In its pressing department, the company employs a machine-vision system from Cognex Corp. (Natick, MA) to detect any tiny holes that may exist in the laser welds used to assemble bodywork components. Although these defects can measure as little as 0.3 millimeter in diameter, they can cause serious problems further down the line, including larger breakages and an unsightly paint finish. In both cases, it is often necessary to scrap the part. Fixing the holes is neither difficult nor expensive-assuming the holes are found in time.
In the past, weld checks were performed by employees, who would often struggle to handle the parts, which can be more than 6 feet long and 4 feet long, and weigh approximately 60 pounds. In addition, because the holes are difficult to see, many defective parts would get through. Clearly, the company needed something better.
To fix the problem, Renault installed a pair of test benches, each equipped with four In-Sight 1000 vision cameras from Cognex Corp. Each test bench includes a sturdy metal frame to hold the parts in place during testing. The cameras are positioned overhead on a crossbeam, and controlled via a touch screen fixed off to one side.
In operation, the body shell parts are fixtured on the test frame by a robot, and the part is lit from underneath. The cameras, which can detect a hole as small as 0.1 millimeter, then inspect a specific target area for light leakage.
If a part is defective, it is displayed on the screen, a red light comes on and the part is marked by a jet of ink, which stops it from being used until it is repaired. Each testing station can accommodate 900 parts per hour and is connected to a PC, which saves the photos of all the defective parts for a year, promoting traceability and allowing engineers to analyze process problems at their leisure.
According to Renault automated systems manager Patrice Dumont, it took some time to ensure the system would be able to detect all the possible defects that can result from the welding process, but it was worth the effort.
“We had to teach the system what faults were to be identified,” Dumont says. “That was done progressively, at the same time that we were familiarizing ourselves with the system. Once the system had memorized the type of fault, the recognition rate was 99.99 percent.”
For on machine vision, call 508-650-3000 or visitwww.cognex.com.