Journal of Process Control 20 (2010) 643652

Contents lists available at ScienceDirect

Journal of Process Control

journal homepage: www.elsevier.com/locate/jprocont

A novel cross directional register modeling and feedforward control in multi-layer roll-to-roll printing
HyunKyoo Kang a,1, ChangWoo Lee b, KeeHyun Shin b,*
a b

Department of Mechanical Design, Konkuk University, 1 Hwayang-Dong, Gwangjin-Gu, Seoul 143-701, Republic of Korea Department of Mechanical Engineering, Konkuk University, 1 Hwayang-Dong, Gwangjin-Gu, Seoul 143-701, Republic of Korea

a r t i c l e

i n f o

a b s t r a c t
For the adaption of roll-to-roll printing method to the printed electronics, it is mandatory to increase the resolution of register control. Therefore, it is desired to derive a mathematical model for register and to develop a controller to reduce the register error. The cross direction register error was derived considering both the lateral motion of a moving web and the transverse position of a printing roll. And a feedforward control method was proposed to cancel out the disturbance of CD register from upstream span. The proposed controller could be used to improve the performance of the CD register controller in a large area roll-to-roll printing machine. The mathematical modeling and proposed control method were validated by numerical simulations and experimental verications in various operating conditions using a multilayer direct gravure printing machine. These results show that the proposed feedforward control scheme greatly improves the control performance of register control in overcoming the upstream disturbances. 2010 Elsevier Ltd. All rights reserved.

Article history: Received 6 July 2009 Received in revised form 22 October 2009 Accepted 27 February 2010

Keywords: 2-D Register Multi-layer registration Feedforward control Roll-to-roll Gravure printing Printed electronics

1. Introduction The rising demand for the fabrication of exible electronics using roll-to-roll technology exposes the abundant challenges of conducting low cost and highly productive printing trials for experiments. Many demonstrations have been conducted using discrete printing methods such as sheet printing, ink-jet printing, spin coating because of their high accessibility and ability to experiment with fewer material [16]. Other researchers carried out laboratory scale continuous printing with gravure, exography, offset, etc. [712], but this is not sufcient to cover the adaption of large area, continuous roll-to-roll printing for printed electronics. If there are several printing rolls between the unwinder and rewinder for multi-layer patterning, each printing roll has not only a variation of phase but the web also goes along the path with a lateral movement and variation of strain. Thus the strain and lateral position of the substrate should be controlled to minimize the register errors in successive roll-to-roll printing. Register error of a moving web is dened in two printing rollers as the relative difference of the distance between the previous printed pattern by the upstream printing roller and the printed pattern in the downstream printing roller. Register error is dened

* Corresponding author. Tel.: +82 2 450 3072; fax: +82 2 447 5886. E-mail addresses: hyunkyoo@gmail.com (H. Kang), khshin@konkuk.ac.kr (K. Shin). 1 Tel.: +82 2 3436 0321; fax: +82 2 447 5886. 0959-1524/$ - see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.jprocont.2010.02.015

as two-dimensional errors: machine direction (MD) and cross direction (CD) errors as shown in Fig. 1. The register error is as critical as surface topography in ensuring the functionality of a printed circuit without mistakes. It determines the printing quality of nal products. In addition, the printing quality is more important in printed electronics because shorts or leakages could be generated due to the register error of multilayered patterns in printed electric devices, such as organic photovoltaic and exible displays. Therefore, the register error of a moving web should be precisely controlled for the fabrication of printed electronics. Brandenburg derived a linear mathematical model of the machine direction register error of a moving web in a rst-order differential equation by using an equilibrium equation of mass which is transported by printing rolls. A non-interacting control method between the tension and cut-off register error was also proposed [13,14]. Yoshida proposed a nonlinear MD register controller to compensate for a downstream register error caused by upstream tension uctuation in gravure printing. Komatsu developed a delay-dependent nonlinear control approach by adjusting the new coordinate and delay-dependent feedback law [15,16]. All of these results were carried out with the typical assumption that a lateral position of a moving web is not changed in the machine direction, so there is no CD register error in multi-layered printed patterns. However, actually the lateral position of a moving web varies along the web path within a certain range, and it makes CD register errors in printed patterns. Therefore, it is necessary to

