Advances in Production Engineering & Management 2 (2007) 4, 185-193

ISSN 1854-6250 Professional paper

185
AN EXPERIMENTAL WEB TENSION CONTROL SYSTEM:
SYSTEM SET-UP

Giannoccaro, N.I.
*
; Oishi, K.
**
& Sakamoto, T.
**

*
Department of Innovation Engineering, University of Salento,
Via per Monteroni, 73100, Lecce, Italy
**
Department of Control Engineering, Kyushu Institute of Technology,
Tobata, Kitakyushu 804-8550,J apan
E-Mail: [email protected], [email protected]


Abstract:
This paper describes the set-up, the technical choices and the preliminary model validation
of a new experimental web tension control system composed by four sections and realized
by the authors. Interesting considerations about the accuracy of the sensors used and about
the applicability of linear models for designing the controller of the system are already shown.
The systems handling web material are very common in the industry; but there exist
many sources of disturbance e.g. roller non circularity, roller change, web sliding, change of
web-elasticity; due to the strong coupling between web velocity and web tension, these
disturbances, introduced by elastic web, are transmitted to the web tension that can result in
a web break or fold. For this reason it is very important to realize a laboratory system that
can simulate the real behavior of a web transport system for applying the theoretical studies
about web handling systems behavior and control.

Key Words: Winding Systems, Web, Tension Control

1. INTRODUCTION

The system handling web material such as textile, paper, plastic, polymer and metal are very
common in the industry. Web is usually transported by many drive rollers and idle rollers and
it is processed in several consecutive sections such as coating, laminating and printing
treatments. Web processing operations contain the functions of both transport speed and
tension controls. Since the tension control covers wide range of applications, keen interest
has been observed in this area of research over a long time in the industry. The process is
composed of many drive rolls connected to motors, guide rolls, and dancer rolls. Then the
system is regarded as a large drive system mechanically connected together with the web,
which is expected to operate at the demanded variable speed and tension.
The modeling and the control of web handling systems have been studied for several
years; in order to design a better control system or to identify the plant parameters
experimentally, correct modeling is necessary. On the web dynamics itself, as already
discussed by one of the authors [1],[2], lumped parameters expressions may be used to
designate a web section between two adjacent drive rolls, but there is the necessity of
incorporating the property of visco-elasticity to the web; at this proposal mainly two models
[3] are used (Maxwell model and Voigt model). Some experimental system results were
recently carried out [4]; but the system designed and developed by the authors has
completely different characteristics of complexity
An investigation on the applicability and validation of Maxwell model and Voigt model is
already shown in this paper. The validation of the models is attempted by tuning a minimum
set of parameters.


Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


186















Figure 1: Schematic diagram of the web tension system.

2. WEB TENSION SYSTEM

The scheme of the system that the authors aimed to realize is shown in a simplified manner
in Fig. 1 and is a complete set for a web-processing model. The system under investigation
is a 4 motors set-up composed by an unwinder section followed by guide rollers that support
a tension sensor, a lead section (speed controlled) with S-wrap, a draw roll section followed
by guide rollers that support a tension sensor and, finally the winder section.
The experimental system was realized towards the Department of Control Engineering
of Kyushu Institute of Technology and is shown in the photo in Figs 2 and 3.
Is possible to note that all the system is mounted on a mechanical frame in order to fix
each roller to the skeleton using appropriate pillows; moreover the four motor shafts are
rigidly connected to the corresponding rollers by means of universal joints (photo in Fig. 3) in
order to avoid slippage of indirect connections (like belts of chain) and are linked to the
skeleton using acrylic bolt in order to reduce the vibration (further photos and details in the
full paper).
An accurate alignment of the shaft of each motor with corresponding roller has been
necessary in the preliminary phase of setting of the system.




Figure 2: Experimental system.




Unwinder motor
Guide roller
Speed motor control
Tension motor
control
Winder
Tension sensor
Tension sensor
motor
Unwinder section Draw-roll section Lead section
Winder section
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


187



Figure 3: Experimental system: a detail.

