Proceedings of the 2016 IEEE

International Conference on Robotics and Biomimetics

Qingdao, China, December 3-7, 2016

Constant Wire Tension Control Using Fuzzy Method in Multi-Wire Saw

Machine
Zaojun Fang, De Xu, Min Tan and Zhiqiang Cao

Abstract— Constant wire tension is one of the key require- and the workpiece in the wafer slicing process based on
ments to achieve high wafer quality in multi-wire saw machine. hydrodynamics using finite element analysis method.
For the constant wire tension control, the wire tension analysis Since the slicing process is based on the friction between
is firstly derived based on the torque equilibrium of the motor
in the tension arm and the wire length variation. Then a closed- the silicon workpiece and the abrasives on the wire, the
loop wire tension control strategy is designed based on fuzzy stability of the wire tension should be controlled to ensure
method due to its complicated nonlinearity. In the controller, high wafer quality. Firstly, the tension behavior and tension
the tension arm angle is considered as the feedback variable, vibration need to be modeled and analyzed [8]–[14]. In [8],
and the angular velocity of the sending reel is selected as the the tension vibration due to the diameter changes of the wire
controlled variable. Experimental results are demonstrated to
verify the effectiveness of the proposed control methods in this barrel was analyzed. In [9] and [10], the tension behavior of a
paper. web in a winding process was studied based on the Hooke’s
law and Coulomb’s law. The tension force mechanism for
I. INTRODUCTION both a web section and the one with a dancer roll was
In solar cell production, a significant step is to cut the formulated in a unified approach in [11]. Song et al. [12]
large silicon crystal into thin wafers. To increase the cutting considered the wire tension to be proportional to the first-
efficiency greatly, a new technique named multi-wire slicing order delay of the line speed difference between two motors.
is invented [1]–[3], which has the merits of high efficiency Hakan et al. [13] studied the dynamic model between the
(thousands of wafers can be obtained per cutting round), web tension and the velocity of each roll in web winding
small kerf loss due to the thin wire and almost no restriction systems.
on the size and shape of the silicon ingot. To guarantee the stability of the wire tension in the
The working principle of the multi-wire saw machine is multi-wire saw machine, effective control strategy should be
described in [1], [4]–[6]. A single wire is supplied from a designed. Several control methods for the constant tension
sending reel (SR) through several pulleys. It is wound on in various systems have been proposed [13]–[19]. In [13], a
two rolls with multi-channels on the surface. Thus, a wire multivariable robust controller was proposed to control the
web is formed between the two rolls. Silicon crystal columns tension in a web transport system. The gains of the controller
are glued to a holder and pressed on the moving wire web. were scheduled according to the tension state. Mu et al.
Liquid abrasive slurry is sprayed through nozzles over the [14] designed a one-step ahead adaptive controller to control
wire web and carried by the wire into the slicing channel. the wire tension in a wire electrical discharge machining.
The friction between the ingot and the abrasives on the wire In [15], an oberver-based tension feedback controller with
slices the material. friction and inertial compensation was proposed. Chol et al.
From the working principle of the multi-wire saw ma- [18] presented a feedback control method for a moving tape
chine, it can be seen that the multi-wire slicing involves tension system, where a slide mode control algorithm was
complicated process between the slurry abrasive, the wire formulated to achieve desired level of tension. Due to the
and the silicon workpiece. Some researchers study the slicing complex interaction between the multivariables in the wire
process from microscopic point of view [1], [5], [7]. Moller tension, some intelligent control strategies are also adopted
[1] investigated the wire slicing process from two aspects: [20]–[23]. In [20] and [21], fuzzy logic control method was
the interaction of an individual particle of the slurry abrasive used in the tension control problem. Luo [23] designed an
with the crystal surface based on the fracture mechanics artificial neural network to isolate the cross coupling between
and the dynamical behavior of the abrasive particles in the speed and tension control loops.
the moving slurry between the wire and the workpiece. The quality of the sliced wafers is determined to a large
Bhagavat et al. [5] studied the interaction of the free abrasive extent by the constant wire tension. Thus, the wire tension
needs to be controlled to keep constant enough. In this
This work was supported by the National Natural Science Foundation paper, the wire tension vibration is analyzed deeply from
of China under Grant (61305024, 61273337). It was also supported by the the torque equilibrium of the motor and the wire length
National Key Technology Support Program under grant 2015BAF01B01.
Z. J. Fang, M. Tan and Z. Q. Cao are with the State Key Laboratory variation. Then unlike the tension control methods mentioned
of Management and Control for Complex Systems, Institute of Automation, above, an indirect wire tension control method based on the
Chinese Academy of Sciences, Beijing 100190, China [email protected] velocity difference between the processing unit and the SR
D. Xu is with the Research Center of Precision Sensing and Control,
Institute of Automation, Chinese Academy of Sciences, Beijing 100190, is proposed.
China. This paper is organized as follows. Section II describes