644

H. Kang et al. / Journal of Process Control 20 (2010) 643652

Fig. 1. Two-dimensional register errors.

derive a mathematical model of the CD register for more accurate register control, but there has been no report that deals with the dynamics of the CD register. In this paper, a mathematical model of the CD register was derived by using both the lateral dynamics of a moving web and the relative difference of lateral positions between a printing roll and a web which goes through the printing section. The CD register was controlled by the translation motion of a printing roll, and the translation could generate a transient CD register in downstream printing sections. A feedforward controller was designed to compensate for the deterministic disturbance of the CD register. The modeling was validated, and the performance of the proposed controller was veried by numerical simulations and experiments. The results show that the proposed modeling is accurate enough to describe the CD register and the performance of the suggested controller is more effective in reducing the deterministic CD register errors than that of the traditional PID controller.
Fig. 2. Boundary conditions.

2. Mathematical modeling In this section, the modeling of the lateral motion of a moving web, which was suggested by Shelton, was summarized and the modeling of CD register was derived by using the lateral motion of a moving web, including translations of printing rolls [17,18]. Shelton derived rst- and the second-order models of the lateral motion of a web. The lateral motion could be described more accurately by the second-order model than the rst-order one. Thus the second-order model was summarized and used for the derivation of the CD register error. 2.1. Second-order lateral motion of a moving web The differential equation of the web elastic curve could be derived from beam theory if the tension acts on the web. The beam equation, which is a fourth-order differential equation, is shown in Eq. (1). Eq. (2) is a general solution of Eq. (1), and the boundary conditions shown in Fig. 2 can be used to determine the coefcients of Eq. (2). The boundary conditions are Eq. (3):

@4y @2y K 2 2 0; 4 @x @x

K2

T EI

where EI is the bending stiffness

y C 1 sinhKx C 2 coshKx C 3 x C 4 y0 y0 ; h0 h0 ; yL yL ; hL hL

2 3

Using Eq. (2), the curvature of the downstream roll could be obtained as Eq. (4):

 @ 2 y   @x2 

f1 KL L2

y0 yL

xL

f2 KL f3 KL hL h0 L L

where
KL coshKL1 f1 KL KL sinhKL2 coshKL2 KLKL coshKLsinhKL f2 KL KL sinhKL2 coshKL2 KLsinhKLKL f3 KL KL sinhKL2 coshKL2 :
2

H. Kang et al. / Journal of Process Control 20 (2010) 643652

645

Substituting Eqs. (4) and (5) into Eq. (6), the second-order differential equation could be derived as Eq. (7):

d yL dt
2

a1

dyL dy d wL a2 yL a3 0 a4 y0 a5 uL a6 u0 b1 2 dt dt dt dwL dw0 b2 b3 dt dt

@y u L f2 KL f1 KL ; a2 ; ; hr ; s ; a1 @x c V s s2 2 f3 KL f1 KL V ; a4 ; a5 a3 f2 KL s s2 Lc
2

Fig. 3. Poles of lateral model (KL is 010).

where a6 V f3 KL; b1 1; b2 f2 KL ; b3 f3 KL, L = span length, Lc s s V = velocity, c = half width of roll and u = arc of rotation. In the printing section, the CD register error should be controlled by only translational motions of printing rollers so that the lateral motion of Eq. (7) yields as in Eq. (8). The transfer function, the response of yL to the input of y0, wu and w0 could be derived as Eq. (9):

The lateral velocity of a web is as shown in Eq. (5) and the lateral acceleration is given by Eq. (6):

d yL dt
2

a1

dyL dy d wL dwL dw0 b2 a2 yL a3 0 a4 y0 b1 b3 2 dt dt dt dt dt 8 9

1 2

dyL @y dwL V hr L dt @x dt  2 2 2  d yL @ y d wL V2 2 2 2 @x  dt dt L
where
@yL @x

5 6

Y L s AsY 0 s BsW L s CsW 0 s

where As s2a3 sa4 ; Bs a sa
1 2

b1 s2 b2 s s2 a1 sa2

b s and Cs s2 a 3sa .

is the slope of the web at the downstream roll.

