This is the Pre-Published Version.

This is the accepted version of the publication Yin R, Tao XM and Xu BG. Systematic investigation of twist generation and propagation in a modified
ring spinning system. Textile Research Journal (vol 90, no. 3-4) pp. 367-375. Copyright © 2020 (The Author(s)). DOI: 10.1177/0040517519866950

Systematic Investigation of Twist Generation and Propagation in A

Modified Ring Spinning System

R Yin, X M Tao* and B G Xu

Institute of Textiles and Clothing, The Hong Kong Polytechnic University, Hong Kong, China
E-mail: [email protected]

Abstract:

Twisting is an important process to form a continuous yarn from short fibers and to determine the
structure and properties of the resultant yarn. This paper systematically examined yarn twisting
process in a modified ring spinning process based on a theoretical model proposed recently. In order
to reduce the number of experiments, response surface methodology (RSM) involving a central
composite design (CCD) in three factors of twist multiplier, speed ratio and wrap angle was
successfully employed for the study and analysis. The significant terms of the models were studied,
and it was discovered that the speed ratio and wrap angle are statistically significant for the responses
of twist efficiency, propagation coefficients of twist trapping and congestion. And more importantly,
linear relationships were found among the three responses.

Keywords: yarn motion, twist generation, twist propagation, ring spinning

Introduction

Spinning is a fundamental method to produce long strands from staple raw fibers of cotton, wool,
flax, or other material.1 Twisting is a vital process to determine the staple yarn structure and
performances like strength (tenacity), elongation, evenness and hairiness.2 Till now, ring spinning
continues to predominate in yarn manufacturing industry due to its high yarn quality and flexibility in
materials and yarn counts.3-5 According to the latest statistics, the total number of spindles around
world was astonishing 244 million.6 With the increasing demand of novel features or improving
qualities, many modifications have been developed, such as Compact7, 8, Siro 9, 10
and Solo11-14. In
recent years, a novel spinning technology, named Nu-Torque15-17, has been developed by introducing
a false-twisting unit into the conventional ring frame for producing low twist and soft handle single
yarns. The modified cotton yarns and fabrics have significant advantages in terms of soft handle,
higher yarn strength at lower twist factor, lower residual torque and low knitted fabric spirality after
washing and tumble-dry cycles. 18-24 Among over ten mills using the technology, 18 to 40% increment
in production rate has been achieved for cotton ring yarns with various versions of the technology. In
addition, a significant average energy saving of 337KWh/ton was reported by a mill producing Ne 30
and 1100 KWh/ton for another mill for spinning Ne 80 yarns.
Despite achieved low twist and soft handle single yarns, fundamental mechanism in terms of
false twisting process needs to be addressed. Twist generation and propagation25, 26, which are key
issues leading to the essence of spinning, including the amount of false-twist generated by the
false-twister, false-twisting efficiency and blockage rate, as well as the relationships between

1
false-twisting efficiency and system parameters. The false-twisting unit employed in this study is
single friction-belt with circular cross-section. It has been identified that the friction-belt generates the
false-twists into yarn, meanwhile traps the upward propagation of the real twist as well as congests the
downward propagation of the false twists. In our recent published paper , a steady-state model of
27

yarn dynamics in the modified ring spinning system has been proposed, which deals with two
important phenomena simultaneously, that is, twist generation and twist propagation. Based on the
proposed model, in this paper a systematic evaluation of twist generation and propagation in the
modified ring spinning system has been carried out to assess effects of system parameters on the
false-twisting efficiency as well as propagation coefficients. In order to minimize the number of
experiments, response surface methodology involving a central composite design in three factors of
twist multiplier, speed ratio and wrap angle has been successfully employed for the study and analysis.
The significant terms of the models were studied and relationships among three responses were
investigated.

