FIELD SERVICE REPORT No. 47

To:	(Proprietary)
From:	Brian Boulter - Chief Consulting Engineer
Date:	March 13th 2001
Subject:	Field Trip Report, 3W Graphics No.2 Printing Line, Monterey Mexico

The author made a field trip to investigate the causes of poor print registration on the 3W Graphics #2 Printing Line in Monterey Mexico. The print registration was reported to be out of specification by the customer, 3W Graphics. The drive vendor Rockwell Automation, commissioned ApICS LLC to work with RA field service engineers to identify the problem, and develop a corrective action plan.

There were two registration systems on the line. The first system was integral to the 5 color printing section, and was not the source of the registration problems. A second registration scheme was in place after the print drying section. This registration controller regulated the placement of glossy coats and other printing effects over the printed color images on the printed paper. Some of the coating material was similar to that found on "scratch'n'sniff" products. This registration scheme was implemeted with a novel neural-network based vision-registration sensor that was located approximately 30 linear [ft] from the imprint roll with associated actuating drives.

The specified tolerances for the specialty coating section were for placement within +/- 0.75 [mm] of set-point. The measured deviations were +/- 2.2 [mm] for print line speeds of 2800 [ft/min], and +/- 2.7 [mm] for print line speeds of 1800 [ft/min]

The trip was made to the site on 6/12/00 – 6/18/00. The following observations and corrective actions were taken.

The following observations were made while on-site:

1. The floatation oven-dryer in the print drying section was observed to introduce a significant amount of "noise" into the paper exit speed from the dryer. This could be observed by using a strobe light and strobing the printing section registration marks. The reason for the observed noise was identified as a result of the variation in stretch rates during drying in the oven, combined with oven floatation flutter. In terms of registration positioning, the noise disturbance was measured to be +/- 0.25 [mm]. Based on this measurement, the registration control needed to provide better than +/- 0.5 [mm] registration control.
2. The existing registration control scheme was observed to be unstable, and oscillated at a fixed frequecy of 2-3 [rad/sec], with a fixed amplitude in terms of positioning, of approximately +/- 2 [mm] at top print speed.
3. The registration sensor feedback was observed to be noisy (about +/30 counts out of 4095 Counts). Given that the registration control range was +/- 6 [mm], the sensor noise translated into +/- 0.04 [mm] of the registration control range.
4. There was significant backlash observed in the drive train of the embossing roll for in the imprint section. This was measured to be approximately +/- 0.015 [rad] and resulted in a measured +/- 45 [counts] of speed feedback jitter out of +/- 4095 counts of speed feedback during normal operation. The backlash in the imprint roll drive section significantly impacted the registration of the imprint (estimated to be +/- 0.25 [mm]).

5. The registration control algorithm was initiated with all feedforward terms set to zero (i.e. inertia compensation etc.) These termes were turned of, as it was explained to the author, as a result of the amplification of system noise (from the backlash in the drive train, and noise in the registration sensor feedback) that was observed when the feedforward compensation schemes were enabled.

6. The distance between the actuation device, and the registration feedback introduced a significant lag in the registration closed loop reponse (the amount of lag in degrees of phase margin lost was dependant on the the line speed - That is: the faster the line speed the smaller the loss of phase margin and visa-versa). At top line speed the phase shift for the existing registration control tuning was estimated to be approximately 75 [deg] at crossover.

1. The location of the registration feedback sensor was to far from the imprint drive. It was moved by the mechanical O.E.M. to be approximately 7 linear [ft] from the imprint roll. This reduced the phase margin loss to approximately 15 [deg]. which enabled the registration loop to be tuned for a significantly higher bandwidth (3 times), without the observed instability. This was due to the gain in phase margin obtained by placing the sensor closer to the actuator.
2. The feedback of the registration system was filtered with an analog 2nd order butterworth filter with a corner frequency of 30 [rad/sec]. The A/D card into which the filtered feedback was injected, was set with an anti-alias filter corner frequency of 30 [rad/sec]. The noise from the registration feedback was reduced to +/- 3 [counts].
3. The gear-box for the driven imprint roll was adjusted to remove as much backlash as possible from the drive train. The resulting reduction in backlash, combined with the low reflected inertia of the imprint roll, enabled the speed loop to be tuned for a higher bandwidth (from 17 [rad/sec] to 40 [rad/sec]). The higher speed loop badnwidth resulted in a significant gain in phase margin for the registration loop.
4. The feedforward terms in the registration controller were enabled and tuned to provide the maximum improvement in registration correction tracking, without exciting system resonances, and amplifying the existing system noise.

After the corrective actions were taken, the registration was improved to +/- 0.85 [mm] of set-point. After meeting with representatives of 3W graphics, and explaining to them the source of registration noise from the drying oven, while at the same time acknowledging that the registration deviation was greater than the agreed to contract specification, 3W Graphics agreed that the final product was more than acceptable, and waved the contractual registration specification requirement.

ApICS LLC would appreciate any feedback about the machine's performance.