For the stability analysis of the lateral model equation (9), poles h 1=2 i of the model are determined as a1 a2 4a2 =2 and 1

Fig. 4. Schematic of three-layer printing system.

Fig. 5. Three-layer gravure printing machine.

646

H. Kang et al. / Journal of Process Control 20 (2010) 643652

Fig. 6. CD register errors and lateral position of the web at each printing roll caused by the translation of the second printing roll (30 m/min, 100 N).

Fig. 7. CD register error and lateral position of the web at each printing rolls caused by translation of the second printing roll (40 m/min, 100 N).

h 1=2 i =2. In Fig. 3, poles of lateral model were dea1 a2 4a2 1 picted with varying KL (010). In roll-to-roll system, the KL is less than 1, because of large amount of Youngs modulus (E) of substrate. In the experiment, the parameter KL is 0.66. Therefore, the range of KL, 010, is reasonable and it can be proved the lateral model is stable in various KL. 2.2. Cross directional register error modeling The schematic view of the three-layer printing system is shown in Fig. 4. When the second printing roll prints a pattern on the web, the CD register error is produced by the relative difference of the lateral positions of both the successive printing roll and the moving web. The CD register may occur due to the relative distances of printing rolls even with a straight moving web. Furthermore, there are no translations of printing rolls, and the CD register error is also induced by the variation of lateral positions of the moving web. Accordingly it is signicant to control the lateral position of both the web and the printing roll, compared to the lateral position of the upstream-pattern that has traveled in the instance of current printing of a pattern for CD registration.

The CD register is composed of two patterns that imply the history of printing as a period of time constant (L/V). This means that the upstream-pattern (roll No. 1 in Fig. 4) made ahead of the time constant and the pattern currently being printed downstream (roll No. 2 in Fig. 4) complete the denition of the CD register as shown in Eq. (10):

ry;n t yn t wn t yn1 t s wn1 t s

10

where ry,i is the ith CD register, yi is the lateral position of the ith web and wi is the lateral position of the ith roll. The lateral position of a moving web is calculated by Eq. (9), and the lateral position of a printing roll is the input value of a translational motion. Eq. (10) could be Laplace-transformed as in Eq. (11):

Ry;n s Y n s W n s Y n1 s W n1 sess

11

Under the assumption that all spans, which are structured between printing rolls have same length, Eq. (9) of the lateral position of the web would be expanded as in Eq. (12):

Y n1 s AsY n2 s BsW n1 s CsW n2 s Y n s AsY n1 s BsW n s CsW n1 s

12

Substituting Eq. (12) into (11) and rearranging the resulting equation yields Eq. (13):

H. Kang et al. / Journal of Process Control 20 (2010) 643652

647

Fig. 8. Block diagram of CD register controller. (a) Conventional CD register controller using feedback. (b) Proposed CD register error controller using feedforward and feedback loop.

Fig. 9. Simulation model.

Ry;n s AsY n1 s Y n2 sess BsW n s W n1 sess CsW n1 s W n2 sess W n s W n1 sess 13

Table 1 Simulation parameters. Simulation parameters Operating tension Operating velocity Length of span (between idle rolls) Length of span (between printing rolls) P gain of CD register feedback controller D gain of CD register feedback controller Thickness of web Width of web Youngs modulus Moment of inertia (I) Number of span between printing rolls Warm gear ratio Values 100 N 30, 40 m/min 1m 8m 0.01 0.15 12 lm 1m 3.6 GPa 1 106 m4 8 4.7746 104 m/rad