Twist generation and propagation

As shown in Figure 1, a simplified system is composed of front rollers at point A, a friction-belt

false-twister, which contacts with the yarn in zone BC and a real-twister at point D. Hence, there are
two twist units in the system: one is the real-twister at point D; another is the false-twister which
generates the frictional moment at zone BC.
Three concepts are introduced to describe the roles played by friction, correspondingly, three
coefficients are defined. The first parameter is the twisting efficiency of the false-twister. To determine
it, let R0 be the radius of the yarn, nc = Tc v be the rotational speed of the yarn, Tc be the total

twist generated by the moving belt, v be the delivery speed of the yarn, and vb be the moving
speed of the belt, the twist efficiency of the moving belt is expressed as
2π R0Tc v
λ= (1)
vb

The second effect is the twist trapping in the up-ward propagation of the real twist inserted by D.
To quantify this effect, let Tt be the total twist lost in zone BC, and TCD be the twist in zone CD,

then the propagation coefficient of twist trapping is defined as

Tt
k = 1− (2)
TCD
The last effect of friction is the twist congestion, which occurs in the downward propagation of
twist in zone AB. Let Th be the total twist increment in zone AB, then the propagation coefficient of

2
the twist congestion is defined as
Tc
η= (3)
Tc + Th
From the kinematic point of view, the twist in zone AB in Figure 1 can be expressed as
λTBC
TAB kTCD +
= (4)
η
vb
where TBC = is the theoretical twist generated by the false-twister, and all three coefficients
2π R0 v
range from 0 to 1.

Figure 1 A simplified spinning system

Experimental design

The experiments were conducted on a ring spinning frame (Zinser 351) by installing a moving
friction-belt with diameter of 6mm between the front rollers and the yarn guide. A cotton yarn with
linear density of Ne 32 and diameter of 0.16 mm was adopted for the experiments. Three important
parameters, twist factor (X1), speed ratio (X2), and wrap angle (X3) were chosen as the potential
influencing, and twist efficiency (Y1), propagation coefficients of twist trapping (Y2) and congestion
(Y3) were the dependent variables calculated by our proposed model27. Although yarn tension may
have a large impact on twist generation and propagation, it was not considered here for further

3
investigation not only because its value cannot be arbitrarily changed at such a wide range for industry
production, but also the fact that it is not a continuously adjustable parameter for the experiment.

Front rollers
X0 X
O

A0 A1
B0
Oo O1 B1

Y0 G0
Yarn guide
Y

Figure 2 Schematic view of belt position in a ring spinning frame

In order to reduce the number of experiments needed, response surface methodology (RSM)
involving a central composite design (CCD) was employed to exam the relationships between
several input parameters and one or more response variables. The range and levels of the independent
variables investigated as listed in Table 1. Speed ratio is defined as the ratio of belt moving speed to
the yarn delivery speed. Wrap angle is defined as the angle formed by the yarn and belt, ∠AOB as
shown in Figure 2. The change of wrap angle can be achieved by adjusting belt position in the
spinning frame. The independent variables were tested in an orthogonal 23 CCD with six center
points and six-star points. Each of the independent variables was conducted at five different levels as
per CCD in three variables with a total of 20 experiments 28. The plan of CCD in coded levels of three
independent variables is shown in Table 2. The statistical significance level (p value) was set at 0.05.
The coded and actual values of the design were generated in a randomized run order using Minitab 16

4
software. Based on the RSM, quadratic polynomial regression equations are developed to fit the
experimental data, as show in the following equation,
K K K −1 K
Y =C0 + ∑ Ci X i + ∑ Cii X i2 + ∑∑ Cij X i X j + ε (i < j ) (5)
=i 1 =j 1 =i 1 =j 2

where Y is the response, C0 is constant, Ci, Cii and Cij are regression coefficients and Xi are the coded
factors, ε is the error and K is the number of independent variables.

Table 1 Coded and actual level for each variables of the CCD
Variation levels
Variables Code
-1.682 -1 0 +1 +1.682
Twist factor X1 2.51 2.91 3.50 4.09 4.49
Speed ratio X2 1.01 1.41 2.00 2.59 2.99
Wrap angle (o) X3 30.00 42.16 60.00 77.84 90.00

