Intelligent cameras detect aluminum roll flaws
reparing aluminum plates for use in the printing industry is no small task. Rolls of raw material from 0.005 to 0.02 in. wide and from 26 to 60 in. in diameter must be prepared, coated with photosensitive coating, and cut to fit specific press sizes.
At the Plate Manufacturing Division of Eastman Kodak Co. (Windsor, CO), aluminum plates are prepared on web-inspection systems that sometimes run at 100 feet per minute. Often, during the coating process, defects due to scratches or foreign matter cause small uncoated areas of the aluminum. If this aluminum were used, printing images would not be formed correctly, thereby rendering poor-quality prints.
According to manufacturing-division engineer Kerry Mori, scratches and defects as small as 0.005 in. must be detected and tagged so that the coated aluminum is not cut into finished plates. To automatically inspect such defects, Mori turned to DNS (Crown Point, IN), a company specializing in vision-systems integration.
Building the system
In building the automated web-inspection system, DNS chose to incorporate four Fineline 2048 ¥ 1, 8-bit CCD cameras from Mayan Automation (Lachine, Que., Canada) into a PC-based system. Mayan Automation, now part of Cognex Corp. (Needham, MA), developed the camera using the 2105 digital-signal processor from Analog Devices Inc. (Norwood, MA) as the on-board processor and the i960 processor from Intel Corp. (Santa Clara, CA) as a communications engine to the CANbus.
"We chose the Fineline cameras because they have a built-in CCD, digital signal processor, memory, and digital I/O controller," says Victor Binkauskas, DNS president. "Although Mayan Automation did provide camera-resident software, we found that significant modifications were required," adds Binkauskas. Consequently, DNS purchased the source code from Mayan Automation and developed its own camera-resident software.
In the system setup DNS installed at Kodak, the company used four such cameras synchronized over the CANbus. Digital signals from each camera are clocked to a specialized interface that, in turn, controls a programmable logic controller. After each defect is detected, the controller triggers a labeling system from Ventura Labeling (Ventura, CA), which places a 3/4-in. paper-thin tag at the center of the aluminum roll. In this way, each imperceptible flaw in the aluminum coating is easily identified by human operators who must cut the finished aluminum.
"Prior to automating this process," says Mori, "there was no way to quantify any defects in the aluminum coating. Now, because such defects can be tagged, we can better manage the inventory and production process while improving the quality of the finished product."
--ANDREW WILSON
Four programmable linescan cameras linked via the CANbus are used to inspect rolls of coated aluminum plates at Eastman Kodak for tiny scratches and dirt particles.
