It's often best to seek outside help when adding vision technology for superhigh production of packaged goods.
Say one thing for industrial vision systems: Despite new technology that simplifies their deployment, setting them up is still no job for amateurs.
"More than any other type of automation, vision-system design requires deep, multidisciplinary knowledge," says Richard Hebel, a vice president at Insight Control Systems International Inc. (ICS), Safety Harbor, Fla. "Putting together turnkey vision solutions is not simple. Any given application requires combinations of electrical, electronic, optical, illumination, software, image and signal processing, statistics, machine design, control, and packaging disciplines. It's a tall order, and that's why the industry has segmented into companies that specialize in various end markets.
"Can users set up their own vision systems?" Hebel asks. "Yes. Should they? I guess that depends on whether the company considers the expertise to do so a crucial strategic capability. Some vendors would have you believe you can buy various vision components and snap them together like Erector sets. That may be true for sophisticated packaging engineers who have a lot of vision system experience, but it won't be the case with most companies."
Lighting and optics remain one of the most difficult aspects of setting up vision systems, admits Darren Hench, in the Cleveland office of Coreco Imaging, headquartered in Quebec. "It's tough to overcome the obstacle of ambient light," he says. "If you can't see a product, naturally you can't detect its defects."
Vision system manufacturer Applied Vision LLC in Akron, Ohio, has run into such problems especially when inspecting metal products such as cans.
"While the highly reflective surfaces of aluminum beverage cans are eye-appealing, inspecting irregularly-shaped mirrored surfaces at rates of 2,700 cans per minute is no simple task," says Todd Bailey, director of sales at Applied Vision. "Proper lighting is essential. Illumination systems must detect tiny flaws without generating glare against the product and without casting shadows that might conceal defects."
Navigating such difficulties makes the process of setting up a vision system more involved than just purchasing parts. "While you might buy a smart camera for $7,000, some lighting for $4,000, and a PLC for $2,000 and think you have a $13,000 vision system, you've actually only just started," Hebel says. "You need to include software and figure in costs for building a hardware framework, and you need to allow for test-equipment cost. Ditto for your time and that of your engineers. And what if you finally get it all put together and it doesn't quite do what you need? A job change takes an engineer two days, and that drives production managers crazy."
Packaged goods that are particularly good candidates for vision are those where one product imperfection can end up ruining a whole shipment, observes Bud Patel, director of marketing and business development for Applied Vision. Acidic soft-drink liquid from a leaky can, for example, may end up destroying stacks of pallets. Other prime applications are those where a broken seal can earn heavy fines for a manufacturer.
Over the last five years, packaging quality, particularly within the food industry, has become a huge issue, Hebel says. Branding in consumer marketing is increasing quality requirements.
"Things are so uniform within today's market that people avoid anything that stands out," Hebel explains. "They won't purchase a dented can or a crushed box. Margins are so tight in retail that manufacturers can't afford any type of loss; they can't afford to distribute damaged goods. Until the mid-90s, overseas manufacturers used manual inspection, but as the use of vision systems pushed quality standards up on the world market, foreign companies realized the quality that labor produces is more important than the cost of that labor, and so many companies are switching to vision systems."
Humans versus machines
Industry veterans note that the original inspection systems had a reputation for being unreliable. Improvements in lighting and imageprocessing techniques over the past decade have turned that around. But many manufacturers still equate industrial vision with replacing human inspectors, an image that is not entirely accurate.
Today machine vision may be installed to improve product traceability and send back production intelligence to factory-floor data-collection networks. And besides, "Human inspectors also clear jammed lines, operate and fix machinery, and record production data," Hebel says. "You can't look at vision simplistically as a human inspector replacement."
The human brain also distinguishes things that cameras can't, points out Coreco Imaging's Darren Hench. For example, humans perceive objects in three dimensions while machines still largely look at products in 2D.
No question, though, inspection via vision systems can relieve potential bottlenecks that would arise from manual methods. Hebel recalls a visit to a bottling plant on the East Coast where workers were inspecting bottles at the prodigious rate of four per second.
"Aside from the question of whether the workers could actually do anything useful at that speed, their heads were wagging back and forth and they had facial expressions of folks who are totally stoned," Hebel remembers. "Maybe they were. How else could a person sit and watch bottles for eight hours a day? That company bought one of our vision systems shortly afterward," he adds.
In contrast, vision systems keep up with the fastest manufacturing lines. According to Hench, the average vision system can handle 1,800 to 2,000 parts or more per minute with a 99.5% accuracy.
Adds Applied Vision's Todd Bailey, "Cap-making machinery for various beverage containers turn out 3,000 caps per minute. Real-time operating systems, inspection algorithms optimized for high speed, and specialized high frame-rate video cameras let inspection systems handle these speeds."
But, "Unfortunately, there is a lack of offthe-shelf hardware able to track parts at these speeds," Patel says. "Systems that do have such tracking devices knock defective products off conveyors before they reach reject stations by using sharp bursts of air."
Not just for the speedy
Not all advances in vision technology target production lines clipping along at breakneck speeds. Camera technology also has progressed in ways that simplify installations on slower lines.
Developments in this category include the "smart-camera" concept. These are cameras with enough built-in electronics to take over some data-processing tasks. Users develop inspection strategies while the camera is attached to a PC, then download the routines to the camera. Multiple cameras can be set up this way, but camera-processing power is still quite limited. So smart cameras are appropriate only where inspections need not be superfast or complex.
"For simple vision tasks, smart cameras offer lower price points," Hebel states. "Ten years ago you needed $70,000 systems to handle the same tasks. Today, these products sell for as little as $3,000. Of course," he adds, "these aren't complete systems. Users must add lights, control, and handling, which increase installation costs."
All in all, self-contained microprocessorbased vision systems can do about 80% of the tasks handled with PC-based systems, and for less than half the price.
According to Hench, the average price for major players today is between $25,000 and $50,000 depending on the level of integration required. Indeed, integration can be a sticky issue.
Hebel states, "At ICS we have systems that range from $15,000 to nearly $200,000. However, integration costs are often overlooked and underestimated."