Table 2 Coded and actual levels in experimental design by CCD method

Coded levels Independent variables Responses
Run Twist Trapping Congestion
Twist Speed Wrap
order X1 X2 X3 efficiency coefficient coefficient
factor ratio angle
(Y1) (Y2) (Y3)
1 0 -1.682 0 3.50 1.01 60.00 0.126 0.873 0.892
2 -1 -1 +1 2.91 1.41 77.84 0.166 0.829 0.860
3 0 0 0 3.50 2.00 60.00 0.124 0.874 0.895
4 -1 +1 -1 2.91 2.59 42.16 0.073 0.926 0.939
5 0 +1.682 0 3.50 2.99 60.00 0.098 0.900 0.916
6 +1 -1 +1 4.09 1.41 77.84 0.143 0.853 0.877
7 +1.682 0 0 4.49 2.00 60.00 0.130 0.867 0.889
8 +1 +1 -1 4.09 2.59 42.16 0.070 0.929 0.941
9 0 0 -1.682 3.50 2.00 30.00 0.043 0.956 0.966
10 -1.682 0 0 2.51 2.00 60.00 0.152 0.844 0.872
11 0 0 0 3.50 2.00 60.00 0.133 0.864 0.887
12 0 0 0 3.50 2.00 60.00 0.131 0.866 0.889
13 +1 -1 -1 4.09 1.41 42.16 0.103 0.896 0.912
14 0 0 +1.682 3.50 2.00 90.00 0.171 0.821 0.854
15 0 0 0 3.50 2.00 60.00 0.131 0.866 0.889
16 0 0 0 3.50 2.00 60.00 0.144 0.853 0.878
17 -1 -1 -1 2.91 1.41 42.16 0.101 0.898 0.915
18 -1 +1 +1 2.91 2.59 77.84 0.124 0.870 0.893
19 +1 +1 +1 4.09 2.59 77.84 0.109 0.886 0.905
20 0 0 0 3.50 2.00 60.00 0.107 0.890 0.908

Results and discussions

Analysis of variance (ANOVA)

ANOVA is a statistical technique to analyze the differences among group means and their

5
associated procedures, which is accomplished by subdividing the total variation in a dataset into
component parts allied with sources of variation for testing hypotheses on the variables of the model 29,
30
. Precision of a parameter estimation depends on the degree of freedom (DF), which equals to the
number of experiments subtract the number of additional parameters estimated for that calculation 31.
Followed by Fisher's statistical test (F test), ANOVA was employed to study the importance of each
independent variable 32. F value is calculated by the regression mean square divided by the real error
mean, which implies the influence of each controlled parameter on the model 33, 34. The ANOVA data
in Table A1 lists F value for twist efficiency, propagation coefficient of twist trapping and congestion
as 10.36, 11.16 and 11.33 respectively, suggesting that the regression equation is highly significant.
Generally, the large Fisher value denotes that the variation in the responses can be interpreted by the
model. The associated p value is an important parameter to estimate if F value is large enough to
display statistical significance 35, 36. The p values is the index of the significance of the test, whose
value below 0.05 indicates that the model and the associated terms are statistically significant at 95%
confidence level 37-39. The significance of each coefficient is also determined by the F and p values 40,
41
. ANOVA indicated that the highest significant level was shown by the wrap angle (X3), followed by
the speed ratio (X2) and lastly, the quadratic wrap angle (X3*X3), while the interaction terms (X1* X2,
X1* X3, X2* X3), the linear and quadratic twist factors (X1, X1* X1) were less significant.
Table A1 also shows multiple correlation coefficient (R2) and adjusted R2. The R2 implies the
variation of the response in the model 42.
The higher the R2, the better the model fits the data. The
values of R2 were calculated to be 90.31%, 90.95% and 91.07% for twist efficiency, propagation
coefficients of twist trapping and congestion, respectively, implying that the experimental data was
well-fitted. The adjusted R2 explains the number of predictors in the model and is useful for comparing
models with different numbers of predictors 43.
The high value of adjusted R2 supports a high
correlation between the experimental and the predicted data 30. In this study, the values of adjusted R2
for twist efficiency, propagation coefficients of twist trapping and congestion were 81.59%, 82.80%
and 83.03%, respectively, which means the regression models were statistically significant. The
lack-of-fit p value of the model for twist efficiency, propagation coefficients of twist trapping and
congestion were 0.334, 0.332 and 0.312, respectively, which were higher than 0.05, indicating that the
models fitted all the design points well.
Since there were some insignificant terms existing in the regression models, it was necessary to
simply them by eliminating insignificant terms. Decreasing the number of terms can make the model
manageable, meanwhile increase the precision of the predictors. By examining the F and p values of
each coefficient, it was found that all terms containing twist factor (X1) over the range of 2.51 to 4.49
has little impact on the responses, thus should be removed from the models. Moreover, the interaction
terms should also be removed as they are not significant for the models. The significant terms for the
models were wrap angle (X3), speed ratio (X2) and quadratic wrap angle (X3*X3). The quadratic wrap

6
angle (X2*X2) was reserved as it is significant at 10% level. Therefore, these four terms were
employed to reconstruct the regression formulas. The results of simplified model and ANOVA are
listed in Table A2, from which we can concluded that the reduced model was superior to the complete
quadratic equations because the reduced models have higher F values and lower p values than those of
previous ones. Moreover, all the four terms were statistically significant for the three responses and
the adjusted R2 were better than the previous ones, implying that the reduced regression models were
significant and adequate to depict the actual relationship between the responses and inputs.

