Imagine an inspection machine that can catch defects as small as 0.3 mm on parts–like cylinder heads–moving down an assembly line. A non-contact machine that can measure flatness, porosity, and the size and location of holes, and check for defects all within 60 seconds.
This machine is real. Sort of. It's the University of Michigan's Reconfigurable Inspection Machine (RIM). Honda has expressed an interest in RIM. So has GM. But neither company can have one–unless an equipment supplier starts manufacturing RIMs. The folks at U of M develop technologies for new devices. They don't produce capital equipment. However, they have a good sense of what the capital equipment needs to do. Says Patrick Spicer, a research fellow at the school, "The machines are modular and designed for specific part families like cylinder heads. By narrowing the scope of flexibility in this way, we've created a machine that is economical, efficiently does the inspection, and can be reconfigured to handle any changes that are made within the family of parts."
The system cuts the part into small slices by running it past a mask that has a vertical slit approximately ¼-in. wide cut into it. This ensures the laser portion of the RIM concentrates its attention within this area, and examines the part and not the surrounding area. The optical unit takes a picture of the part as it travels past, and takes a 2D picture of the part's surface. Software looks for defects on the surface, measures the location and diameter of holes, and compares this to a model of an ideal part. Concurrently, the laser concentrates its beams within the open area of the mask and measures 750 points per second, allowing the unit to scan several thousand data points as the part passes the station. These images are then stitched together into a seamless picture nearly 200 megabytes in size.
"It's a very large file and it must be analyzed quickly," says Spicer, "and the result is displayed in a low-resolution format on the oper-ator's screen." If the operator wants to inspect a particular feature, the full-resolution image is used. Spicer expects that, if the manu-facturer wants to retain a file of potentially defective parts, "chips" corresponding to areas where items are found on the surface would be archived along with a low-resolution image of the entire part.
There are some system sensitivities that must be taken into account. As with all vision-based inspection systems, it's possible that polished surfaces will distract the sensors and lead to unreliable results without proper lighting. One solution that is commonly used is spraying the part with a powder that cuts the glare. Unfortunately, the powder also can add as much as 100 microns to the surface of the part, which affects the flatness measurement. "If your flatness spec is 100 microns across the entire part," says Spicer, "the powder could throw off the measurement enough to cause you to scrap or rework the part." So the RIM's laser sensing system and on-board lighting are engineered to help eliminate this problem.
Spicer's group also is working to give the RIM expert knowledge such that it can determine when a scratch, discoloration, or slight porosity is important, or can be ignored. By indicating those areas where adherence to spec is critical and those where it is not, he hopes to help OEMs and suppliers reduce waste without adversely affecting quality. "We're not sure we can give the industry everything it's asking for, but we're attempting to push the boundaries," says Spicer. Part of that effort is focused on gaining access to an OEM niche engine line to refine the system and create a production-capable unit. "The folks in engine plants get a little leery when you come to them with a technology that doesn't touch the part," says Spicer, "but I'm pretty certain the potential cost savings will be more than enough to overcome their fears."