978-1-5090-4364-4/16/$31.00 ©2016 IEEE 1736

 vp Po
1 3 T Pl T Pr
vr
Pulley
P1 T T
4
2 5
L θ Arm
P2

7 6 Disk
9 8

Fig. 2. The tension arm position in the tension adjustment

Fig. 1. The configuration of the multi-wire saw machine, 
1 ingot holder,
2 processing unit, 
3 pulley, 
45 TA,  6 7 PAU, 8 SR, 9 WR
a large extent. At the worst, the wire will be broken, and
the rest of the silicon ingot will become useless. Thus, the
the system configuration and the working principle of the controller for the wire tension should be designed effectively.
multi-wire saw machine. In section III, the analysis of the In the following sections, the wire tension analysis is derived
tension vibration and the coordinate control of the TA and the based on the torque equilibrium of the motor in the TA and
SR are presented in detail. Finally, experimental results are the wire length variation. Then, the controller design for the
illustrated to verify the effectiveness of the proposed method. constant wire tension is given in detail.
II. S YSTEM D ESCRIPTION
A. Wire Tension Analysis
A. System Configuration
The wire tension is adjusted through the TA, which is
The configuration of the multi-wire saw machine is shown composed of a pulley, a arm and a disk, as shown in Fig.
in Fig. 1, which consists of five parts, i.e., sending part, wind- 2. The wire is wound through the pulley. One end is to the
ing part, processing unit, ingot holder, and other auxiliary processing unit. The other one is from the SR. The disk is
parts. Sending part is composed of a SR, a position adjust installed at the axis of a torch motor, which exerts force on
unit (PAU) and a Tension arm (TA). The SR sends new wire the disk and adjusts the wire tension through the arm and
through some pulleys to the processing unit. The PAU moves the pulley. As shown in Fig. 2, if the wire tension is the
the pulley P2 in the vertical direction, and keeps pace with setting one, the arm will be static at the position Po . On the
the SR. Thus it can guarantee that the wire section near the other hand, if the wire tension is smaller or larger than the
sending point is horizontal and that the wire is smoothly sent. setting one, the arm will move right or left. From the torque
Both the SR and the PAU are driven by server motors. The equilibrium principle of the motor, the following equations
TA adjusts the wire tension to keep it constant in the slicing can be achieved.
process, which is driven by a torch motor. The configuration
of the winding part is similar with the sending one except J θ̈ = Tm − (2Ft L + Ft D p ) cos θ − T0 (1)
that the winding reel (WR) is used to wind the used wire. where Tm is the electromagnetic torque of the motor, which
The processing unit is composed of two rollers grooved with is constant for the torch motor, T0 is the brake torch of the
constant pitches. The pitch size depends on the thickness motor, Ft is the wire tension at the time t, L is the length of
requirement of the wafers. One roller is driven by a server the arm from the center of the disk to the lower edge of the
motor, while the other is a slave roller without a motor. It is in pulley, D p is the diameter of the pulley, θ is the angle of the
the processing unit that the silicon crystal is sliced. The ingot arm from the vertical direction, J is the inertial moment of
holder is installed upside of the processing unit, which holds the motor.
the silicon ingot in the slicing process. The other auxiliary Suppose the steady state of the TA is F0 and θ0 and the
parts includes twelve pulleys, a blurry blender, a freezing variations of F and θ about F0 and θ0 are dF and d θ , the
machine and control devices. The pulleys guides the wire equation (1) can be expanded at F0 and θ0 using the Tayler
moving and makes it more stable. The blurry blender is used expansion method. If the terms with higher than two orders
to supply blurry abrasive with proper viscosity. The freezing are ignored, the expansion is given by
machine supplies cold water to the axes of the server motors
for temperature reduction. The control devices can guarantee J θ̈ = k0 + k1 dF + k2 d θ (2)
that the saw machine is in proper action.
where k0 , k1 and k2 are constants described as
III. T HE C ONSTANT W IRE T ENSION C ONTROL ⎧
⎪
⎨k0 = Tm − (2F0 L + F0 D p ) cos θ0 − T0
The constant wire tension is the core requirement for the
k1 = −(2L + D p ) cos θ0 (3)
multi-wire saw machine. If the wire tension changes sharply, ⎪
⎩
the quality of the resulted silicon wafers will degrade to k2 = (2F0 L + F0 D p ) sin θ0