Model validation
It is necessary to check the regression model to guarantee that adequate prediction to the actual
data is obtained 44. Diagnostic plots such as normal probability plots shown in Figure 3 enable to
judge the normal distribution of the residuals. The residual is the error between the experimental data
and the simulated value by the theoretical model. A small residual value represents a high accuracy of
the prediction by the model. In Figure 3, the data points on the figure were close to the straight line,
implying that the data was normally distributed.
The regression models were also investigated by nine sets of randomly selected data. As shown
in Table 3, the estimated results generally agreed well with the experimental data, particularly for
trapping and congestion coefficients. Errors between the estimated and experimental values for three
responses were generally less than 10%. All the above results implied that the reduced regression
equations provided sufficient accuracy for predicting the responses. Based on the analysis above, the
reduced regression models were capable of estimating and explaining the actual relationships between
twist efficiency, propagation coefficients and the various system parameters of twist factor, speed ratio
and wrap angle.

7
(a) 99 Twist efficiency
98
95
90
80
70
Percent (%) 60
50
40
30
20
10
5
2
1
-0.03 -0.02 -0.01 0.00 0.01 0.02 0.03
Residual

(b) 99 Twist trapping

98
95
90
80
70
Percent (%)

60
50
40
30
20
10
5
2
1
-0.03 -0.02 -0.01 0.00 0.01 0.02 0.03
Residual

8
 (c) 99 Twist congestion
98
95
90
80
70
Percent (%) 60
50
40
30
20
10
5
2
1
-0.02 -0.01 0.00 0.01 0.02
Residual
Figure 3 Normal probability plots of residuals for responses

Table 3 Model verifications for 9 cases

Order 1 2 3 4 5 6 7 8 9
Twist factor 2.5 2.8 3.3 3.7 4.2 2.7 3 4.3 4.1
Speed ratio 2.5 1.5 2.2 1.8 2.5 3 2 2.8 1.8
Wrap angle 30 40 50 65 50 60 45 75 85
A 0.039 0.098 0.110 0.143 0.095 0.085 0.102 0.109 0.163
Twist efficiency P 0.036 0.091 0.106 0.142 0.094 0.086 0.099 0.117 0.158
E (%) 7.69 7.14 3.64 0.70 1.05 1.18 2.94 7.34 3.07
A 0.957 0.901 0.888 0.854 0.904 0.913 0.897 0.892 0.830
Trapping coefficient P 0.963 0.908 0.892 0.854 0.904 0.912 0.900 0.879 0.836
E (%) 0.63 0.78 0.45 0.00 0.00 0.11 0.33 1.46 0.72
A 0.967 0.917 0.906 0.879 0.919 0.928 0.914 0.910 0.860
Congestion coefficient P 0.971 0.924 0.911 0.879 0.920 0.926 0.917 0.899 0.865
E (%) 0.41 0.76 0.55 0.00 0.11 0.22 0.33 1.21 0.58
A: actual value P: predicted value E: error

Effect of control variables on the responses

Equation 6 is the empirical equations for the twist efficiency (Y1), propagation coefficients of
twist trapping (Y2) and congestion (Y3) as a function of the independent variables of speed ratio (X2)
and wrap angle (X3) in coded units. It was found that the twist efficiency was increased with the
decrease of X2 and the increase of X3. Besides, the trapping and congestion coefficients were increased
with the increment of X2 and the decrement of X3. Moreover, the sign of the coefficients except
constant for Y1 was opposite to that of Y2 and Y3, while the coefficients of Y2 and Y3 had the same sign

9
and similar value.

