FORD CANADA - Ford engineers are using 3D mapping in what is believed to be the technology's first application in the auto industry to better analyze rear axle parts of F-Series pickups during assembly, leading to a smoother, quieter ride for drivers of Ford trucks.
Similar to Google Earth, which uses three-dimensional imaging to map the world, Ford's photogrammetric pattern reader (PPR) uses a pair of digital cameras to combine photos of the axle gears into a series of 3D pictures that are compared against an ideal computer model of the gears. Gears that don't meet "Built Ford Tough" specifications are discarded.
"PPR technology is the next evolution in quality control for our commercial trucks," said David Gravel, an engineer in Ford's advanced manufacturing group. "While traditional, visual inspections of our axle gears ensure we're delivering dependable, tough trucks to our customers, this new technology allows us to conduct our inspections faster, and at a level of detail the human eye just cannot discern."
"We constantly look for ways to go further for our customers," he added, "and this system is one example of how we are making our trucks even more reliable."
Ford's system uses line scan cameras and infrared lights to turn a series of two-dimensional image slices into a single three-dimensional image for analysis. It was developed with Madison, Wis.-based Automated Vision and ATM Automation in Livonia, Mich.
"This technology is part of a trend where companies like Ford are using advanced automation to increase accuracy and consistency in production," said Nan Zhang, a scientist at Automated Vision. "Computer vision is booming and is a very important topic for the next decade."
Now in use at Ford's Sterling Axle Plant in Sterling Heights, Mich., PPR stations are being rolled out in Ford plants worldwide.
How does it work?
The complex curvature of the hypoid gears used on the rear axle ring and pinion makes it impossible to see both sides of each gear tooth in a single image. The two cameras used on the inspection rig capture 9,000 1024x1-pixel images from each side of every gear tooth in a matter of seconds as the gears rotate. The processing system contains a digital model of the gear profile that is then used to stitch these images together and flatten out the gear teeth into a single 3D panoramic image that can easily be scanned to ensure the teeth are meshing correctly.
Any parts that show anomalies that could cause noise or durability issues are scrapped. On average, only two to five parts per every 1,000 parts fall outside the tight tolerance range for acceptability.