In addition, the roll-to-roll printing machine consists of many idle rolls in the middle of the printing sections for the prevention of wrinkling of a moving web or the construction of desired span length and wrap angle for dryers, prevention of slippage, and so on, as Fig. 5. Therefore, Eq. (10) of the CD register should be expanded in order to include idle rolls as in Eq. (14). Substituting Eq. (12) of the lateral position of the web into Eq. (14) and its Laplace-transformation yields Eq. (15):

r y;n t yn t wn t ynm t sm wnm t sm

14

648

H. Kang et al. / Journal of Process Control 20 (2010) 643652

Ry;n s Asm1 AsBs Cswnm s Bswn s wn s wnm se

sm s

Bswnm se

sm s

15

which were assumed to be the same. The difference is also due to mechanical loss such as belting and inertia effect. 3. Proposed CD register controller 3.1. Block diagram of CD register controller In this section, experiments on the CD register control of a moving web using a feedforward controller are described. For the experiment, the step input of the translation motion is generated by the second printing roll as a disturbance. The third printing roll is used to control the CD register error in this experiment. The block diagram of a conventional CD register controller using feedback control is illustrated in Fig. 8(a). The CD register control system is composed of a side-lay motor, motor driver and optical sensors for measuring register error. The CD register error could be regulated by the translation motion of the printing rolls compared with the relative transverse position of a pair of printed patterns. The mathematical model of side-lay motor is described in Eq. (16):

where m is the number of spans between a pair of printing rolls structured by idle rolls and s is the complex number. Step responses of the lateral position and the CD register caused by a translational motion of the second printing roll are depicted in Figs. 6 and 7 at different operating conditions. The translation generates a transient variation of the lateral position of y1, and the lateral variation transferred to the downstream as y2 after time delay (L/V), as shown in Figs. 6(b) and 7(b). On the other side, the translation motion induces steady-state CD register disturbances at the second printing roll (Ry1), and they are transferred to the downstream printing roll (Ry2). The transferred CD register has the same magnitude and reverse direction compared with the upstream CD register in Figs. 6(a) and 7(a). Therefore, in order to increase the resolution of the registration, the transferred CD register of the downstream should be compensated. In Figs. 6 and 7, there were difference in the settling time of Ry2 between the estimated and experimental results because each actual span length of printing sections structured by idle rolls are different from estimated ones

Gm s

hs km Us s2 am s km

16

Fig. 10. Performance of the feedforward including PID controller (computer simulation: T = 100 N, V = 30 m/min). (a) CD register error. (b) Lateral position.

Fig. 11. Performance of the feedforward including PID controller (computer simulation: T = 100 N, V = 40 m/min). (a) CD register error. (b) Lateral position.

H. Kang et al. / Journal of Process Control 20 (2010) 643652

649

Fig. 12. Experimental setup.

where U(s) is the motor input, am, km are motor constants and h is the angular position of motor. The rotation of the side-lay motor is converted into the translation motion of the printing roll using the worm gear as in Eq. (17):

Ks

ws kw hs

17

where kw is constant of worm gear ratio. The lateral dynamics of a web caused by translation motion is determined by a transfer function as in Eq. (18) and the effect of translation motion of upstream printing roll which was transferred several idle rolls is expressed as in Eq. (19):

Bs

s2

b 1 s2 b 2 s a1 s a2

18 19

Ds Asm1 fAsBs Csg

where m is the number of spans between a pair of printing rolls structured by idle rolls and s is the complex number. A block diagram of the proposed feedforward controller is shown in Fig. 8(b). A translation motion of the printing roll could reduce the CD register error in current roll but it also causes a CD disturbance for the CD register error to the following printing roll. As a result of that the proposed controller is aimed to compensate for the CD register error generated by translation of the upstream printing roll. Substituting Eqs. (9) and (12) into Eq. (18) and the Laplace-transforming CD register error yields Eq. (20): The control input to eliminate the effect of the upstream disturbance of wnm is calculated as Eq. (21) and the control input of the CD register error is shown in Eq. (22):

Fig. 13. Three-layer gravure printing machine.

3.2. Simulation results Numerical simulations to verify the performance of the proposed feedforward CD register controller were carried out using SIMULINK (The Mathworks Inc., Natick, MA). The three-layer printing system is designed as shown in Fig. 9. The system includes two kinds of measuring sensors: EPS (edge position sensor) for the lateral position of the web and OS (optical sensor) for the register error of the printed pattern. All of the tensions except the printing section are controlled with a PI controller and loadcell. In the printing sections, the printing rolls are controlled to preserve the same phase of the printing patterns for MD register control. The printing rolls also have sidelay motors and worm gears to regulate the CD register by translation. The simulation conditions are described in Table 1.