0.131 − 0.013 X 2 + 0.030 X 3 − 0.008 X 22 − 0.010X 32

Y1 =

0.866 + 0.013 X 2 − 0.032 X 3 + 0.008 X 22 + 0.009 X 32

Y2 = (6)

0.889 + 0.011X 2 − 0.026 X 3 + 0.006 X 22 + 0.008 X 32

Y3 =
The contour plot of control parameters on twist efficiency is shown in Figure 4a, where it could
be seen intuitively that the twist efficiency was reduced as the wrap angle decreased and speed ratio
increased. The wrap angle has a more significant effect than the speed ratio. Figure 4b shows that as
the wrap angle went up, the trapping coefficient dropped sharply. Meanwhile, the trapping coefficient
was reduced as the speed ratio decreased. A similar trend was noted in the congestion coefficient in the
contour plot of Figure 4c.

10
 Figure 4 Response surface plots for (a) twist efficiency, (b) twist trapping, and (c) twist congestion

Relationships among three responses

Based on the analysis of the reduced regression equations, it was interesting to unveil whether
the three responses were related with each other. Thus, a fitting scheme was carried out in order to
check their relationships. Figure 5a displays the experimental data of the twist efficiency and trapping
coefficient. Results showed an approximately linear relationship between Y1 and Y2 with a high
correlation coefficient of 0.998 and could be well explained by the following linear regression
equation.
Y2 1.002 − 1.042Y1
= (7)

Moreover, the relationship between Y1 and Y3 are shown in Figure 5b and a linear regression
equation with a high correlation coefficient of 0.998 was indicated to explain the relationship as
below,
Y3 1.001 − 0.860Y1
= (8)

In addition, Figure 5c depicts the experimental data of propagation coefficients of twist trapping
and congestion, from which a linear relationship could be obtained with the same high correlation
coefficient. The relationship was expressed by the following linear regression formula.
Y3 0.175 + 0.825Y2
= (9)

Finally, it was concluded that the three responses have linear relationships.

11
 (a) 1.00
Experimental data
Linear fitted line

Trapping coefficient (Y2)

0.95

0.90

0.85
1.002 − 1.042Y11
Y2Y2=1.002-1.04202Y
=
2
Adj.
Adj .RR2 =0.998
= 0.998
0.80
0.00 0.05 0.10 0.15 0.20
Twist efficiency (Y1)

(b) 1.00
Experimental data
Congestion coefficient (Y3)

Linear fitted line

0.95

0.90

0.85
1.001 − 0.860Y11
YY33=1.00118-0.86027Y
=
2
Adj.
Adj.RR 2=0.998
= 0.998
0.80
0.00 0.05 0.10 0.15 0.20
Twist efficiency (Y1)

(c)1.00
Experimental data
Congestion coefficient (Y3)

Linear fitted line

0.95

0.90

0.85
= 3 0.175 + 0.825Y22
YY3=0.17479+0.82463Y
2
Adj.
AdjR.R=0.998
2
= 0.998
0.80
0.80 0.85 0.90 0.95 1.00
Trapping coefficient (Y2)
Figure 5 Linear relationships among three responses

12
Conclusions

In this paper, a systematic investigation was carried out to study yarn twisting process in a modified
ring spinning system using central composite response surface design. It was found that the speed
ratio and wrap angle are statistically significant for the twist efficiency, propagation coefficients of
twist trapping and congestion at 5% significant level, while the twist factor has little effect on the
responses. Then, the complete quadratic regression models were further simplified by eliminating
insignificant terms, which were validated by normal probability analysis and another nine randomly
selected experiments. Finally, it was important to discover that relationships among three responses
can be expressed by linear regression equations with a high correlation coefficient of 0.998. The
findings can be used to predict yarn twist in the spinning zone under given operation parameters and
facilitate further research in improving spinning technology as well as reforming machine design.

Acknowledgment
This research was funded in part through a research grant from Innovation and Technology
Commission of the Hong Kong Special Administrative Region, China (Project No: ITP/021/17TI) and
a postgraduate scholarship by the Hong Kong Polytechnic University.

