You are on page 1of 13

This article was downloaded by: [Universitat Politcnica de Valncia]

On: 26 October 2014, At: 00:19


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Polymer-Plastics Technology and Engineering


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/lpte20

Application of BoxWilson Optimization Technique to


the Partially Oriented Yarn Properties
a

Glay zkan , Gray rkmez & Gksel zkan

Department of Chemical Engineering , Ankara University , Tandogan, Ankara, Turkey

Nergis A. . , Organize Sanayi Blgesi , Bursa, Turkey

Chemical Engineering Department , Gazi University , Ankara, Turkey


Published online: 14 Feb 2007.

To cite this article: Glay zkan , Gray rkmez & Gksel zkan (2003) Application of BoxWilson Optimization Technique
to the Partially Oriented Yarn Properties, Polymer-Plastics Technology and Engineering, 42:3, 459-470, DOI: 10.1081/
PPT-120017964
To link to this article: http://dx.doi.org/10.1081/PPT-120017964

PLEASE SCROLL DOWN FOR ARTICLE


Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content) contained
in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no
representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the
Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and
are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and
should be independently verified with primary sources of information. Taylor and Francis shall not be liable for
any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever
or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of
the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any
form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at http://
www.tandfonline.com/page/terms-and-conditions

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

POLYMER PLASTICS TECHNOLOGY AND ENGINEERING


Vol. 42, No. 3, pp. 459470, 2003

Application of Box Wilson Optimization


Technique to the Partially Oriented
Yarn Properties
zkan,1,* Guray U
zkan3
rkmez,2 and Goksel O
Gulay O
1

Department of Chemical Engineering, Ankara University, Tandogan,


Ankara, Turkey
2
Nergis A. S., Organize Sanayi Bolgesi, Bursa, Turkey
3
Chemical Engineering Department, Gazi University, Ankara, Turkey

ABSTRACT
A static model for partially oriented yarn (POY) production was
examined in a commercial POY spinning plant. A two-level factorial,
industrial experimental design was used to identify a statistical model.
Three independent variables effecting the elongation at break (a POY
property of interest) were selected. Namely, yarn winding speed,
spinneret flow rate, and spinneret hole diameter. Elongation at break was
chosen as the dependent variable. A first-order statistical model was
considered to show the dependence of the elongation at break on the
operating parameters: 145%; for elongation at break, the optimum values
of yarn winding speed, spinneret flow rate, and spinneret hole diameter

zkan, Department of Chemical Engineering, Ankara


*Correspondence: Gulay O
University, Tandogan 06100, Ankara, Turkey; E-mail: gozkan@eng.ankara.edu.tr.
459
DOI: 10.1081/PPT-120017964
Copyright q 2003 by Marcel Dekker, Inc.

0360-2559 (Print); 1525-6111 (Online)


www.dekker.com

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

zkan, U
zkan
rkmez, and O
O

460

were found to be as 2938 m/min, 49.74 g/min, and 205 mm, respectively.
The statistical model and Box Wilson steepest ascent method were used.

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

Key Words: Partially oriented yarn; Effective experimentation;


Optimization; Elongation at break.

INTRODUCTION
A synthetic polymer needs to have fiber characteristics for its use as a textile
material, suchas high-softening and low-melting points thatallow ironing,enough
load at break, elongation at break, and reasonable boiling shrinkage. Polyethylene
terephthalate (PET) is an important synthetic fiber that is classified as a polyester.
The yarns produced from POY (partially oriented yarn), a semi-finished product of
the first production stage of polyester yarn, have a wide range of use. For the
polymer to gain these characteristics in the production stage, its degradation must
be under certain values and enough orientation has to be reached by the time it
comes to the spinning beam.
Fiber spinning is a complex manufacturing process in which the final
properties of the spun fiber depend on the polymer and the process operation. The
latter includes the details of the heat transfer, the aerodynamics, and the stress
development, all of which are ultimately related to a variety of factors that include
both the spinning speed and the polymer properties at the spinneret. The
uniformity and birefringence of the spun fiber are the process variables that are
probably of the greatest interest; both are directly related to product quality and
generally correlate with final product properties such as elongation.[1]
Small changes in polymer properties or in operating variables are
sometimes known to cause large changes in fiber properties, thus an
understanding of the interrelationships is essential. This understanding is
becoming increasingly important as fiber properties move to polymer
modifications and to new process conditions that may provide increased
productivity or improved properties at the expense of process sensitivity.[2]
By considering these effects on the POY productivity and quality, we aimed
to find the optimum values of important parameters for 100 denier and 96
filaments in a commercial POY spinning plant with a capacity of 70 ton/day.
To determine the optimal operating conditions, we used a statistical
model. The experimental data used in the preparation of this model were
obtained according to the use of a two-level, factorial experimental design.
This is called a statistical experimental design technique that has been used
over a wide range of industrial process, such as ternary composites based