Ry;n s Hus swnm s Hds swn s Hus s U FF s wnm s Hds s Us U FB s U FF s

where Hus s Ds 1 Bsesm s ; Hds s Bs 1 U FB control input of PID controller.

20 21 22
and

650

H. Kang et al. / Journal of Process Control 20 (2010) 643652

Fig. 14. Loadcell (left), optical sensor (middle) and infrared sensor (right).

Fig. 15. Performance of the feedforward including the PID controller. (a) CD register error at 100 N, 30 m/min. (b) CD register error at 100 N, 40 m/min.

Fig. 16. Lateral position of the web (experiment: T = 100 N, V = 30 m/min). (a) Lateral position of the PD control. (b) Lateral position of the feedforward including the PD control.

Figs. 10 and 11 show the performance of the conventional and the proposed controller at different operating velocities. The disturbance of the translation of the second printing roll starts at

5 s, and the third printing roll controls the CD register with PD and feedforward controller. The operating tension is 100 N. The line speed is 30 and 40 m/min in Figs. 10 and 11, respectively.

H. Kang et al. / Journal of Process Control 20 (2010) 643652

651

amplier (Dover, TI17), as shown in Fig. 14(left). The register marks were measured by an optical sensor and amplier (ArTec. Co.), as shown in Fig. 14(middle). The lateral positions of the moving web were measured by infrared sensor (FIFE, SE-23) with an amplier, as shown in Fig. 14(right). Data acquisition software, LabView 8.2 (National Instruments), was used for collecting and saving the signal of loadcell, infrared sensor and optical sensors with DAQ board (National Instruments, SCC-68) and an A/D converter (National Instruments, PXI-6251). The experimental results are shown in Figs. 1517, and the operating conditions are 100 N of tension and 30 and 40 m/min of velocity. The proposed controller reduced the transient CD register error 0.2 mm lower than the PD controller in simulation, as shown in Figs. 10(a) and 11(a). The CD register error was decreased 0.50.3 mm lower than by the PD controller in experimental result as shown in Fig. 15. The difference in performance is due to the modeling error, mechanical loss and cross direction slip between the web and the nip roller in printing. In Figs. 16 and 17, the second lateral position of web (y2) of the feedforward including the PD control is changed more than that of the PD control. However, the larger displacement lateral (y2) of the feedforward control increases the performance of the CD register. It is concluded that even though the proposed controller cannot reject the CD register error perfectly, the proposed control method reduces the transient CD register error more than 50% in settling time and 40% in overshoot compared with the PD control only.

4. Conclusion A novel mathematical model for the CD register error was derived by using the lateral position of a moving web, translation motion of printing rolls and time constant (L/V). The register error occurred because of the relative difference between adjacent printed patterns on the web. Therefore, the lateral position of a web should be considered to control the CD register error. A feedforward controller was designed to reject the deterministic disturbance in a previous span caused by the translational motion of the printing roll. The performance of the proposed controller was demonstrated to be very effective by the numerical simulations and experimental studies. Generally printing rolls should be translated to control the CD register; therefore, the proposed feedforward controller is very useful for improving the performance of the CD registration. Acknowledgements This research was supported by the Seoul R&BD Program (10848) and the Korea Foundation for International Cooperation of Science & Technology (KICOS) through a grant provided by the Korean Ministry of Education, Science & Technology (MEST) in K20701040600-09A0404-05410. References
[1] H. Yan, Z. Chen, Y. Zheng, C. Newman, J. Quinn, F. Dotz, M. Kastler, A. Facchetti, A high-mobility electron-transporting polymer for printed transistors, Nature 447 (2009) 679687. [2] T. Sekitani, M. Takamiya, Y. Noguchi, S. Nakano, Y. Kato, T. Sakurai, T. Someya, A large-area wireless power-transmission sheet using printed organic transistors and plastic MEMS switches, Nature 6 (2007) 413417. [3] S. Moller, C. Pelov, W. Jackson, C. Taussig, S. Forrest, A polymer/semiconductor write-once read-many-times memory, Nature 426 (2003) 166169. [4] S. Forrest, The path to ubiquitous and low-cost organic electronic appliances on plastic, Nature 428 (2004) 911918. [5] A. Briseno, S. Mannsfeld, M. Ling, S. Liu, R. Tseng, C. Reese, M. Roberts, Y. Yang, F. Wudl, Z. Bao, Patterning organic single-crystal transistor arrays, Nature 444 (2006) 913917.