References
1. Hearle JWS, Grosberg P and Backer S. Structural Mechanics of Fibers, Yarns, and Fabrics. New York:
Wiley-Interscience, 1969.
2. A. LC. Advances in yarn spinning technology. Cambridge: Woodhead Publishing Ltd, 2010.
3. Yin R and Gu HB. Numerical simulation of quasi-stationary ring spinning process linear elastic yarn.
Text Res J. 2011; 81: 22-7.
4. Yin R and Gu HB. Accurate prediction of the ring-spinning equation in zero air drag based on
homotopy perturbation method. J Text I. 2011; 102: 763-6.
5. Fraser WB. On the Theory of Ring Spinning. Philos T Roy Soc A. 1993; 342: 439-68.
6. Batra SK and Fraser WB. Engineering fundamentals of ring spinning/twisting, over-end unwinding
and two-for-one twisting in textile processes. Lancaster, Pennsylvania: Destech Publications 2015.
7. Skenderi Z and Vitez D. Compact spinning - A new chance for ring spinning. Tekstil. 2003; 52: 11-7.
8. Nikolic M, Stjepanovic Z, Lesjak F and Stritof A. Compact spinning for improved quality of
ring-spun yarns. Fibres Text East Eur. 2003; 11: 30-5.
9. Plate DEA and Lappage J. An Alternative Approach to 2-Fold Weaving Yarn .1. Control of Surface
Fibers. J Text I. 1982; 73: 99-106.
10. Emmanuel A and Plate DEA. An Alternative Approach to 2-Fold Weaving Yarn .2. The
Theoretical-Model. J Text I. 1982; 73: 107-16.
11. Prins MW, Lamb PR, Naylor GRS and Tao XM. Yarn spinning from fibre sub-assemblies with
variation of their paths of travel, relative positions or twist levels. U.S. Patent Number 6,012,277 2000.
12. Lappage J. End breaks in the spinning and weaving of weavable singles yarns - Part 2: End breaks in
weaving. Text Res J. 2005; 75: 512-7.
13. Lappage J. End breaks in the spinning and weaving of weavable singles yarns - Part 1: End breaks in
spinning. Text Res J. 2005; 75: 507-11.
14. Wronz, Iws and Csiro. Weavable singles yarn - A technological breakthrough in worsted spinning.
Wool Tech Sheep Bree. 1998; 46: 46-9.
15. Chi PC, Ki LC and Tao XM. Method of Industrially Producing Yarn at a Lower Twist Multiplier for

