effect as a finer roving has a higher unevenness value and this can be transferred onto the yarn. Table 5 shows the results of the regression analysis. The predictive ability of our model is very high as shown in Figure 3. We obtained very high R2 (0.952), and adjusted R2 (0.949) values, and the standard error of estimates value was 0.533.
Figure 3 Experimental and predicted values of yarn unevenness.
Influence of Fiber Properties on Yarn Hairiness One of the basic conditions of linear regression is that the relationship between the dependent variable and each independent variable should be linear. The curve estimation analysis showed that the micronaire value (Mic) is related linearly to yarn hairiness by the following quadratic form: Quad(Mic) = 32.894 – 12.327 × Mic + 1.405 × Mic2 (1) Table 6 shows the regression coefficients, t-values and significance level of each variable. For the independent variables’ the signs (+) or (–) indicates the direction of their effect on hairiness. From the table, it can be seen that yarn twist is the most important factor affecting the hairiness value. Among fiber properties strength has the maximum effect on yarn hairiness and is followed by elongation and length.
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu
365
Table 6 Regression coefficients, t-values and significance level of variables of our linear regression model for yarn hairiness. Independent variables (Constant) b* t Sig.
18.958 5.752 0.000
Tpi –0.129 –9.447 0.000
Strength –0.107 –8.580 0.000
Roving CVm% 0.394 11.575 0.000
Elongation
UHML
Quad (mic.)
Yarn cnt. (Ne)
+b
Uniformity
1.107 11.467 0.000
–0.113 –1.932 0.050
1.063 5.560 0.000
–0.020 –3.148 0.002
–0.372 –4.507 0.000
–0.173 –3.709 0.000
*Partial regression coefficient.
Figure 4 Experimental and predicted values of yarn hairiness.
The R2 values show that the model gave good prediction results (R2 = 0.843, adjusted R2 = 0.835, std. error of estimate = 0.326). Figure 4 show the scatter plot of predicted values versus experimental values and the regression line of the model. The model indicates that increasing fiber strength, length, yellowness, uniformity index, yarn twist and yarn count (finer yarn) will reduce yarn hairiness. A greater unevenness of rovings and fiber elongation values increased the hairiness value.
Conclusion Yarn properties are influenced by fiber properties, roving properties, yarn count and twist. In this study we tried to predict the most important yarn parameters of ring spun cotton yarns with linear multiple regression analysis by using these parameters. Our curve estimation tests indicate
that the relationships between fiber properties and yarn properties are almost linear. Therefore our models have good prediction performance. The regression coefficient R2 values of our models are very high and have greater significance. We also selected roving number and roving unevenness value as independent variables in addition to fiber properties. Our tests indicated that roving properties have considerable effects on all yarn properties. Yarn count and twist are the most decisive factors for yarn properties. Among the fiber properties, strength, elongation and fineness also have great importance and other important fiber properties are uniformity index, length and short fiber content. Our models show that reflectance (Rd) and yellowness (+b) also have important effects on some yarn properties. These fiber properties are most directly linked to the growth environment. Bright, creamy-white fibers, which have higher reflectance, are more mature. Premature termination of fiber maturation by application of growth regulators, frost, or drought characteristically increases the saturation of the yellowness (+b). It means that higher yellowness is related to poorer quality. Obviously, we expected a negative relationship between Rd and (+b) values. However we found a positive effect of both parameters on yarn properties in the regression models. There are some possible reasons for this result. First, we used good quality cotton samples. All samples were grown under normal weather conditions and environment and so their color values are acceptable for all spinners. Second, the cotton samples were tested after finisher drawframe passages. The carding and combing processes removes trash and dust and once these are removed, the Rd values increase and (+b) values decreases. Therefore both parameters seem to have good effect on yarn properties. Ethridge [2], Sasser [13] and some other researchers’ models, and also Uster’s spinning consistency index (SCI) includes both Rd and (+b) values with positive signs. Autocorrelation between fiber properties can cause instability in coefficients. Higher correlation between Rd and (+b) is another possible reason for this results. The same problem also arose for the prediction of tensile properties.