1737
 P1 Lˌ P2
The equation (2) can be rewritten as P2'
O1 O2
1
dF = (−k2 d θ − k0 + J θ̈ ) (4) P3
O'2
k1 P4

Now, we analyze the wire tension from the wire length O3 P4'
variation, which can be computed from the adjustment of
the TA. Suppose the TA moves left, the computation of the vp vr
θ
wire length variation is shown in Fig. 3. The pulleys O1 and
O3 are fixed. The wire length between P1 and P2 is L0 . When
the arm swings from O2 to O2 , the wire length between P1
O4

and P2 is calculated as follows.

Three lines O1 O2 , O1 O2 and O2 O2 are drawn as auxiliary
lines. The length O2 O2 is computed as Fig. 3. The wire length variation due to the TA adjustment

1
O2 O2 = 2L1 sin θ (5)
2 where
⎧
where ⎪ −2L12 sin θ + 2L0 L1 cos θ
⎪
⎪ f ( θ ) =
1 ⎪
⎪
1 1
(L02 + 2L12 − 2L12 cos θ − 2L0 L1 sin θ ) 2
L1 = L + D p ⎪
⎪
2 ⎪
⎪
⎪
⎪
⎪
⎪ +L13 sin(α − β )β̇
Since ∠O1 O2 O2 = θ /2 and O1 O2 = L0 , in the triangle ⎪
⎪
⎪
⎨
O1 O2 O2 , the length O1 O2 is computed using the cosine law −2L12 cos θ − 2L0 L1 sin θ
f 2 (θ ) =
as follows. ⎪
⎪
1
(L02 + 2L12 − 2L12 cos θ − 2L0 L1 sin θ ) 2
 ⎪
⎪
⎪
⎪
  1 ⎪
⎪ 4(L12 sin θ − L0 L1 cos θ )2
P1 P2 = O1 O2 = L02 + 4L12 sin2 θ − 2L0 L1 sin θ (6) ⎪
⎪ −
2 ⎪
⎪ 3
⎪
⎪ (L02 + 2L12 − 2L12 cos θ − 2L0 L1 sin θ ) 2
⎪
⎪
The length P3 P4 is computed as ⎩
−L13 cos(α − β )β̇ 2 + L13 sin(α − β )β̈
P3 P4 = L0 − L13 cos α (7)
From (4) and (11), the wire tension vibration can be
where L13 is the distance between O1 to O3 , α = ∠O2 O1 O3 . computed as
As similar to (6), the length P3 P4 is given by J f2 (θ ) 2 k2 J k0
dF = − θ̇ − d θ + v˙s − (12)
P3 P4 = P1 P2 − L13 cos(α − β ) (8) k1 f1 (θ ) k1 k1 f1 (θ ) k1
From (12), it can be seen that the wire tension vibration is
where β = ∠O2 O1 O2 , which changes with θ . caused by the following factors: the angle, the angle variation
From the wire length of P1 P2 , P1 P2 , P3 P4 and P3 P4 , the and the angle velocity of the TA, and the acceleration of the
wire length variation can be computed as SR. Due to f1 (θ ) and f2 (θ ), the relationship between the
dL = P1 P2 − P1 P2 + P3 P4 − P3 P4 (9) wire tension vibration and these factors are highly nonlinear
and time-variant.
= P1 P2 − P1 P2 + P1 P2 − L13 cos α − P1 P2
+ L13 cos(α − β ) B. Controller Design for the Wire Tension