13
Textile Products. 2010.
16. Tao XM and Xu BG. Manufacturing Method and Apparatus for Torque-free Singles Ring Spun Yarns.
US 6,860,095 B2, 2005.
17. Tao XM, Xu BG and Wong SK. Method and Apparatus for Manufacturing A Singles Ring Yarn. US
7,096,655 B2, 2004.
18. Guo Y, Feng J, Yin R, Wang XG, van der Sluijs M and Tao XM. Investigation and evaluation on fine
Upland cotton blend yarns made by the modified ring spinning system. Text Res J. 2015; 85: 1355-66.
19. Hua T, Tao XM, Xu BG and Murrells C. Comparison of the Properties of Nu-Torque (Tm),
Convensional Ring, Compact and Vortex Yarns and Their Fabrics. 86th Textile Institute World Conference,
Vol 4, Conference Proceedings. 2008: 2497-506.
20. Yang K, Yip YK, Tao XM, et al. Developing Nu-Torque (TM) singles ring yarn to reduce spirality of
single Jersey knitted fabric. Quality Textiles for Quality Life, Vols 1-4. 2004: 515-8.
21. Xu BG, Tao XM, Wong SK, et al. Nu-Torque (TM) singles ring yarn and its production technology.
Quality Textiles for Quality Life, Vols 1-4. 2004: 494-6.
22. Wong KK, Hua T, Leung CL, et al. Development of Nu-Torque (TM) singles ring yarns for industrial
application. Quality Textiles for Quality Life, Vols 1-4. 2004: 482-4.
23. Murrells CM, Wong KK, Hua T, et al. Study of yarn snarling in Nu-Torque (TM) singles ring yarns.
Quality Textiles for Quality Life, Vols 1-4. 2004: 401-4.
24. Hua T, Cheng KPS, Tao XM, et al. A method for improvement of denim fabric appearance by using
Nu-torque((TM)) singles, ring spun yarns. Quality Textiles for Quality Life, Vols 1-4. 2004: 285-9.
25. Miao MH and Chen RZ. Yarn Twisting Dynamics. Text Res J. 1993; 63: 150-8.
26. Yin R, Tao XM and Xu BG. Variation of false twist on spinning process stability and resultant yarn
properties in a modified ring spinning frame. Text Res J. 2018; 88: 1876-92.
27. Yin R, Tao XM and Xu BG. Mathematical Modeling of Yarn Dynamics in a Generalized Twisting
System. Sci Rep-Uk. 2016; 6.
28. Feng J, Xu BG and Tao XM. Systematic investigation and optimization of fine cotton yarns produced
in a modified ring spinning system using statistical methods. Text Res J. 2013; 83: 238-48.
29. Siddh MM, Gadekar G, Soni G and Jain R. Integrating Lean Six Sigma and Supply Chain Approach
for Quality and Business Performance. 2014 2nd International Conference on Business And Information
Management (Icbim). 2014.
30. Swamy GJ, Sangamithra A and Chandrasekar V. Response surface modeling and process optimization
of aqueous extraction of natural pigments from Beta vulgaris using Box-Behnken design of experiments.
Dyes Pigments. 2014; 111: 64-74.
31. Anawa EM and Olabi AG. Using Taguchi method to optimize welding pool of dissimilar laser-welded
components. Opt Laser Technol. 2008; 40: 379-88.
32. Siyal AN, Memon SQ and Khaskheli MI. Optimization and equilibrium studies of Pb(II) removal by
Grewia Asiatica seed: a factorial design approach. Pol J Chem Technol. 2012; 14: 71-7.
33. Oliveira R, Oliveira V, Aracava KK and Rodrigues CED. Effects of the extraction conditions on the
yield and composition of rice bran oil extracted with ethanol-A response surface approach. Food Bioprod
Process. 2012; 90: 22-31.
34. Braga ARC, Gomes PA and Kalil SJ. Formulation of Culture Medium with Agroindustrial Waste for
beta-Galactosidase Production from Kluyveromyces marxianus ATCC 16045. Food Bioprocess Tech. 2012;
5: 1653-63.
35. Altemimi A, Lightfoot DA, Kinsel M and Watson DG. Employing Response Surface Methodology for
the Optimization of Ultrasound Assisted Extraction of Lutein and beta-Carotene from Spinach. Molecules.
2015; 20: 6611-25.
36. Wang AX, Li YF, Yang B, Xu BQ, Kong LX and Liu DC. Process optimization for vacuum distillation
of Sn-Sb alloy by response surface methodology. Vacuum. 2014; 109: 127-34.
37. Ghaedi M, Jaberi SYS, Hajati S, Montazerozohori M, Asfaram A and Zareh M. Modified Carbon
Paste Electrode for Pb2+ Ion Determination: Response Surface Methodology. Ieee Sens J. 2015; 15:
2974-83.
38. Samuel MS, Abigail EAM and Chidambaram R. Isotherm Modelling, Kinetic Study and Optimization
of Batch Parameters Using Response Surface Methodology for Effective Removal of Cr(VI) Using Fungal
Biomass. Plos One. 2015; 10.
39. Valencia P, Espinoza K, Ceballos A, Pinto M and Almonacid S. Novel modeling methodology for the
characterization of enzymatic hydrolysis of proteins. Process Biochem. 2015; 50: 589-97.

14
40. Madani S, Gheshlaghi R, Mahdavi MA, Sobhani M and Elkamel A. Optimization of the performance
of a double-chamber microbial fuel cell through factorial design of experiments and response surface
methodology. Fuel. 2015; 150: 434-40.
41. Savasari M, Emadi M, Bahmanyar MA and Biparva P. Optimization of Cd(II) removal from aqueous
solution by ascorbic acid-stabilized zero valent iron nanoparticles using response surface methodology. J
Ind Eng Chem. 2015; 21: 1403-9.
42. Chang BP, Akil HM, Nasir RB and Khan A. Optimization on wear performance of UHMWPE
composites using response surface methodology. Tribol Int. 2015; 88: 252-62.
43. Nagar S, Gupta VK, Kumar D, Kumar L and Kuhad RC. Production and optimization of cellulase-free,
alkali-stable xylanase by Bacillus pumilus SV-85S in submerged fermentation. J Ind Microbiol Biot. 2010;
37: 71-83.
44. Diwaniyan S, Sharma KK and Kuhad RC. Laccase from an alkalitolerant basidiomycetes Crinipellis
sp RCK-1: Production optimization by response surface methodology. J Basic Microb. 2012; 52: 397-407.

