AI-Based Component Control For Faster Changeover Times
Checking components on the go and in real time – this is made possible by the MARQUIS software from Fraunhofer IGD. Users in the automotive sector benefit from short changeover times, user-friendly guidance and reduced costs as a result of less downtime in the production lines. Researchers presented the software for mobile quality testing at the recent CONTROL exhibition.
A missing screw, a displaced strut or a wrench lying around: There are many sources of error on production lines in the automotive sector. Efficient and user-friendly test systems are all the more important. The MARQUIS software combines augmented reality with machine learning methods – and thus enables automated testing in real time.
“In order to train the artificial intelligence and to mark test cases, the software does not need any real photos – it uses synthetically generated images from the CAD model,” explains Holger Graf, department head at the Fraunhofer Institute for Computer Graphics Research IGD. The software then compares the CAD specifications with the existing components. The system trained in this way recognizes these without ever having seen real structures beforehand. MARQUIS is characterized by its full automation: With this approach, the lead time is reduced from one to two weeks to just a few days. Users in the automotive sector benefit from faster changeover times and correspondingly more drop tests per day. Corrections can be made directly on the production line without stopping it. This reduces costly downtimes,
Flexible Component Control In The Production Process
MARQUIS can be used not only at stationary test centers, but also on the move. Fitters can use smartphones, tablets or AR glasses to check the components in the dynamic process. “The system recognizes which component it is, how it is positioned in space and whether it deviates from the target dimensions,” says Graf. A visual feedback shows user-friendly when the software detects errors in the structure: A component then lights up red. The AI-based software also recognizes deformable objects such as hoses.
With the latest generation, it is even possible to connect experts remotely. These can remotely support the person on site with the control. In addition to testing on the production line, automobile manufacturers also benefit from the software in incoming and outgoing goods inspection. Where previously spot checks were the norm, each individual component can now be scrutinized on a test bench.
The Fraunhofer researchers are working closely with the spin-off Visometry to develop the prototype into a market-ready product.
For more information: www.igd.fraunhofer.de