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

Box Wilson Optimization Technique and POY Properties

461

on isotactic polypropylene, thermoplastic elastomer (ethylene-octhene


copolymer and polyethylene terephtalate textile fibers have been processed
and their properties analyzed). The effect of matrix composition and fiber
content on the final properties of the composite were investigated by means of
an experimental design based on a Doehlert uniform net.[3]
Two-level, factorial experimental design is an active experimentation. It
has been used in mathematical modeling of the industrial process and
optimization. This experimental modeling technique allows the determination
of the regression equation in a very short time. The parameter (factor), such as
the winding speed, spinneret hole diameter, and spinneret throughput, are
changed according to predetermined levels. When S and K represent the level
of one parameter and number of parameters, respectively, then the number of
experiments for a linear model can be shown as N SK
The main difference between active and passive experimentation is the
change of parameters with time in the latter. In the traditional or passive,
method all the parameters are kept constant, changing only one parameter.
This requires more experiments. This means that considerably less
experiments are needed with the two-level, factorial experimentation design
compared to the passive method. For this reason, it is quite economical.
The parameters of the statistical model were found and then BoxWilsons
steepest ascent method was used to find optimal operating conditions.[4 6]

MATERIALS AND METHODS


To calculate optimal operating conditions on the system during POY
production, the yarn samples taken from the winding system were sent to the
laboratory and analyzed for the properties of dtex (10,000 m weight of fiber per
g), oil pick up, uster (POY change of weight in the unit of POY fiber), load at
break, elongation at break and boiling shrinkage, to decide about its quality. For
yarn, the polymer melt as a thread should be fed to distributors called spinnerets
with a flow rate. The spinnerets, which process the polymer melt as a thread,
consist of holes with small diameter. The melt solidifies as it passes through the
spinneret channel and is stretched as a thread in the fiber elongation channel under
it. Figure 1 depicts the route of the polymer melt from its outlet at the spinneret
and yarned as a thread. The melt not only solidifies on this route but it is also
subjected to a certain amount of orientation
The investigations were carried out by observing different plant conditions
for the production and the product quality. In the production of (100 denier 96
filaments) POY, the optimal values of winding speed, spinneret flow rate, and
spinneret hole diameter that effects the characteristics of the yarn, and of the yarn

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

462

zkan, U
zkan
rkmez, and O
O

Figure 1. The yarn production system, which is similar in the direct spinning system
and the yarn production system with extruder.

characteristic, the values of elongation at break, were determined by using the


experimental design and the BoxWilson optimization method.

STATISTICAL AND BOX WILSON


OPTIMIZATION TECHNIQUE
Operating parameters relating to the fiber properties were investigated
and the effect of these parameters on POY productivity was examined by
static model identification. It was clearly observed that these three
independent variables, namely spinneret hole diameter, spinneret flow rate,
and spinning speed, had an effect on the elongation at break.

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Box Wilson Optimization Technique and POY Properties

463

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

To obtain the available elongation at break, optimal values of these


variables were evaluated by utilizing the Box Wilson optimization method.
In this method, the form of the relationship between the independent variables
is unknown. Therefore, a suitable function, a relationship between the
dependent and independent variables, can be obtained by using a statistical
model. Its form, which is studied in this work, is shown below:
Y bo b1 X1 b2 X2 b3 X3 b4 X1 X2 b5 X1 X3 b6 X2 X3
b7 X 1 X 2 X 3

This first-order regression model, including second- and third- order


interaction terms was considered. In this equation, Xi i 1. . .n are the coded
values of the spinneret hole diameter, spinneret flow rate, and spinning speed,
respectively, and bi i 07 is the model coefficient. Y is the measured of
values elongation at break. In two-level, factorial experimental design there
are n variables (factors). These factors are controlled at two different levels,
namely minimum and maximum levels. In these types of designs, one works in
a dimensionless coordinate system using following definitions.
In a dimensionless coordinate system, the maximum level is +1, minimum
level is 21, and the central point is zero. In such a design, the number of
experiments to be carried out is 2n (in the preparation of the design matrix the
coded values of the parameters are used; the reason for this is to render the
parameters dimensionless, thus, making the calculations dimensionless).
In this work, the possible maximum and minimum values of the selected
operating variables (parameters) for fiber properties were studied in a commercial
POY spinning plant. The levels of these parameters are given in Table 1. For this
reasons, 23 8 experiments were done to identify the regression model.
The design matrix included coded values of the independent and un coded
dependent variables. Table 2 shows the interrelation between the coded
independent variables. The column (see Table 2) under X0 contains the
dummy variable, which is equal to 1, so that the model includes an intercept.
All the weights will be unity. The interaction terms, including second- and

Table 1.