1 1) Choice of the Feedback and Controlled Variables: The
= 2L0 − 2 L02 + 4L12 sin2 θ − 2L0 L1 sin θ wire tension adjustment is through the TA, which is driven
2
− L13 cos α + L13 cos(α − β ) by a torch motor. If the tension deviates from the setting
one, the arm will swing to the left or the right, as shown
From another point of view, the wire length variation in Fig. 2. The size of the swing angle is proportional to the
can be computed from the velocity difference between the amount of tension deviation. On the other hand, the tension
processing unit and the SR as follows. deviation is mainly caused by the velocity difference between
 t the processing unit and the SR.
dL = (vs (t) − v p (t))dt (10) Thus, the arm of the TA can be considered as a sensor and
0 its swing angle is selected as the feedback variable. Since
where vs and v p are the wire velocities of the SR and the the velocity of the processing unit is constant, the angular
processing unit, respectively, as shown in Fig. 3. velocity of the SR is selected as the controlled variable.
Compute second derivative on the two sides of (9) and 2) Controller Structure: Since the controlled system is
(10), we can get highly nonlinear, time-variant and with a big lag, a feed-
forward fuzzy logic controller is designed for the tension
v˙s = f1 (θ )θ̈ + f2 (θ )θ̇ 2 (11) control. To compensate the radius and inertia variation of