3. THE THEORETICAL MODEL

The model developed in this work for the tuning of control system will follow the general lines
expressed in [1] and is based on the following laws;
- the law of conservation of mass for the web section for evaluating the relation between
the speeds (named v
a
and v
b
) of two adjacent rolls and the strain in the web (Equation 1).
For the web of the ordinary two drive rolls, the length L of the span is considered constant,
because the strains in general have fairly small values compared to unity;

dt t v t v
L
t
a b
] ) ( ) ( [
1
) (

= (1)

- assuming that the web does not completely slide on the roll, the web velocity is
considered equal to the roll linear velocity. The velocity v
k
of the k
th
roll can be obtained
through a torque balance in function of the tension forces F
k+1
and F
k
applied to the roll from
the web, (Equation 2).
k
is the rotational speed of roll k, J
k
is the roll inertia, r
k
is the roll
radius, U
k
is the motor torque applied to that roll;


k k k k
k k
U F F r
dt
J d
+ =

+
) (
) (
1

(2)

- for taking into account the linear viscoelasticity of web-material expressed as
combination of linear springs and dashpots, two models were considered and compared in
this work; the Maxwell model and the Voigt model [1,2,3]. The relation between the tension
force F applied to the web and the strain in the Laplace domain is expressed in Equations 3
(Voigt model) and 4 (Maxwell model).


) ( )
1
( ) ( s s
s T
s T
A s F
v
v

+
=
(3)



) (
1
) ( s s
s T
A
s F
m

+

=
(4)

where A is cross sectional area of the web, is the viscosity modulus, T
m
=T
v
=/E, E is
the elastic modulus; the Equations (3) and (4) can be expressed in an unified way like in
Equation (5) adapting P(s) to the equations (3) and (4) [1].
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


188

) ( ) ( ) ( s s s P s F = (5)

Additional assumptions are taken into account; constant value of the cross-sectional area
A of the web for all the system, constant value of the radius of the unwinder and winder
section and neglect able effects of guide rolls on the web tension.
Therefore, the model of the system is shown in a block diagram in Fig. 9, where J
uw
, J
l
, J
dr

and J
w
are the total moment of inertia of each roll referred to the motor shaft, L
uw
, L
l
and L
dr

are the web lengths of unwinder section, lead section and draw roll section, r
uw
, r
l
, r
dr
and r
w
,
are the radius of driven rollers in unwinder section, lead section, draw roll section and winder
section respectively, u
1
, u
2
, u
3
and u
4
are the manipulated variable corresponding to the torque
control signals (u
u
, u
l
, u
dr
, u
w
) of the four servomotors used, and y
1
, y
2
, y
3
and y
4
are the
controlled variable corresponding to the unwinder tension (F
uw
), lead speed (v
l
), draw-roll
section tension (F
dr
) and winder speed (v
w
).



Figure 4: Block model of the system.


4. PRELIMINARY EXPERIMENTAL IDENTIFICATION OF THE THEORETICAL MODEL

The experimental system was initially experimentally identified using acquisition cards and
sensors as schematically shown in Fig. 10. The inputs to the system are the torque control
signals (u
u
, u
l
, u
dr
, u
w
) of the four servomotors used; the servomotors are the same type but
have different ratings. The servomotors of the unwinder and winder section (model Sgmas-
08ACA21 Yaskawa 750 W maximum power), powered [5] by single phase 200 VAC filtered
by analogical filter (SCHAFFNER mod FN2070-16-07); the servomotors of the lead section
and draw-roll section (model Sgmas-01ACA21 Yaskawa 100 W maximum power) powered
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


