Vision Sensors for Gaging
High-precision machine vision sensor technologies aid in-process inspection
By Patrick Waurzyniak
Machine vision sensors offer manufacturers highly accurate yet cost-effective methods to help boost quality on the factory floor with the latest vision sensors for in-process inspection systems.
Advances in machine-vision gaging systems include more powerful sensors and laser-based gaging systems. With the latest techniques, manufacturers can improve shop-floor measurement with systems featuring lower cost, greater durability, increased precision, and ease-of-use that can make machine-vision sensors for gaging a more viable alternative to contact gaging.
High-precision applications such as the automotive and aerospace industries demand in-process gaging systems capable of delivering the goods when manufacturers employ vision systems on the factory floor. Vision-based systems introduced recently by vision system suppliers like Cognex Corp. (Natick, MA) add rugged industrial vision sensors. Laser-based sensors from Omron Electronics LLC (Schaumburg, IL) use CCD technology similar to those of vision systems but instead employ a laser light source with very high precision for gaging applications.
Vision sensors for gaging usually hold cost advantages compared to PC-based vision systems. In some industrial factory applications, resolution requirements for sensor gaging also are not as high as those provided by higher-end PC-based vision systems that measure sub-micron-sized components for industries including semiconductors or nanotechnology.
Built for the factory, two recent additions to the Cognex In-Sight 5000 Series vision sensors, the 5400C and 5401 vision sensors, are designed as self-contained, industrial-grade sensors that meet IEC specifications for shock and vibration and achieve an IP67 rating for protecting against factory contaminants. With a DSP-based design, the In-Sight vision sensors utilize fast, inexpensive DSPs as a CPU instead of using more complex, expensive PC microprocessors. The sensors detect color, and the 5401 model features 1024 X 768-pixel resolution for applications requiring higher-accuracy gaging.
Machine vision sensors allow manufacturers to perform in-process inspection directly on a factory's assembly lines, giving vision sensor systems an advantage over PC-based vision systems located off the shop floor. "Depending on how and when the inspection takes place, we try to keep the user away from inspecting within the machine itself, in the machining center, just because there's mechanical apparatus flying around," notes Mark Sippel, Cognex's In-Sight vision sensors product manager. "There's obviously a lot of lubricating fluids being used. It's not hard on our product, although you have to keep the lens clean."
With an in-process gaging application, manufacturers could use a variety of automation equipment to move parts to the vision sensors, which can accurately inspect a part regardless of its fixturing or orientation on an assembly line. "Sometimes they're robotically fed, other times they're manually fed, depending on the type of machining center the user has, so it just depends on the level of automation a user might have," Sippel says. "Even if the user is pulling parts on and off the machine, at some point there's usually a point of inspection where the camera can be used, whether it's a conveyor they're feeding the part to or something else."
The Ethernet-based Cognex vision sensors can be configured remotely over the network, making it easy to deploy the technology. "Picture this as a vision system built into the camera--we call them vision sensors," Sippel adds. "The reason we call them sensors, as opposed to a vision system, is that generally the word 'system' refers to something that requires the combination of several parts, usually like a frame grabber separate from a camera. And that frame grabber board has to be put into a PCI slot in a computer."
Once it's configured, the vision sensor functions as a standalone device and doesn't need the PC. "Sometimes people leave them hooked up to the PC because it gives them the ability to see some images, and maybe transfer data, but it's not necessary," says Sippel. "The I/O is separate from the configuration or Ethernet port. It has the ability to be unplugged from the PC and just sit there and bang away with a PLC, or the CNC controllers, or whatever it's hooked up to. So it's literally meant to be a standalone unit on the shop floor.
With shock and vibration resistance designed into the system, the Cognex 5100 and 5400 In-Sight vision sensors are rated for 80 g of shock and 10 g of vibration. "It's designed to be rugged, to be on the shop floor," Sippel says. "People don't always think about g-force shock until the forklift comes by right next to it and drops a pallet of steel on the floor. The camera's designed to tolerate that; you won't hurt the camera in a normal environment."
The ruggedness of IP67-rated Cognex vision sensors enables manufacturers to deploy the system in very harsh factory environments. "The way we've designed our product, the camera is in an industrial enclosure and has a cover over the lens that protects it from any type of dust or grit," Sippel notes. "Even machining oil doesn't have any effect on our camera--it can sit in that environment and be very happy with it. Again, the only limitation is that you still must keep the lens clean."
Manufacturers keep the camera lens clean in a variety of ways, notes Sippel. "It's all in relation to how bad the environment is. They can mount the camera facing down, or use something like an air knife, so you can blow air across the lens and remove dust. With machine-tool oil or oil mist, you may need a regular maintenance schedule where you come in once a day and wipe down the lens cover. The bottom line is that the cover keeps fluid out of the lens; it's not in the mechanism of the lens, and it cannot access any of the electronics in the camera.