1738
 μe, μu μde
the SR in the slicing process, and improve the performance
of the controller, the output membership functions are self-
NB NM NS 1 ZE PS PM PB NB NS 1 ZE PS PB
tuned.
The controller diagram for the wire tension is shown in
Fig. 4, which is composed of two closed-loops, a feedforward
channel and a parameter tuning channel. The inner loop is
-1 -2/3 -1/3 0 1/3 2/3 1 e, u -1 -1/2 0 1/2 1 de
velocity control for the SR, which is realized in the drive of
the servo motor. Its feedback signal is from the encoder of Fig. 5. The membership functions of the fuzzy controller
the motor. The outer loop is wire tension control, which is
based on the fuzzy controller. Its feedback signal is from the TABLE I
swing angle of the TA. The feedforward channel provides the RULE BASE OF THE FUZZY CONTROLLER
base value of the angular velocity of the SR. The parameter
tuning channel adjusts the parameters of the controller to de/e NB NM NS ZE PS PM PB
improve its performance. NB PB PB PB PM PS PS ZE
NS PB PM PM PS PS ZE NS
3) Feedforward Signal: Since the control objective of the ZE PM PM PS ZE NS NM NM
SR is to ensure that its peripheral velocity is equal to the PS PS ZE ZE NS NM NB NB
wire speed. Thus, the angular velocity of the SR computed PB ZE NS NS NM NB NB NB
from the wire speed is used as the feedforward signal, which
can be considered as the base angular velocity of the SR. The
feedforward angular velocity is computed as 4) Membership Functions: The error and the error change
of the swing angle of the TA are selected as the inputs of
vr
ωt = (13) the fuzzy controller as follows.
Rt
e(t) = θs − θ f (t) (18)
where vr is the wire moving speed, ωt is the base angular
de(t) = e(t) − e(t − 1) (19)
velocity of the SR, Rt is the radius of the SR at time t, which
is given by (14). where e(t) and de(t) are the error and the error change at
sample time t, θs is the setting swing angle of the TA where
Rt = R0 − 2 ∗ Rw ∗ N p (14) the tension is the setting one, θ f (t) is the swing angle of the
TA at real time. The left angle of the TA is defined as the
where Rt is the radius of the SR at the sample time t, which positive one.
varies in the slicing process. R0 is the initial radius of the The membership functions for the inputs and the output
SR, Rw is the radius of the wire, N p is the number of moving of the fuzzy controller are shown in Fig. 5. There are seven
direction changes of the PAU, which can be counted in the fuzzy sets for the input e and the output u. The fuzzy sets
slicing process as follows. NB, NM, NS, ZE, PS, PM, PB means negative big, negative
n median, negative small, zero, positive small, positive median,
N p = ∑ Dc (15) positive big, respectively. There are five fuzzy sets for the
i=1 input de, which are defined similarly.
where Dc is a flag indicating whether the moving direction Note that the universes of discourse for the inputs and
of the PAU is changed or not as follows. output of the fuzzy controller are all defined in the range
 of [-1,1], which is achieved by the input and output scaling
1, if Vp (t − 1) > 0 and Vp (t) < 0 factors.
Dc = (16) 5) Rule Base: The rule base of the fuzzy controller is
0, otherwise
established based on our understanding for the multi-wire
where Vp denotes the moving direction of the PAU, Vp > 0 saw machine and by trial and error, which is shown in Table
if the PAU moves up. I.
Due to the variation of the radius of the SR, the angular IF-THEN form is adopted for defining the rule in the base.
velocity of the SR computed with (13) is not accurate The k-th rule is described as
enough. It needs to be tuned precisely. The adjustment of the Rule k: if e is EFk and de is DEFk , then u is UFk . (20)
angular velocity of the SR is given by the fuzzy controller.
Thus the overall setting angular velocity of the SR is given where EF , DEF and UF are the fuzzy sets for the e, de and
by u, respectively.
According to the positive direction of the swing angle of
ωs = dω l + ωt (17) the TA and the relationship between the TA and the SR, the
first rule in the rule base is described as
where ωs is the setting angular velocity of the SR, dω l is its ”If the TA is in the left position and it tends to swing
adjustment part given by the fuzzy controller. The design of to the left, then the angular velocity of the SR should be
the fuzzy controller is given in detail as follows. increased.”

1739
 vr
Fuzzy controller
1/Rt

Defuzzification
θs

Fuzzification
θe wt
ke
Inference Servo vs F
1 − z −1
δku PI Driver SR TA
kde engine dw motor
T
θf θde

Rule s Encoder
base

Torch motor

Fig. 4. The controller for the wire tension

6) Defuzzification: The widely used defuzzification

method named center of gravity is adopted in the fuzzy
controller as follows.
n
∑k=1
l
ch μk (E j , DEk ,Uh , e, de)
u= (21)
∑i=1 μk (E j , DEk ,Uh , e, de)
35 Ingot holder

where u is the crisp output of the fuzzy controller, ch is

the center of area of the membership function of the fuzzy
Abrasive
set Uh in the consequent part of the given If-Then rule, nozzles
μk (E j , DEk ,Uh , e, de) is the membership value of the implied
fuzzy set k of the k rule, which is computed used the Max-
Min inference rule, nl is the number of the rules.