Fig. 17. Lateral position of the web (experiment: T = 100 N, V = 40 m/min). (a) Lateral position of the PD control. (b) Lateral position of the feedforward including the PD control.

As shown in Figs. 10 and 11, the proposed controller reduced the CD register error to less than 70% of the overshoot of the system response where the feedback control was used. The residue of CD register errors are caused by the feedforward controller not including the side-lay motor dynamics. The proposed control method could be used to reject the CD register caused by transition of a printing roll between adjacent spans.

3.3. Experimental results The experimental studies were carried out in order to verify the performance of the proposed feedforward CD register controller using the three-layer gravure printing machine. The experimental setup is shown in Fig. 12, and experimental conditions are same as those of Table 1. Fig. 13 shows the system conguration for the experiment. The system includes unwinding, rewinding, infeeding, outfeeding, three printing rollers, a commercial register controller (ArTec. Co.), and a main controller (Bosch Rexroth, PPC). In each span, tensions were measured by the loadcell (Dover, Model C2DFL) with an

652

H. Kang et al. / Journal of Process Control 20 (2010) 643652 completely by means of mass printing technologies, Org. Electron. 5 (2007) 480486. G. Brandenburg, New mathematical models for web tension and register error, in: Proc. Third Int. IFAC Conf. on Instrum. Automation in the Paper, Rubber Plast. Ind., Brussels, May, 1976, pp. 2426. G. Brandenburg, S. Geibenberger, A. Klemm, Noninteracting control of web forces and cut-off register errors in rotary printing presses with electronic line shafts, EPE 16 (2006) 3844. H. Komatsu, T. Yoshida, S. Takagi, T. Shen, Y. Muto, Improvement of printing accuracy via web handling control, in: Int. Conf. on Control Autom. Syst., 2007, pp. 953956. T. Yoshida, S. Takagi, Y. Muto, T. Shen, Register control of sectional drive rotogravure printing press, in: Proc. of the 41st CIRP Conf. Manuf. Syst., 2008, pp. 417420. J.J. Shelton, K.N. Reid, Lateral dynamics of a real moving web, ASME J. Dyn. Syst. Meas. Control 93 (1971) 180186. J.J. Shelton, Lateral dynamics of a moving web, Ph.D. Dissertation, Oklahoma State Univ. Stillwater, 1968.

[6] H. Kang, H. Sung, T. Lee, D. Kim, C. Kim, Liquid transfer between two separating plates for micro-gravure-offset printing, J. Micromech. Microeng. 19 (2009) 19. [7] A. Vombrock, J. Ding, D. Sung, H. Tseng, V. Subramanian, Printing and scaling of metallic traces and capacitors using a laboratory-scale rotogravure press, in: Proc. Conf. Flexible Electron. & Displays, 2009. [8] T. Kawahara, K. Doushita, H. Tada, A large-area patterned TiO2/SnO2 bilayer type photocatalyst prepared by gravure printing, J. SolGel Sci. Technol. 27 (2003) 301307. [9] H. Nokki, J. Kallioinen, T. Kololuoma, V. Tuboltsev, J. Tommola, Dynamic preparation of TiO2 lms for fabrication of dye-sensitized solar cells, J. Photochem. Photobiol. 182 (2006) 187191. [10] M. Pudas, J. Hagberg, S. Leppvuori, Printing parameters and ink components affecting ultra-ne-line gravure-offset printing for electronics applications, J. Eur. Ceram. Soc. 24 (2004) 29432950. [11] M. Pudas, J. Hagberg, S. Leppvuori, IEEE Trans. Electron. Packaging Manuf. 25 (2002) 335343. [12] A.C. Huebler, F. Doetz, H. Kempa, H.E. Katz, M. Bartzsch, N. Brandt, I. Hennig, U. Fuegmann, S. Vaidyanathan, J. Granstrom, Ring oscillator fabricated

[13]

[14]

[15]

[16]

[17] [18]