TRJ
TRJ
366
Textile Research Journal 76(5)
The good performance of the linear regression method in explaining the yarn properties indicates that the relationships between fiber properties and yarn properties are almost linear. However, further research should be done to eliminate the effects on yarn properties of nonlinear impacts and the autocorrelation problems between the independent variables.
9.
10.
11. 12.
Literature Cited
13.
1.
14.
2.
3.
4.
5.
6.
7. 8.
El Mogahzy, Y., Broughton, R. M., and Lynch, W. K., Statistical Approach for Determining the Technological Value of Cotton Using HVI Fiber Properties, Textile Res. J. 60, 495–500 (1990). Ethridge, M. D., and Zhu, R., Prediction of Rotor Spun Cotton Yarn Quality: A Comparison of Neural Network and Regression Algorithms, in “Proceedings of the Beltwide Cotton Conference,” Vol. 2: 1314–1317 (1996). Hunter, L., Prediction of Cotton Processing Performance and Yarn Properties from HVI Test Results, Melliand Textilber. 69, E 123–124 (1988). Majumdar, P. K., and Majumdar A., Predicting the Breaking Elongation of Ring Spun Cotton Yarns Using Mathematical, Statistical, and Artificial Neural Network Models, Textile Res. J., 74, 652–-655 (2004). Ramey, H. H., Jr., Lawson, R., and Worley, S., Jr., Relationship of Cotton Fiber Properties to Yarn Tenacity, Textile Res. J. 47, 685 (1977). Hunter, L., Predicting Cotton Yarn Properties from Fibre Properties in Practice, presented at “27th International Cotton Conference,” Bremen, March 24–27 (2004). Cheng, L., and Adams, D. L., Yam Strength Prediction Using Neural Networks, Textile Res. J. 65 (9), 495–500 (1995). Ramesh, M. C., Rajamanickam, R., and Jayaraman, S., Prediction of Yam Tensile Properties Using Artificial Neural Networks, J. Textile Inst. 86, 459–469 (1995).
15. 16. 17.
18.
19.
20.
21.
Zhu, R., and Ethridge, D., Predicting Hairiness for Ring and Rotor Spun Yarns and Analyzing the Impact of Fiber Properties, Textile Res. J. 67, 694–698 (1997). Chattopadhyay, R., and Guha, A., “Artificial Neural Networks: Applications to Textile”, The Textile Institute, Manchester, 2004. Frydrych, I., A New Approach for Predicting Strength Properties of Yarns, Textile Res. J. 62, 340–348 (1992). Pan, N., Relationship between Fiber and Yarn Strength, Textile Res. J. 71, 960–964 (2001). Sasser, P., Shofner, C. K., Chu, Y. T., Shofner, F. M., and Townes, M. G., Interpretations of Single Fiber, Bundle, and Yam Tenacity Data, Textile Res. J. 61, 681 (1991). Subramanian, T. A., Ganesh, K., and Bandyopadhyay S., A Generalized Equation for Predicting the Lea Strength of Ring Spun Cotton Yarns, Textile Res. J. 43, 307–313 (1973) Subramanian, T.A., A Solution for Estimating Cotton Yarn Tenacity, Int. Textile Bul., 4, 36–38 (2004). Hearle, J. W. S., “Structural Mechanics of Yarns and Fabrics,” Vol. 1, Wiley-Interscience, New York, 1969. Fiori, L. A., Sands, J. E., Little, H. W., and Grant, J. N., Effect of Cotton Fiber Bundle Break Elongation and Other Fiber Properties on the Properties of a Coarse and a Medium Single Yam, Textile Res. J. 26, 553–564 (1956). Aggarwal, S. K., A Model to Estimate the Breaking Elongation of High Twist Ring Spun Cotton Yarns, Part I: Derivation of the Model for Yarns from Single Cotton Varieties, Textile Res. J. 59, 691–695 (1989). Aggarwal, S. K., A Model to Estimate the Breaking Elongation of High Twist Ring Spun Cotton Yarns, Part II: Applicability to Yarns from Mixtures of Cottons, Textile Res. J. 59, 717–720 (1989). Zurek, W., Frydrych, I., and Zakrzewski, S., A Method of Predicting the Strength and Breaking Strain of Cotton Yarn, Textile Res. J. 57, 439–444 (1987). El Mogahzy, Y., and Broughton, R. M., Regressional Observations of HVI Fiber Properties, Yarn Quality, and Processing Performance of Medium Staple Cotton. Part I: HVI Fiber Parameters, Textile Res. J. 62(4), 218–226 (1992).