IV. E XPERIMENTS AND RESULTS Mai rolls

To investigate the effectiveness of the proposed control Wire array

methods, experiments are well conducted in a multi-wire pulley

silicon saw machine. The setup of the machine is shown

in Fig. 6. Fig. 6(a) shows the processing unit of the machine
(a)
including two main rolls, a wire array between the two rolls,
two abrasive nozzles and a ingot holder. Fig. 6(b) shows the
sending and the winding parts of the machine. They were
TA
structured in the same manner. Both of them consisted of a TA

SR or WR, a PAU and a TA. The diameter of the wire used

in this machine was 0.12 mm, and the thickness of the sliced
wafer was 0.2 mm. The largest amount of wafers that can be
obtained in one slicing round were 2300. The highest wire
speed of this machine was about 800 m/min. This machine WR SR
adopts FANUC servo motors and related numerical control PAU
PAU
system. The basic logic and flow controls were implemented
in the programmable logic controller (PLC), and the smooth
wire sending and constant wire tension controls proposed in Reel base
this paper were implemented using C language.
To test the effectiveness of the proposed constant wire
tension control strategy, experiments about the coordinate
(b)
control of the TA and the SR are conducted. The diameter of
the disk in the TA was 850 mm, and the length of the arm in
Fig. 6. The experimental setup, (a) the processing unit of this machine,
the TA was 3100 mm. The parameters in the controller were (b) the configuation of the sending and winding parts
set as follows. vt = 600 m/min, S p1 = {0.3, 0.005, 0.002},
S p2 = {0.1, 0.005, 0.06}, dt1 = 0.5rad/s, dt2 = 0.2rad/s.
The angle of the TA and the angular velocity changes of
the SR and in the experiment are shown in Fig. 7. From Fig.

1740
 [2] A. Mueller, N. Cherradi, P. M. Nasch, ”The challenge to implement
2
thin wafer potential with wire saw cutting technology”, in Proceedings
Angle of the TAU (deg)

1 of 3rd World Conference on Photovoltaic Energy Conversion, May 11-

18, Osaka, Japan, 2003, pp. 1475-1478.
0 [3] A. Bidiville, K. Wasmer, J. Michler, P.M. Nasch, M. V. Meer, C. Ballif.
”Mechanisms of wafer slicing and impact on wafer properties”, vol.
−1 18, pp. 563-572.
[4] K. Saito, F. Furuta, H. Umezawa, ”multi-wire slicing of large grain
−2 ingot material”, in Proceedings of the 14th International Conference
0 1000 2000 3000 4000 5000 6000 7000
Control period (8mm) on RF Superconductivity, Berlin, Germany, 2009, pp. 467-472.
Angular Velocity of the SR (rad/s)