Software environments for PC-based vision systems also are typically more advanced and costly than those used by a vision sensor system, Sippel notes. "PC systems are going to have some programming capability. They work within computer environments, and are more embedded systems. With a vision sensor, everything is self-contained in the camera. You use a PC and its software simply to be able to talk to the camera, and interface to the sensor."
Aside from ruggedness, Cognex vision sensors offer advanced inspection tools with software in the camera. "Inside the camera the inspection software has several different tools that can be used on the image once the image is captured, to process or make the measurements or the gaging," Sippel says. Cognex's patented Pat-Max geometric pattern-matching software recently was added to the sensor lineup.
With the patented PatMax technology, sensors can perform high-accuracy feature location despite changes in the image angle, size, rotation, and appearance of parts on an assembly line. Cognex claims that PatMax provides 10 times greater accuracy than other pattern-matching technologies. This enhanced accuracy reportedly allows sensors to handle differences in part appearance caused by process variations, reflective surfaces, partial occlusion, nonlinear changes in lighting, or uneven image formation.
Costs for vision sensors typically start at about $3500, but adding options, such as a more expensive telecentric lens, can push that figure upward. Still, vision sensors tend to be much less expensive than PC-based systems that cost about $10,000 for a turnkey system that targets more high-end, high-accuracy applications.
Cognex vision sensors start at 640 X 480 resolution as standard, giving the camera the ability to see features as small as 0.0021" (0.0528 mm). Resolution isn't accuracy, of course. "In the long run, true accuracy has to be determined by running the application. Because environment has so much to do with accuracy, and repeatability, it's dangerous to throw something out there arbitrarily. My rule of thumb is to use three to five times the resolution; at three times, it would be roughly 0.0063" (0.159 mm). If the application's robust enough, accuracy could be considerably better. If the application is really sloppy, it could be worse."
Edge detection allows users to perform edge detection in what the vision industry calls sub-pixellation, Sippel notes. "By comparing one group of pixels to another, we can actually calculate less than a pixel. So with some of the tools, depending on the environment, you can go anywhere from one-quarter of a pixel down to as little as 1/10th of a pixel in resolution."
Such tools enable accurately calculating the width of a machined area on a part by detecting edges on the part. "This is not a physical measurement; you're not taking two edges of a caliper and placing it across the part, you're doing it visually. Therefore it's critical to accurately detect that edge," Sippel says. "That's why being able to sub-pixellate with these tools is really a significant advantage, because you can pick up a very tight tolerance on an edge."
With Cognex's PatMax software, In-Sight sensors can help edge detection work better on randomly oriented parts on an assembly line. "If you have a part coming down a line that's not fixtured and is dumped on the conveyor, its orientation or rotation could be random," Sippel says. "It is critical to identify the orientation of that part. The ability to accurately and repeatably do this in gaging is absolutely vital."
Laser-based gaging with sensors from Omron feature the same CCD technology found in vision systems, according to Reno Suffi, Omron's Sensor Group Manager. Omron also makes a wide range of vision sensors used in assembly applications, but Suffi says the units aren't used in gaging as widely as in assembly and other applications. "We do have some sensors that use CCD technology, which is the same technology you have in a vision camera, but it uses a laser light source as a point of reference, and those types of sensors are really used more in gaging, because of the precision involved."
For gaging, Omron's laser-based Z500 sensors enable higher precision than its more standard line of sensor products. "In a vision system, you have your light source and camera, and the light source bounces off the object and goes back and creates an image in the camera. With laser-based sensors, their own laser light source is like a highly controlled light element, a pinpoint beam with approximately a 10 µm spot diameter. That beam reflects off the object and hits a CCD that's looking for that type of light, whereas vision can be susceptible to ambient light and other conditions."
With its vision sensors, users are doing primarily pattern matching for surface defects, Suffi adds. "When we do a gaging application, it's generally much broader than with the laser system. We're not looking at 10 µm-level gaging inspection systems. When we use the laser sensor, we go into the CCD receiver, the laser-light-source sensor, so it's almost a cross between a sensor and the 2-D vision CCD element.
"It's a combination of what you would call a traditional sensor but with a CCD element, so it resembles a cross between a sensor and a vision system, but it's high precision. You can do an inspection, and the system gives you a gage or a measurement, the ability to measure within several microns, and it gives you height profile. The CCD element monitors the profile of the reflected light, so light comes down like a fan, like a very thin beam, and then it can profile that measurement."
Because of their high precision, aerospace manufacturers like Boeing Co. (Seattle) deploy the laser sensors to accurately gage the overlap on the wing skins of airplanes where the material overlaps slightly. "They have to ensure that the layers overlap each other," Suffi says, "and this sensor would gage the thickness and make sure the amount of overlap is correct."
This article was first published in the January 2005 edition of Manufacturing Engineering magazine.