Article
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties Abstract
The main aim of the present study was to predict the most important yarn quality characteristics derived from cotton fiber properties that were measured by means of an HVI system. With this aim 15 different cotton blends were selected from different spinning mills in Turkey. The cotton fibers were processed in the short staple spinning line at Ege University Textile and Apparel Research–Application Center and were spun into ring yarns (20s, 25s, 30s and 35s). Each count was spun at three different coefficients of twist (αe 3.8, αe 4.2 and αe 4.6). Linear multiple regression methods were used for the estimation of yarn quality characteristics. Yarn count, twist and roving properties all had considerable effects on the yarn properties and therefore these parameters were also selected as predictors. After the goodness of fit statistics very large R2 (coefficient of multiple determination) and adjusted R2 values were observed. Furthermore, analysis of variance tables showed that our equations were significant at the α = 0.01 significance level.
Mustafa E. Üreyen1 and Hüseyin Kadoglu2 Textile Engineering Department, Ege University, 35100 Izmir, Turkey
Key words cotton yarn, HVI, cotton fiber, ring spinning, regression analysis, yarn quality estimation
The physical characteristics of a fiber determine its processing behavior, production efficiency and finally yarn and fabric quality. Therefore, predicting the quality characteristics of yarns, such as the tensile properties from the raw material properties, was the main purpose of many studies in the last century. In addition to raw material processing conditions, preparation stages, machine parameters and the spinning method also have considerable effects on the yarn properties. Generally two approaches were used in these studies for predicting yarn quality from fiber and yarn characteristics: • an empirical and statistical approach; and • a theoretical or analytical approach.1
Textile Research Journal Vol 76(5): 360–366 DOI: 10.1177/0040517506062262 Figures 1–4 appear in color online: http://trj.sagepub.com
The empirical and statistical approach to establishing a relationship between fiber and yarn quality characteristics has been the most popular method during the second half of the twentieth century. Fast and accurate measurement of fiber properties by means of high volume instruments (HVI) and more powerful computers are the two main reasons for this tendency. With this method, important fiber and yarn properties can be measured for a range of samples and by using these results empirical relationships have been established by means of statistical analysis. One of the
1 2
E-mail: [email protected] Corresponding author: e-mail: [email protected]
www.trj.sagepub.com © 2006 SAGE Publications
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu most common statistical approaches is the multiple regression method (e.g. linear, log–log, etc.). Such an approach is used to investigate the interdependence of the different fiber properties and to estimate the relative contribution of each fiber property to the overall yarn properties. Several researchers [1–5] have established various regression equations using this method. The theoretical approach is based on physical and mechanical principles. These models usually yield good information about interactions between different fiber properties and yarn characteristics. However, practical applications are almost impossible because of the complexities of the models. They are usually based on certain assumptions and their success is determined by the feasibility of these assumptions [4, 6]. In recent years some researchers [2, 4, 7–9] have shown an interest in the use of artificial neural networks (ANN) to predict yarn characteristics. This analytical system is also useful for discovering relationships between variables [9]. Chattopadhyay and Guha [10] have reviewed textile applications of artificial neural networks in detail. The tensile properties of a spun yarn have always been very important in determining the quality of the yarn, since they directly affect the winding and knitting efficiency as well as warp and weft breakages during weaving. It is, therefore, important to establish which