Maximum (+1)
Minimum (-1)

Levels of independent variables.

Spinneret hole
diameter U1 (mm)

Spinneret flow
rate U3(g/min)

Winding speed
U2 (m/min)

250
170

52
47

3100
2850

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016

X0

X1

X2

X3

X1X2

X1X3

X2X3

X1X2X3

Y
1
2
3
4
5
6
7
8

1
1
1
1
1
1
1
1

1
1
1
1
21
21
21
21

1
1
21
21
1
1
21
21

1
21
1
21
1
21
1
21

1
1
21
21
21
21
1
1

1
21
1
21
21
1
21
1

1
21
21
1
1
21
21
1

1
21
21
1
21
1
1
21

136.6
146.2
133.8
144.3
137.8
153.8
135.1
149.1

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

Experiment
number

zkan, U
zkan
rkmez, and O
O

2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

464

Optimal design matrix.


Table 2.

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Box Wilson Optimization Technique and POY Properties

465

third- order, were obtained by multiplying the appropriate column. The


relationship between coded and real values is shown below:

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

Xi

U i 2 U iav
DU i

i 1; 2; . . .. . .. . .; n

U iav

U max:
U min:
i
i
2

i 1; 2; . . .. . .. . .; n

DU i

U max:
2 U min:
i
i
2

i 1; 2; . . .. . .. . .; n

The model coefficients, b0 2 b7, by using experimental data, were calculated


by the following equations by applying the least-squares estimation technique:
3
2
2 3
2 3
y1
X 01 X 11 X 21 * * X n1
b0
7
6
6 7
6 7
6 X 02 X 12 X 22 * * X n2 7
6 b1 7
6 y2 7
7
6
6 7
6 7
7
6
6 7
6 7
7 *6 * 7 6 * 7
6*
5
7
6
6 7
6 7
7
6
6 7
6 7
*
*
*
7
6
6 7
6 7
5
4
4 5
4 5
yn
bn
X 0n X 1n X 2n * * X nn
n* n
n* 1
n* 1
X* B Y

B X T * X21 X T * Y

The design matrix has orthogonal properties, the coefficient matrix (X*X)
becomes diagonal, and its diagonal elements are equal to the number of
experiments, N. For this reason, X* X21 N1 and the values of the constants
in the regression equation were calculated by the following equation:
N
X

bi

N
X

Xij Y i

j1
N
X

X2ij

Xij Y i

j1

; i 0; 1; . . .N

j1

In this work, the model coefficients, b0 b7, by using experimental data, were
calculated using MATLAB code. And then, Students t and Fishers F tests
were applied using the following equations. The error mean square was

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

zkan, U
zkan
rkmez, and O
O

466

calculated from the equations n0 2:


n0
X

S2e

y0i 2 y 0

i1

n0 2 1

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

The static model constants were found as


j bi j
i 0; 1; . . . 7
Sbi
Se
Sbi p i 0; 1; . . . 7
N

ti

10
11

Fishers F test was applied to see the fitness of the new estimated regression
equation, which was obtained by the removal of the insignificant coefficient
from the experimental data. According to this, the F value was determined by
using the following equation:
F

Sr 2
Se 2

12

In this equation, Sr2 is the residual mean square, and it was determined as
N
X

S2r

yi 2 y^ i 2

i1

13

N2l

Then, parameters of the process were determined by the Box Wilson


optimization technique, as follows. Starting from the central points of the
parameters [this central point is the origin (O)]. The purpose is to move from
this point, where the coordinates are (0, 0,. . ..0), to the maximum point (P),
where the coordinates are (X1,X2,X3. . .XN). The point where the response was
the maximum was reached using the estimated derivatives of the modeling
parameters. This was achieved by applying a 2 0.50 unit of change to the hole
diameter and the new levels of the other independent variables were
determined the use of the program given in Table 3.

RESULTS
By applying the design matrix given in Table 2, the productivity values
were used to evaluate the values of the constants in a linear regression model
as illustrated previously. The identified statistical model is given by Eq. (14)

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Box Wilson Optimization Technique and POY Properties

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

Table 3.

MATLAB m files.