[5] M. Bhagavat, V. Prasad, I. Kao, ”Elasto-hydrodynamic interaction in

48.8
the free abrasive wafer slicing using a wiresaw: modeling and finite
48.6 element analysis”, Journal of Tribology-Transactions of the ASME,
vol. 122, No. 2, 2000, pp. 394-404.
48.4 [6] L. Q. Zhu, I. Kao, ”Galerkin-based modal analysis on the vibration
of wire-slurry system in wafer slicing using a wiresaw”, Journal of
48.2 Sound and Vibration, vol. 283, No. 3-5, 2005, pp. 589-620.
[7] T. Watanabe, ”Optimization quartz crystal slicing conditions using
48 multi-wire-saw”, in Proceedings of IEEE International Frequence
0 1000 2000 3000 4000 5000 6000 7000
Control period (8mm) Control Symposium and PDA Exibition, 2002, pp. 386-393.
[8] J. L. He, H. Y. Dong, X. Y. Ye, ”Research of multi-wire saw cutting
slurry control system”, in Proceedings of International Conference on
Fig. 7. The results of constant wire tension control Advanced Technology of Design and Manufacture, 2010, pp. 403-406.
[9] T. Bastogne, H. Noura, P. Sibille, A. Richard, ”Multivariable identifi-
cation of a winding process by subspace methods for tension control”,
Control Engineering Practice, vol. 6, pp. 1077-1088, 1998.
[10] D. Knittel, E. Laroche, D. Gigan, H. Koc, ”Tension control for
winding systems with two-degree-of-freedom H∞ controllers”, IEEE
Transactions on Industry Applications, vol. 39, No. 1, pp. 113-120,
2003.
[11] T. Sakamoto, Y. Fujino, ”Modelling and analysis of a web tension
control system”, in Proceedings of the IEEE International Symposium
on Industrial Electronics, July 10-14, 1995, pp. 358-362.
[12] S. H. Song, S. K. Sul, ”A new tension controller for continuous strip
processing line”, IEEE Transactions on Industry Applications, vol. 36,
No. 2, pp. 633-639, 2000.
[13] H. Koc, D. Knittel, M. Mathelin, G. Abba, ”Modeling and robust
control of winding systems for elastic webs”, IEEE Transactions on
Control Systems Technology, vol. 10, No. 2, pp. 197-208, 2002.
[14] M. T. Yan, P. H. Huang, ”Accuracy improvement of wire-EDM by
real-time wire tension control”, International Journal of Machine Tools
Manufacture, vol. 44, pp. 807-814, 2004.
[15] K. C. Lin, ”Observer-based tension feedback control with friction
Fig. 8. The sliced silicon wafers and inertia compensation”, IEEE Transactions on Control Systems
Technology, vol. 11, No. 1, pp. 109-118, 2003.
[16] B. J. An, S. H. Park, B. Y. Kim, T. D. Yum, D. H. Kang, M. H.
7, it can be seen that the largest adjusting angle of the TA Lee, ”Design and tension control of filament winding system”, in
Proceedings of the IEEE International Conference on Systems, Man,
was −1.15o , and the mean value was 0.28o . From Fig. 7, it and Cybernetics, vol. 2, 1999, pp. 660-670.
can be seen that the tension was favorably controlled. The [17] S. L. Ren, Y. N. Lai, Y. Z. Wang, H. Lu, ”A new fiber winding pre-
sliced silicon wafer is shown in Fig. 8. It can be seen that the cision tension control system”, in Proceedings of IEEE International
Conference on Automation and Logistics, Shenyang, China, August,
surface of the sliced wafers were very smooth. Thus, high 2009, pp. 233-236.
wafers quality can be obtained using the proposed method. [18] S. B. Chol, C. C. Cheong, G. W. Kim, ”Feedback control of tension
in a moving tape using an ER brake actuator”, Mechatronics, vol. 7,
V. C ONCLUSIONS No. 1, pp. 53-66, 1997.
[19] M. A. Valenzuela, J. M. Bentley, R. D. Lorenz, ”Sensorless tension
This paper presents a control strategy for constant wire control in paper machines”, IEEE Transactions on Industry Applica-
tension in multi-wire saw machine. The wire tension vibra- tions, vol. 39, No. 2, pp. 294-304, 2003.
tion is analyzed deeply based on the torque equilibrium of [20] F. J. Sharifi, ”A neuro-fuzzy system for looper tension control in
rolling mills”, Control Engineering Practice, vol. 13, pp. 1-13, 2005.
the motor in the TA and the wire length variation. Then a [21] K. Okada, T. Sakamoto, ”An adaptive fuzzy control for web tension
closed-loop wire tension control strategy is designed with control system”, in Proceedings of the 24th IEEE Annual Conference
the TA angle as the feedback sensor and SR velocity as the on Industrial Electronics, vol. 3, 1998, pp. 1762-1767.
[22] D. C. Chen, H. Zhang, M. F. Wang, ”An intelligent tension control
controlled variable. To solve the complicated nonlinearity in system in strip unwinding process”, in Proceedings of IEEE Inter-
the tension control, fuzzy method is adopted in the controller national Conference on Control and Automation, Guangzhou, China,
design. Experimental results illustrate that high wafer quality May 30 to June 1, 2007, pp. 342-345.
[23] F. L. Luo, ”Multiple-page-mapping backpropagation neural network
can be obtained using the proposed control methods in this for constant tension control”, in Proceedings of the IEE Electric Power
paper. Applicaton, vol. 145, No. 3, 1998, pp. 239-245.

R EFERENCES
[1] H. J. Moller, ”Basic mechanisms and models of multi-wire slicing”,
Advanced Engineering Materials, vol.6, No.7, pp. 501-513, 2004.

1741