fiber and yarn parameters influence the yarn tensile properties and if possible, to derive functional relationship between them. So far, numerous mathematical and empirical models have been established for the estimation of single yarn tenacity [3, 5, 11–13] and count strength product (CSP) [1, 3, 14, 15] using fiber properties and some yarn parameters. Hearle [16] reviewed various mathematical and empirical studies concerning yarn strength, which were published between 1926 and 1965. Hunter [6] reported on more than 200 published papers about the prediction of yarn quality parameters, particularly tensile properties, up to 2004. Another important yarn parameter is breaking elongation. It influences the performance of spun yarns during winding, warping, and weaving. Yarn elongation is chiefly influenced by fiber properties, yarn twist, and yarn count
361
[17]. However, prediction models dealing with the breaking elongation of cotton yarns are few in number. Mathematical models have been proposed by Aggarwal [18, 19], Frydrych [11], and Zurek et al. [20]. Statistical models have been developed by Hunter [3] and ANN models produced by Majumdar [4]. Unevenness is a very important factor for the yarn and fabric quality. Cross-sectional fiber variation is the basic reason for unevenness. In addition to machine parameters, spinning method, yarn count and some fiber parameters have decisive influence on the unevenness of the yarn. Hunter [3] and Ethridge et al. [2] have developed some models to determine yarn irregularity by using fiber parameters. Hairiness, another measurable yarn characteristic, is usually an undesirable property. Acceptable measuring devices for the determination of hairiness, such as the Uster Tester and Zweigle Hairiness Tester are relatively recent and therefore fewer research articles have been published on the estimation of hairiness by using fiber parameters. In this study, we developed statistical models to estimate ring yarn tensile properties, unevenness and hairiness values by using multiple linear regressions. In addition to HVI fiber properties, we also used roving properties, yarn count, and yarn twist.
Experimental In this work, a total of fifteen different cotton samples were collected in roving form from various spinning mills. The spinning operations can affect the fiber properties in different ways, depending on the machinery line and adjustments, etc. For the elimination of these effects, fiber properties were measured from finisher drawing slivers by using the Uster HVI testing system. Table 1 shows the fiber properties measured by HVI. All samples were spun into yams on ring spinning machine (Rieter Model G30) at a yarn count of Ne 20 (29.53 tex), Ne 25 (23.63 tex), Ne 30 (19.69 tex), and Ne 35 (16.88 tex). Each yarn count was spun at three different twist multipli-
Table 1 Main fiber properties of the tested samples. Property Fineness (microner) Strength (g/tex) Length (UHML) Uniformity (%) SFI (%) Elongation (%) Reflectance (Rd) Yellowness (+b)
Sample no. 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
4.7 31.6 30.30 86.6 3.5 6.1 82.2 9.4
4.0 50.6 32.60 86.8 3.5 7.4 77.7 9.2
4.3 33.2 28.55 86.5 3.5 6.8 79.9 8.9
4.8 32.0 29.35 85.9 3.5 6.2 81.9 8.9
4.1 34.2 27.70 83.5 6.7 5.8 78.2 9.8
4.7 33.7 29.95 86.9 3.5 6.4 82.4 9.3
4.2 47.8 35.10 90.8 3.5 7.7 80.6 9.4
3.9 30.3 29.90 84.8 4.1 6.8 68.1 11.0
4.5 33.8 29.25 85.5 3.6 6.7 80.9 9.4
4.9 34.3 29.50 86.2 3.5 6.6 81.8 9.0
4.9 31.7 27.60 83.4 6.8 5.6 80.2 9.1
4.3 33.7 29.45 86.2 3.5 6.7 80.9 9.5
4.9 33.6 29.90 87.4 3.5 6.4 82.9 9.1
4.5 38.8 30.55 87.7 3.5 7.6 81.4 9.4
4.5 33.1 27.45 82.4 8.1 5.7 77.8 8.9
TRJ
TRJ
362
Textile Research Journal 76(5)
Table 2 Main properties of rovings. Property Roving cnt. (Ne) Um (%) CVm (%) CVm (1m) (%)
Sample no. 1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
1.09 4.11 5.06 2.96
1.08 4.95 6.29 2.82
0.66 3.23 4.06 2.17
0.93 4.00 5.07 2.48
1.22 4.68 5.92 3.01
1.22 3.90 4.91 1.96
1.11 5.24 6.69 4.06
0.89 5.35 6.77 3.04
1.05 4.90 6.16 2.31
1.00 4.42 5.56 1.61
1.01 3.91 4.93 2.68
1.02 5.72 7.24 2.82
1.00 5.06 6.42 4.05
0.90 3.61 4.56 1.46
1.00 4.98 6.29 2.02