467

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

zkan, U
zkan
rkmez, and O
O

468

below:
Y 141:7125 2 1:4875X1 1:1375X2 2 5:8875X3
0:0375X1 X2 0:8625X1 X3 0:2375X2 X3
14

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

2 0:0125X1 X2 X3

There were replicate observations performed at the center points of the


three parameters. The error mean square was found from Eq. (9) as 1.2025.
The tabulated value of the Students t distribution was taken as t0:6 1 0:325
for a significance level of P :6 and f 1 degrees of freedom. Since the t4, t7,
values (Eq. 10) were less then the tabulated t value, constants of b4, b7 were
omitted from the regression equations. The next step was the application of
Fishers F test to see the fitness of the new estimated regression equation. The
residual mean square calculated from Eq. (13) was 0.00625. For a significance
level P 0:6 and v1 1 and v2 2 degrees of freedom, an F value of 2.5714
was found from the related table. Since the calculated F value (0.069) was less
than the tabulated F value, it was decided that the estimated regression
equation fits the experimental data.
Consequently, the estimated regression equations are
Coded : Y 141:7125 2 1:4875X1 1:1375X2 2 5:8875X3
15

0:8625X1 X3 0:2375X2 X3
Uncoded : E 487:65 2 0:5487* U1 2 1:82* U2 2 0:121* U3

16

0:00017* U1 U3 0:000768* U2 U3

This equation was validated against new data, which is given in Table 4, and it
was seen that the model was suitable for this process.
After the identification of the statistical model for the process, Box
Wilsons steepest ascent optimization method was applied theoretically on the
Table 4.
Elongation at
break E (%)
(theoretical)
141.8

New data.

Elongation at
break E (%)
(experimental)

Spinneret hole
diameter: U1
(mm)

Winding
speed U3
(m/min)

Spinneret
flow rate U2
(g/min)

142.3

170

3000

49

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

Box Wilson Optimization Technique and POY Properties

469

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

basis of the statistical model using the model constant by applying 2 0.50
change to the growth medium spinneret hole diameter and 145% for
elongation at break, the optimum values of yarn winding speed, spinneret flow
rate, and spinneret hole diameter were then found as 2938 m/min, 49.74 g/min,
and 205 mm, respectively.

CONCLUSION
In the present study, an orthogonal, two-level, factorial experimental
design technique was found suitable for the first-order regression equation.
Model constants were calculated using MATLAB coded and Students t and
Fishers F test were applied to see whether the identified statistical equation fit
the experimental data. Then, with the help of the identified model coefficient
for 145% elongation at break, the optimal operating parameters of the process
were determined by the Box Wilson steepest ascent optimization technique
as follows: spinneret flow rate 49.74 g/min, spinneret hole diameter 205 mm,
yarn winding speed 2938 m/min.

NOMENCLATURE
bi
D
E
l
Q
S
Sbi
Sr 2
Se 2
Ui
Uiav
Ui*
DUi
Xi
Xi*
X1
X2
X3
Yi

Statistical model coefficients


Spinneret hole diameter
Elongation at break
Number of the coefficients in the regression equation
Spinneret throughput
Winding speed
Standard error of the coefficient
Residual mean square
Error mean square
Real value of the parameters
Average values of the parameters
Average value of the independent variables at center points
Incremental value of the parameters
Independent variables matrix
Average value of the independent variables at center points
Coded value of the spinneret hole diameter
Coded value of the spinneret flow rate
Coded value of the winding speed
ith experimental value of the elongation at break

MARCEL DEKKER, INC. 270 MADISON AVENUE NEW YORK, NY 10016


2003 Marcel Dekker, Inc. All rights reserved. This material may not be used or reproduced in any form without the express written permission of Marcel Dekker, Inc.

zkan, U
zkan
rkmez, and O
O

470

v 1, v 2
Y

Degrees of freedom
Estimated value of the elongation at break

Downloaded by [Universitat Politcnica de Valncia] at 00:19 26 October 2014

ACKNOWLEDGMENT
.B.A.P. 2002 07 45
Ankara University Research fund through project A.U
009 is gratefully acknowledged.

REFERENCES
1.
2.
3.

4.
5.
6.

Denn, M.M. Spinning: where are we and where are we going? Fiber Prod.
1983, August, 20 26.
Davis, G.W.; Everage, A.E.; Talbot, J.R. Polyester fibers: high speed melt
spinning. Fiber Prod. 1984, 22 28, 22 27.
Lopez-Manchado, M.A.; Arroyo, M. Effect of the incorporation of pet
fibers on the properties of thermoplastic elastomer based on PP/elastomer
blends. Polymer 2001, 42 (15), 6557 6563.
Box, G.P.E.; Wilson, K.B. J.R. Stat. Soc. Ser. B. 1951, 13, 1 45.
Montgomery, D.C. Design and Analysis of Experiments, 3rd Ed.; John
Wiley and Sons: New York, 1991.
Alpbaz, M.; Burasli, N.; Ertunc, S.; Akay, B. Application of a statistical
technique to the production of,saccharomyces cerevisiae (bakers yeast).
Biotechnol. Appl. Biochem. 1997, 26, 91 96.