Appendix

Table A1 Analysis of variance for responses

Term DF Seq. SS Adj. MS F p
Twist efficiency
Regression 9 0.017579 0.001953 10.36 0.001
X1 1 0.000423 0.000423 2.24 0.165
X2 1 0.002481 0.002481 13.16 0.005
X3 1 0.012325 0.012325 65.35 0.000
X 1* X 1 1 0.000106 0.000106 0.56 0.471
X 2* X 2 1 0.000819 0.000819 4.34 0.064
X 3* X 3 1 0.001249 0.001249 6.62 0.028
X 1* X 2 1 0.000001 0.000001 0.01 0.940
X 1* X 3 1 0.000171 0.000171 0.91 0.363
X 2* X 3 1 0.000028 0.000028 0.15 0.707
Residual error 10 0.001886 0.000189
Lack-of-fit 5 0.001131 0.000226 1.50 0.334
Pure error 5 0.000755 0.000151
Total 19 0.019465
R2=90.31% R2(adj.)=81.59%
Trapping coefficient
Regression 9 0.019255 0.002139 11.16 0.000
X1 1 0.000465 0.000465 2.43 0.150
X2 1 0.002383 0.002383 12.44 0.005
X3 1 0.014050 0.014050 73.32 0.000
X 1* X 1 1 0.000122 0.000122 0.64 0.444
X 2* X 2 1 0.000934 0.000934 4.88 0.052
X 3* X 3 1 0.001106 0.001106 5.77 0.037
X 1* X 2 1 0.000001 0.000001 0.01 0.940
X 1* X 3 1 0.000190 0.000190 0.99 0.343
X 2* X 3 1 0.000021 0.000021 0.11 0.747
Residual error 10 0.001916 0.000192
Lack-of-fit 5 0.001152 0.000230 1.51 0.332
Pure error 5 0.000765 0.000153
Total 19 0.021171
R2=90.95% R2(adj.)=82.80%
Congestion coefficient
Regression 9 0.013140 0.001460 11.33 0.000

15
 X1 1 0.000234 0.000234 1.82 0.207
X2 1 0.001745 0.001745 13.54 0.004
X3 1 0.009509 0.009509 73.79 0.000
X 1* X 1 1 0.000077 0.000077 0.60 0.458
X 2* X 2 1 0.000519 0.000519 4.02 0.073
X 3* X 3 1 0.000950 0.000950 7.37 0.022
X 1* X 2 1 0.000000 0.000000 0.00 1.000
X 1* X 3 1 0.000113 0.000113 0.87 0.372
X 2* X 3 1 0.000008 0.000008 0.06 0.808
Residual error 10 0.001289 0.000129
Lack-of-fit 5 0.000791 0.000158 1.59 0.312
Pure error 5 0.000498 0.000100
Total 19 0.014429
R2=91.07% R2(adj.)=83.03%
Seq. SS: sequential sum of squares, Adj. MS: adjusted mean squares

Table A2 Analysis of the reduced regression model

Term DF Seq. SS Adj. MS F p
Twist efficiency
Regression 4 0.016850 0.004212 24.16 0.000
X2 1 0.002481 0.002481 14.23 0.002
X3 1 0.012325 0.012325 70.69 0.000
X2* X2 1 0.000888 0.000888 5.09 0.039
X3* X3 1 0.001335 0.001335 7.66 0.014
Residual error 15 0.002615 0.000174
Lack-of-fit 4 0.000994 0.000248 1.69 0.223
Pure error 11 0.001621 0.000147
Total 19 0.019465
R2=86.56% R2(adj.)=82.98%
Trapping coefficient
Regression 4 0.018455 0.004614 25.49 0.000
X2 1 0.002383 0.002383 13.16 0.002
X3 1 0.014050 0.014050 77.61 0.000
X2* X2 1 0.001012 0.001012 5.59 0.032
X3* X3 1 0.001191 0.001191 6.58 0.022
Residual error 15 0.002716 0.000181
Lack-of-fit 4 0.000997 0.000249 1.60 0.244
Pure error 11 0.001719 0.000156
Total 19 0.021171
R2=87.17% R2(adj.)=83.75%
Congestion coefficient
Regression 4 0.012708 0.003177 27.70 0.000
X2 1 0.001745 0.001745 15.21 0.001
X3 1 0.009509 0.009509 82.90 0.000
X2* X2 1 0.000565 0.000565 4.92 0.042
X3* X3 1 0.001015 0.001015 8.85 0.009
Residual error 15 0.001720 0.000115
Lack-of-fit 4 0.000690 0.000172 1.84 0.192
Pure error 11 0.001031 0.000094
Total 19 0.014429
R2=88.08% R2(adj.)=84.90%

16