ers (αe 3.8, αe 4.2, αe 4.6). A total of 180 spinning trials were done. The appropriate drafting ratios were adjusted on the ring spinning machine for each sample. Other spinning conditions were kept constant. Orbit rings (42 mm diameter) and travellers (suitable weights were selected for each yarn count) were used. For each yarn sample ten cops were produced and tested. The tensile properties of the yarns were evaluated on an Uster Tensorapid tensile testing machine. Unevenness and hairiness tests were performed on an Uster Tester 3. Rovings were tested on the Uster Tester 3. The measurements of the main properties are shown in Table 2. The multiple regression method has the advantage of simplicity in describing the quantitative relationship between textile material properties. Therefore this method was selected for establishing the relationships between fiber and yarn properties. At the beginning the types of relationship between selected properties (independent variables) and yarn properties (dependent variable) were checked individually. The statistical tests indicated that there was a nearly linear relationship between fiber properties and yarn properties. Therefore a linear multiple regression analysis method was chosen for our study in order to establish a quantitative relationship of yarn properties with respect to fiber properties, roving properties, yarn count, and yarn twist. Forward stepwise (for estimation of tenacity, unevenness and hairiness) and backward elimination (for estimation of elongation) procedures were selected for the linear regression analysis. Statistical analyses were performed using the SPSS 11.0.1 and Minitab 11.12 programs.
Results and Discussion Influence of Fiber Properties on Yarn Tenacity As might be expected, a high positive correlation was found between fiber and yarn strength (0.905). After regression analysis it was seen that fiber strength was the most important parameter for yarn tenacity. Fiber elongation, fiber length (UHML), uniformity index, fiber fineness, yarn count, yarn twist, roving count and unevenness of roving are other parameters that can have a significant influence on yarn tenacity. Table 3 shows the regression coefficients of variables, the t-values and significance level of each variable. The estimated coefficients of variables indicate the direction of influence. As the strength of a yarn depends on the weakest place, an increase in irregularity should lead to a lower strength [16]. Therefore any unevenness of the rovings has a negative effect on the yarn tenacity. Roving count determines the draft on the ring spinning machine. Our model shows that the drafting ratio on the spinning machine has a positive influence on tenacity, but it should be noted that in our study the drafting ratio ranged from 18.80 to 54.20. We found a negative effect between fiber elongation and yarn tenacity. A stepwise method was used for the regression analysis. The significance values obtained were based on fitting a single model and in this situation the significance values can sometimes be invalid when a stepwise method is used. Furthermore, autocorrelation between fiber properties is another basic problem for linear regression analysis. How-
Table 3 Regression coefficients, t-values and significance level of t-values of our linear regression model for yarn tenacity. Independent variables (Constant) Strength b* –19.883 t –3.632 Sig. 0.000
0.701 29.900 0.000
*Partial regression coefficient.
Twist co. (αe) 1.880 10.749 0.000
Elongation Uniformity –3.504 –12.527 0.000
0.287 2.741 0.007
Yarn cnt. (Ne)
Roving CVm %
UHML
Roving cnt. (Ne)
Mic.
–0.089 –8.063 0.000
–0.687 –7.559 0.000
0.556 4.611 0.000
–2.639 –3.981 0.000
–1.145 –3.853 0.000
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu
363
Figure 1 Experimental and predicted values of yarn tenacity.
Figure 2 Experimental and predicted values of yarn elongation.
ever, Ethridge and Zhu [2], Hunter [3], Ramey [5] and some other researchers have also found a similar negative trend. The model’s coefficient of multiple determination R2 (= 0.954), adjusted R2 (= 0.952) and standard error of estimate (= 0.838) show that the predictive power of the model is very high. Figure 1 show the scatter plot of predicted values versus experimental values and regression line of our model. Due to extremely high strength values of the second and seventh cotton samples in comparison with the values of the other samples, there are two clusters in the figure.