189
by the same type of power supply but with different analogical filters (SCHAFFNER mod.
FN2070-6-07). The 4 servomotors are set in torque control mode, and the inner control
torque loop is considered immediate compared with the system dynamic in such a way that
the torque values are effectively inputs for the system. At this proposal several regulations
were made in Servopack set-up in order to optimise the inner control loop. The outputs of the
systems are the tension forces measured by tension sensors after the unwinder section and
the draw-roll section (termed F
uw
and F
dr
respectively, defines as average value of the couple
of tension sensors for each section), and the speed measured by encoder mounted on the
servomotors for the lead section and the winder section (termed v
l
and v
w
). In details the
tension sensors (model MG010 Nireco) used are placed at each side of the web after the
unwinder section and the lead section in such a way to have a double value of tension for
each section one referred to the right side and one referred to the left side of the web. The
guide roller positioned before and after the tension sensors are necessary for having a
definite slope of the web before and after tension sensors [6]. A consistent phase of
calibration of the tension sensor was already necessary for studying the repeatability of the
measures.
The inputs signals (Fig. 10) are sent to the motors by using a 4 channels D/A board
(model PCI 3343 Interface), the tension signals feed a 4 channels A/D board (model PCI
3180 Interface), the 4 motor encoder signals (including the speed signals of unwinder section
and lead section) feed a digital counter (model PCI-632206 TK Interface). Several tests were
carried out for studying the accuracy of the encoder feedback considering each motor
independently from the system.
In particular the possibility of driving the motors in "speed control mode" [5] was used to
estimate the behaviour of the motors encoder feedback.
In Fig. 5 is shown the calibration curve for the feedback encoder related to winder motor
with a sampling acquisition time of 100 ms and considering the average speed on several
samples; the behaviour shows for all the encoder considered, a good linearity.
The noise added to the encoder feedback in the modality "speed control mode" was
already analysed (e.g. Fig. 6). In particular in Fig. 6 is shown the digital encoder feedback
when a speed of 1900 RPM is requested to the winder motor; the average value is 8746
pulses for each sampling time (100 ms), but a certain oscillation is evident; statistically
defined by a standard deviation of 29 (0.3 % of average value).

Winder Motor
0
500
1000
1500
2000
2500
0 200 400 600 800
speed [ r ad/s]
N
u
m
b
e
r

o
f

p
u
l
s
e
s


Figure 5: Calibration curve for encoder of winder motor.

Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


190
Encoder feedback for wi nder motor
8600
8650
8700
8750
8800
8850
1 5 9 13 17 21 25 29 33 37 41 45 49
Sampl es
S
p
e
e
d

[
P
u
l
s
e
s
/
S
a
m
p
l
i
n
g

T
i
m
e
]

Figure 6: Analysis of the noise added to the encoder.

Similar behaviour was observed for all the speed values used in the calibration curve; in
Fig. 7 is shown the percentage standard deviation/average value of all the speed used for
encoder calibration for motor 4 in "speed control mode".

Wi nder motor
0
0.005
0.01
0.015
0.02
0.025
0 500 1000 1500 2000
speed [rad/sec]
S
t
.

d
e
v
.
/
A
v
e
r
a
g
e


%

Figure 7: Analysis of the percentage.

The analysis carried out in Figs 6-7 was repeated decreasing the sampling time of the
acquisition; but worst results were obtained with a sampling time of 10 ms about the
oscillation of digital encoder feedback in a situation of "speed control mode" [5]. For this
reason, the sampling time of the experimental tests for the estimations of other parameters of
the system model was fixed at 100 ms.
Some digital filter were furthermore tested to decrease the noise of the digital encoder
feedback; good results were obtained with moving average filters; in Fig. 8, e.g., is shown the
behaviour of the encoder feedback of Fig. 6 with a moving average digital filter applied on 5
samples.
Encoder feedback for wi nder motor
8600
8650
8700
8750
8800
8850
1 5 9 13 17 21 25 29 33 37 41 45 49
Sampl es
S
p
e
e
d

[
P
u
l
s
e
s
/
S
a
m
p
l
i
n
g

T
i
m
e
]

Figure 8: Analysis of the noise added to the encoder feedback with digital filter.
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


191
Moreover some tuning parameters are necessary for tuning the models. The parameters
of the experimental systems are shown in Table I.

Table I: Data assumed, measured and identified for modelling the system.

