A Narrow View No More

Reprinted with permission from Managing Automation , March 2002

Machine vision systems come down in price, size, and complexity, and join with robots for advanced factory applications.

At the DaimlerChrysler factory in Twinsburg, Ohio, the loading of car and SUV doors onto the production line to be assembled into finished vehicles used to cause innumerable safety and efficiency problems.

For one thing, people had to move these large pieces of razor-sharp metal off the racks to put them onto the line. Accidents were common. For another, although the system demanded that a new door be loaded onto the line every 15 seconds, that wasn’t always possible since it was a manual (and back-breaking) process. And the racks that stored the parts had to be re-tooled every year, which meant huge costs. In fact, this process was the Achilles heel of the entire production line. The situation was primed for automation, thought Brad Dailey, assembly project and tooling manager for DaimlerChrysler AG (Auburn Hills, MI).

One clear solution was to use robots to unload the treacherous doors from the racks. But that had proved impossible, says Dailey, since the size and position of the doors are too variable for standard robots to cope with. The answer: a combination robot/machine vision application based on technology from Cognex Corp. (Natick, MA), Nachi Robotic Systems Inc. (Novi, MI), and Shafi Inc. (Brighton, MI).

‘‘I thought we could improve the safety issue and improve our productivity by using a vision application,’‘ says Dailey. However, this type of application had never been done before in the auto industry. So, he enlisted the help of Adil Shafi of Shafi Inc. to provide the software to knit together the Cognex vision system and the Nachi robot.

Together, they created and rolled out the system in six weeks last summer. ‘‘There are a multitude of benefits. We’ve had great improvements in safety and productivity. We can bring in six door lines where we could only do three before,’‘ says Dailey with understandable enthusiasm. The project cost $300,000 to put in place; Dailey and his team expect cost savings to be $440,000 annually.

Bearing little resemblance to their large, unwieldy, and exorbitantly priced forbearers, machine vision systems are coming into their own. Today’s models are smaller, much cheaper (about $7,000 today for one system installed versus anywhere from $25,000 to $100,000 about six years ago), and much easier to use. Some vision systems have color capabilities, enabling them to identify presence or absence of particular colors. Many have Ethernet networking capabilities built in so the cameras can share data with other controllers on the plant floor, as well as pass information up to enterprise IT systems.

And machine vision applications have expanded. In addition to robot guidance (such as DaimlerChrysler’s application), there are inspection systems that check for presence/absence of certain characteristics, pass/fail of specified tests, and perform optical character recognition. Then there are gauging vision systems, which perform sophisticated measurements. These are not your father’s machine vision systems.

‘‘They’ve gone from these large, very expensive products and migrated down to a smaller footprint and smaller price. It’s like packaging the power of a mainframe computer in a PDA,’‘ says Mark Sippel, vision product marketing manager for Omron Electronics LLC (Schaumburg, IL). The market has responded favorably, growing 26% last year, according to the Automated Imaging Association (AIA, Ann Arbor, MI).

The total worldwide vision system market reached $6.2 billion in 2000 (the last year for which numbers were available), according to the AIA. That estimate includes cameras, lenses, optical equipment, software, frame grabbers, lighting, power supplies—every component that is needed in a vision system. (It does not include robots such as those that DaimlerChrysler used.)

Michael Lewis can attest to how much machine vision has changed in just the last few years. Lewis, president of Automation Technologies Inc. (Escondido, CA), has been building custom automated assembly equipment for the past 16 years. ‘‘We used vision back when it was ugly to use,’‘ he says. Vision was worth doing even when it cost at least $25,000 per system because it is such a flexible technology. ‘‘Now it’s getting down to the price where you can reasonably use it. You can implement one of these systems for about $7,500 including programming and lighting,’‘ he says.

‘‘The price points are good. And the users can configure the systems themselves,’‘ confirms Dick Slansky, senior analyst at ARC Advisory Group Inc. (Dedham, MA). Slansky warns that most manufacturers need major systems integration help before putting a vision system together. ‘‘The average manufacturer does not have the expertise to put one of these together, especially if it involves integrating robots and vision sensors,’‘ he says.

Automation Technologies uses vision systems from DVT Corp. (Norcross, GA) to build equipment that makes smart cards the Chinese government will use to identify its 20-million-plus-person population. ‘‘Since they have a chip embedded in them, these cards store a lot more information than your average magnetic strip card,’‘ says Lewis. He has just installed 30 units for the project and expects to install a total of 80 by the end of the year.

The DVT inspection system checks the soundness of the chip on a variety of parameters and then checks its position. ‘‘In order to know where to put it, we identify the position of the chip to the bonding head. We send that position to a robot, which moves it into place,’‘ says Lewis. The combination robot/machine vision application has reduced the amount of scrap by 10%.

Machine vision is especially useful in perfecting execution. ‘‘We can watch a part and see when something is starting to move out of spec. The system can transfer information to another device down the line and make adjustments without having to bring the whole line down,’‘ explains Phil Heil, chief technology officer for DVT.

Although many systems (including those from Cognex and DVT) now come standard with Ethernet networking capability, that feature may be a bit ahead of the curve. ‘‘Our sensors can become a direct part of the LAN. An engineer can query a system from their desk onto the factory floor. But we’re running a little ahead of where most people feel comfortable,’‘ says George Blackwell, director of product marketing for Cognex. Blackwell notes most customers have not yet networked their vision systems.

Like most current vision customers, DaimlerChrysler and Automation Technologies have not yet networked their vision systems, although both companies are interested for the future. ‘‘With the DVT camera, you can plug it in to Ethernet and give it an IP address. I could look at it from across the world if I wanted to. We could see real-time pictures come across here. That would be very convenient,’‘ says Lewis.

As a long-time vision user, Lewis is amazed at how far the systems have come in a few short years. Says Lewis, ‘‘The speed at which these systems are advancing is breathtaking to me.’‘

*Reprinted from Managing Automation March/2002 © 2002 Thomas Publishing Company