roving count (i.e. draft ratio on the ring spinning machine). The elongation increased with higher twist and decreased with finer rovings. Coarser yarns have more elongation than finer yarns. Among fiber properties fiber elongation and fiber fineness have more influence on the yarn elongation and both parameters had a negative effect. Other important fiber parameters are reflectance, fiber strength, short fiber index and length, respectively. Table 4 shows the linear regression analysis results. We had worse R2 (0.720), and adjusted R2 (0.705) values in comparison with other yarn properties. The analysis shows that a lot of other factors have effects on yarn elongation in addition to the fiber parameters, yarn count and twist. The standard error of estimates value for our model was 0.454. Figure 2 shows the wide spread of values around the regression line. The presence of nonlinearity between the independent variables and the dependent variables reduces the success of a linear regression model. Our statistical curve estimation analysis showed that the relationship between fiber
Influence of Fiber Properties on Yarn Elongation The backward elimination method was used for the prediction of elongation. According to our model, the breaking elongation is highly influenced by yarn count, twist and
Table 4 Regression coefficients, t-values and significance level of t-values of our linear regression model for yarn elongation. Independent variables
b* t Sig.
(Constant)
Mic.
Strength
UHML
SFI
Elongation
Rd
10.237 5.421 0.000
–1.119 –5.296 0.000
0.051 3.269 0.001
0.132 2.731 0.007
–0.142 –3.008 0.003
–0.948 –5.973 0.000
0.067 3.596 0.000
*Partial regression coefficient.
Yarn cnt. (Ne) –0.142 –16.041 0.000
tpi 0.136 7.168 0.000
Roving cnt. (Ne) –2.819 –7.875 0.000
TRJ
TRJ
364
Textile Research Journal 76(5)
Table 5 Regression coefficients, t-values and significance level of variables of our linear regression model for yarn unevenness. Independent variables (Constant) b* 110.491 t 20.808 Sig. 0.000
Yarn cnt. (Ne)
Strength
0.204 29.942 0.000
–0.669 –23.972 0.000
Roving CVm% 0.989 17.710 0.000
Rd –0.482 –11.076 0.000
Elongation Uniformity 5.485 22.462 0.000
–0.513 –9.100 0.000
+b –4.504 –13.817 0.000
Roving cnt. (Ne) 9.164 11.627 0.000
Mic.
SFI
–1.049 –4.132 0.000
0.170 3.119 0.002
*Partial regression coefficient.
properties and yarn elongation was not very linear and so the prediction power of our regression model was not very high. Some other studies [2–5, 21] have shown that there is a contrary relationship between fiber strength and fiber elongation. One possible explanation of the negative sign for fiber elongation, similar to the prediction of yarn tenacity, is that autocorrelation (between fiber strength and fiber elongation) could cause instability in the estimation coefficients. The positioning of fibers in the yarn cross-section, gripping of fibers by other neighboring fibers, and slippage of fibers during yarn rupture are all decisive factors for both tenacity and elongation of yarns. These factors increase the non-linear impacts of the fiber properties on yarn tensile properties. Therefore it is thought that further investigations are needed to provide better explanations of the effects of fiber strength and fiber elongation on yarn tensile properties.
Influence of Fiber Properties on Yarn Unevenness Evenness is one of the most important properties. Our analyses show that yarn unevenness is mainly affected by yarn count and roving unevenness. Among fiber properties, strength has the greatest effect on the yarn unevenness and higher fiber strength leads to a better yarn evenness value and furthermore, fibers may be prevented from rupture due to higher strength. Reflectance, elongation, uniformity index, yellowness, fineness and short fiber content are other important factors that influence yarn evenness. The roving count also has an important effect as a finer roving has a higher unevenness value and this can be transferred onto the yarn. Table 5 shows the results of the regression analysis. The predictive ability of our model is very high as shown in Figure 3. We obtained very high R2 (0.952), and adjusted R2 (0.949) values, and the standard error of estimates value was 0.533.
Figure 3 Experimental and predicted values of yarn unevenness.