Name Symbol Value
Total length between winder and unwinder L [m] 3.15
Unwinder section length L
uw
[m] 1.2
Lead section length L
l
[m] 1.02
Draw-roll section length L
dr
[m] 0.93
Unwinder roller radius r
uw
[m] 0.025
Lead section roller radius r
l
[m] 0.01
Draw-roll section roller radius r
dr
[m] 0.01
Winder roller radius r
w
[m] 0.025
Web thickness T
h
[mm] 0.04
Young modulus of the web E[N/m
2
] 109
Web width W [m] 0.35
Moment of inertia for unwinder J
uw
[kgm
2
] 0.006
Moment of inertia for lead section roller J
l
[kgm
2
] 0.0008
Moment of inertia for draw-roll section roller J
dr
[kgm
2
] 0.001
Moment of inertia for winder J
w
[kgm
2
] 0.006
Viscosity coefficient (Voigt model) [Ns/m
3
] 10
9

Viscosity coefficient (Maxwell model) [Ns/m
3
] 10
9


In particular, the moments of inertia, were experimentally carried out (using the
experimental set-up previously shown) driving each motor with constant values of torque and
estimating the proportionality factor with the angular acceleration (read by encoder feedback)
of the corresponding roller. In such a way the inertia of each motor was either included in the
inertia of the corresponding roller. An example of this experimental identification for the
unwinder motor is shown in Fig. 9.

Inertia estimation for unwinder motor
0.00
20.00
40.00
60.00
80.00
100.00
120.00
140.00
160.00
180.00
0 0.5 1 1.5 2 2.5 3
Time [s]
S
p
e
e
d

[
r
a
d
/
s
]


Figure 9: Experimental inertial estimation for winder motor.

Moreover the viscosity coefficient was estimated minimizing the difference between
the experimental behaviour (outputs y
1
, y
2
, y
3
, y
4
) of the system respect to a known
sequence of inputs (u
1
, u
2
, u
3
, u
4
) with the corresponding behaviour of the Voigt and
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


192
Maxwell models. One of the result of the comparison between experimental and theoretical
behaviour with the estimated viscosity coefficients is shown in Figs 11 (Voigt model) and
12 (Maxwell model) considering a constant input u
4
of 2.34 Nm and evaluating the draw-roll
section force F
dr
(output y
3
). Voigt model and Maxwell models have often in the past used for
describing the properties of visco-elasticity of the web.
This is only a first preliminary phase in the modelling the system; no friction torques were
considered in this treatment, so not a good overlapping between theoretical and
experimental data is expected by the authors for long time tests. This choice is justified from
the thing that, in the preliminary phase, several control simulations are necessary to
understand if the experimental realised system may be controlled in speed and tension.




Figure 10: Description of the experimental system.




Figure 11: Comparison using Voigt model.
Giannoccaro, Oishi, Sakamoto: An Experimental Web Tension Control System: System Set-Up


193


Figure 12: Comparison using Maxwell model.

5. CONCLUSION

The set-up of a new experimental web tension control system has been shown in this paper.
The experimental system is characterized by four different sections, high number of rollers
positioned at different levels, long way of the web (polypropylene as example) in the different
sections. A mechanical support was conveniently designed to mount all the elements of the
experimental system.
The main scope of this paper is to show the difficulties and the design choices for
realising an experimental complex web tension control system including the commercial
elements used for building such a system. A detailed calibration of the
The preliminary model shown in this paper is the starting point to try, from now on, to
design a control algorithm in such a way to control simultaneously the web tension (lead
section and winder section) and the web speed (unwinder section and draw-roll section).

6. ACKNOWLEDGMENT

The authors wish to thank the "Matsumae International Foundation" for financial support
permitting to the corresponding author to work in the laboratory of Kyushu Institute of
Technology as visiting researcher.

REFERENCES

[1] Sakamoto, T. (1998). PI control of Web tension control system based on overlapping
decomposition, Proceedings of the IEEE Nordic Workshop on Power and Industrial Electronics
[2] Sakamoto, T (1999). Decentralized controller design of web tension control system in terms of
interactions, Industrial Electronics, 1466-1471.
[3] Akatsuka, M. (1991). Models of Longitudinal Web behaviour incorporating Visco-Elasticity,
Mitshubishi Juko Gihos, vol. 28, 4
[4] Kim, J . (2006). Development of hardware simulator and controller for web, Trans. Of ASME, vol.
128, 378-381.
[5] Yaskawa, (2002).User"s manual, Manual No. SIEP S800000 00F.
[6] Nireco, (2002). Instruction manual, Mg tension sensor. QJ 3734.1-E.