Grand Rapids, MI —While many manufacturers rely on end-of-line inspection or lot inspection to assure quality, some companies are making the transition to machine vision systems.
One such company is Monroe Inc., a manufacturer of approximately 60 million automotive speedometer needles per year. With such a high-speed operation, Monroe couldn't afford to continue doing quality inspections downstream from production.
"We had to come up with a way in real time [to inspect product] because the screening people are downstream and catch it in delay. You can build up a pile of scrap [in a short amount of time]. We can sort 100% at the end of line, but that doesn't catch it in real time," reports Mike DeBat, manufacturing engineer for Monroe. "We were operating at 98-99% (good parts), but our customers were telling us that that wasn't good enough. Today [with machine vision] we're at 99.99%—in the neighborhood of 50 defective parts per million."
Machine vision systems are big business. According to a study to be released later this year from the Automated Imaging Association (AIA), total sales of imaging systems worldwide in 2000 exceeded $6 billion, with North American accounting for over $4 billion.
In the past, vision systems tended to be PC-based, very powerful, and very expensive. Five years ago a typical PC system could easily cost more than $25,000. But in recent years, self-contained microprocessor-based vision systems have been developed which can do approximately 80% of the things that the PC-based systems can do, for less than half the price. For Monroe, the switch from PC-based to Omron Electronics' (Schaumburg, IL) microprocessor system was obvious. "The PC-based was overkill for what we wanted to do. Price was also a consideration—Omron was about $5,000-$6,000 for a complete system; PC-based was roughly $20,000," says Debat.
But how accurate is computerized quality control? Monroe places Omron vision systems immediately downstream from the various production stages to catch defects before they are packaged. Then robotic sorters pick and place good and bad parts in the appropriate containers. Here's what they're looking for:
Absence of a counterweight that balances the needle
Absence of a pair of heat stakes that hold the needle to a mounting bushing
Improper heat stake size
Needles of incorrect color
Improper mounting hole size
Pinholes, dirt, scratches
Absence of decorative hot-stamp foil
Nicks and tears in the foil
Improper location of the foil
To detect the first six defects, Monroe uses Omron F-150 black-and-white systems with dual cameras. Dual cameras are necessary because Monroe often runs two assembly lines side by side to meet their production needs. These systems are capable of doing dimensional inspections in the 0.005–0.010-inch range. Notice that a black-and-white system is used to do gross color sorting of needle color. For cosmetic defects related to the hot-stamp foil, Monroe uses Omron's F-400 color system. This 512- x 484-pixel system is capable of extracting up to eight colors per scene.
Despite their high accuracy and low cost, vision system manufacturers still must address stereotypes. "For people who are on the fence and say, 'We can't afford this,' we have to reeducate them. That's been an ongoing process," says Jon Wright, controls engineer at Omron. "What customers have found is that we can do with a system under $10,000 what they were previously doing with a PC-based system which costs considerably more." Another stereotype is the reliability of human operators to catch defects. According to a study by Omron that compared the human eye and the vision system's abilities to catch a defect the first time on a molded part, the human eye was only 50% reliable in catching pin holes and flow lines compared to the vision system's 100% reliability. Wright sums it up with an Omron motto—"For the machine, the work of the machine; the man, the thrill of the development."
For more information about vision systems from Omron: Enter 541