Influence of Fiber Properties on Yarn Hairiness One of the basic conditions of linear regression is that the relationship between the dependent variable and each independent variable should be linear. The curve estimation analysis showed that the micronaire value (Mic) is related linearly to yarn hairiness by the following quadratic form: Quad(Mic) = 32.894 – 12.327 × Mic + 1.405 × Mic2 (1) Table 6 shows the regression coefficients, t-values and significance level of each variable. For the independent variables’ the signs (+) or (–) indicates the direction of their effect on hairiness. From the table, it can be seen that yarn twist is the most important factor affecting the hairiness value. Among fiber properties strength has the maximum effect on yarn hairiness and is followed by elongation and length.
Regressional Estimation of Ring Cotton Yarn Properties from HVI Fiber Properties M. E. Üreyen and H. Kadoglu
365
Table 6 Regression coefficients, t-values and significance level of variables of our linear regression model for yarn hairiness. Independent variables (Constant) b* t Sig.
18.958 5.752 0.000
Tpi –0.129 –9.447 0.000
Strength –0.107 –8.580 0.000
Roving CVm% 0.394 11.575 0.000
Elongation
UHML
Quad (mic.)
Yarn cnt. (Ne)
+b
Uniformity
1.107 11.467 0.000
–0.113 –1.932 0.050
1.063 5.560 0.000
–0.020 –3.148 0.002
–0.372 –4.507 0.000
–0.173 –3.709 0.000
*Partial regression coefficient.
Figure 4 Experimental and predicted values of yarn hairiness.
The R2 values show that the model gave good prediction results (R2 = 0.843, adjusted R2 = 0.835, std. error of estimate = 0.326). Figure 4 show the scatter plot of predicted values versus experimental values and the regression line of the model. The model indicates that increasing fiber strength, length, yellowness, uniformity index, yarn twist and yarn count (finer yarn) will reduce yarn hairiness. A greater unevenness of rovings and fiber elongation values increased the hairiness value.
Conclusion Yarn properties are influenced by fiber properties, roving properties, yarn count and twist. In this study we tried to predict the most important yarn parameters of ring spun cotton yarns with linear multiple regression analysis by using these parameters. Our curve estimation tests indicate
that the relationships between fiber properties and yarn properties are almost linear. Therefore our models have good prediction performance. The regression coefficient R2 values of our models are very high and have greater significance. We also selected roving number and roving unevenness value as independent variables in addition to fiber properties. Our tests indicated that roving properties have considerable effects on all yarn properties. Yarn count and twist are the most decisive factors for yarn properties. Among the fiber properties, strength, elongation and fineness also have great importance and other important fiber properties are uniformity index, length and short fiber content. Our models show that reflectance (Rd) and yellowness (+b) also have important effects on some yarn properties. These fiber properties are most directly linked to the growth environment. Bright, creamy-white fibers, which have higher reflectance, are more mature. Premature termination of fiber maturation by application of growth regulators, frost, or drought characteristically increases the saturation of the yellowness (+b). It means that higher yellowness is related to poorer quality. Obviously, we expected a negative relationship between Rd and (+b) values. However we found a positive effect of both parameters on yarn properties in the regression models. There are some possible reasons for this result. First, we used good quality cotton samples. All samples were grown under normal weather conditions and environment and so their color values are acceptable for all spinners. Second, the cotton samples were tested after finisher drawframe passages. The carding and combing processes removes trash and dust and once these are removed, the Rd values increase and (+b) values decreases. Therefore both parameters seem to have good effect on yarn properties. Ethridge [2], Sasser [13] and some other researchers’ models, and also Uster’s spinning consistency index (SCI) includes both Rd and (+b) values with positive signs. Autocorrelation between fiber properties can cause instability in coefficients. Higher correlation between Rd and (+b) is another possible reason for this results. The same problem also arose for the prediction of tensile properties.
TRJ
TRJ
366
Textile Research Journal 76(5)
The good performance of the linear regression method in explaining the yarn properties indicates that the relationships between fiber properties and yarn properties are almost linear. However, further research should be done to eliminate the effects on yarn properties of nonlinear impacts and the autocorrelation problems between the independent variables.
9.
10.
11. 12.
Literature Cited
13.
1.
14.
2.
3.
4.
5.
6.
7. 8.
El Mogahzy, Y., Broughton, R. M., and Lynch, W. K., Statistical Approach for Determining the Technological Value of Cotton Using HVI Fiber Properties, Textile Res. J. 60, 495–500 (1990). Ethridge, M. D., and Zhu, R., Prediction of Rotor Spun Cotton Yarn Quality: A Comparison of Neural Network and Regression Algorithms, in “Proceedings of the Beltwide Cotton Conference,” Vol. 2: 1314–1317 (1996). Hunter, L., Prediction of Cotton Processing Performance and Yarn Properties from HVI Test Results, Melliand Textilber. 69, E 123–124 (1988). Majumdar, P. K., and Majumdar A., Predicting the Breaking Elongation of Ring Spun Cotton Yarns Using Mathematical, Statistical, and Artificial Neural Network Models, Textile Res. J., 74, 652–-655 (2004). Ramey, H. H., Jr., Lawson, R., and Worley, S., Jr., Relationship of Cotton Fiber Properties to Yarn Tenacity, Textile Res. J. 47, 685 (1977). Hunter, L., Predicting Cotton Yarn Properties from Fibre Properties in Practice, presented at “27th International Cotton Conference,” Bremen, March 24–27 (2004). Cheng, L., and Adams, D. L., Yam Strength Prediction Using Neural Networks, Textile Res. J. 65 (9), 495–500 (1995). Ramesh, M. C., Rajamanickam, R., and Jayaraman, S., Prediction of Yam Tensile Properties Using Artificial Neural Networks, J. Textile Inst. 86, 459–469 (1995).
15. 16. 17.
18.
19.
20.
21.
Zhu, R., and Ethridge, D., Predicting Hairiness for Ring and Rotor Spun Yarns and Analyzing the Impact of Fiber Properties, Textile Res. J. 67, 694–698 (1997). Chattopadhyay, R., and Guha, A., “Artificial Neural Networks: Applications to Textile”, The Textile Institute, Manchester, 2004. Frydrych, I., A New Approach for Predicting Strength Properties of Yarns, Textile Res. J. 62, 340–348 (1992). Pan, N., Relationship between Fiber and Yarn Strength, Textile Res. J. 71, 960–964 (2001). Sasser, P., Shofner, C. K., Chu, Y. T., Shofner, F. M., and Townes, M. G., Interpretations of Single Fiber, Bundle, and Yam Tenacity Data, Textile Res. J. 61, 681 (1991). Subramanian, T. A., Ganesh, K., and Bandyopadhyay S., A Generalized Equation for Predicting the Lea Strength of Ring Spun Cotton Yarns, Textile Res. J. 43, 307–313 (1973) Subramanian, T.A., A Solution for Estimating Cotton Yarn Tenacity, Int. Textile Bul., 4, 36–38 (2004). Hearle, J. W. S., “Structural Mechanics of Yarns and Fabrics,” Vol. 1, Wiley-Interscience, New York, 1969. Fiori, L. A., Sands, J. E., Little, H. W., and Grant, J. N., Effect of Cotton Fiber Bundle Break Elongation and Other Fiber Properties on the Properties of a Coarse and a Medium Single Yam, Textile Res. J. 26, 553–564 (1956). Aggarwal, S. K., A Model to Estimate the Breaking Elongation of High Twist Ring Spun Cotton Yarns, Part I: Derivation of the Model for Yarns from Single Cotton Varieties, Textile Res. J. 59, 691–695 (1989). Aggarwal, S. K., A Model to Estimate the Breaking Elongation of High Twist Ring Spun Cotton Yarns, Part II: Applicability to Yarns from Mixtures of Cottons, Textile Res. J. 59, 717–720 (1989). Zurek, W., Frydrych, I., and Zakrzewski, S., A Method of Predicting the Strength and Breaking Strain of Cotton Yarn, Textile Res. J. 57, 439–444 (1987). El Mogahzy, Y., and Broughton, R. M., Regressional Observations of HVI Fiber Properties, Yarn Quality, and Processing Performance of Medium Staple Cotton. Part I: HVI Fiber Parameters, Textile Res. J. 62(4), 218–226 (1992).
