On its premises, the partners have taught the industrial robots how to inspect the quality of processes in component production of electric drive systems used in BMW’s e-vehicles. More precisely, the project partners installed a lightweight robot equipped with a camera in the production line of electric engines. Then they taught the robot how to pan, tilt, and focus the camera to inspect the previously completed production steps on the electric engine from the decisive angles. In detail, the system learned how to inspect not only the applied codes and serial numbers, but also the previously pulled silicone beads. In addition, the robot also had to be able to sort out the deci-sive pictures from an abundance of photos taken during the process and to identify defects on the silicone beads.

Experienced BMW employees helped with the interpretation and marking (“labeling”) of the image data so that it was possible for Robotron to generate the respective models for artificial intelligence (AI) on this basis with its “Realtime Computer Vision” (RCV) platform. The cloud resources for this AI were provided by Microsoft. The sensor pens (“TracePens”) with which the robot was taught the right movements came from Wandelbots. In addition, interfaces were created between the diverse systems of Robotron, Microsoft, and BMW.

According to Wandelbots, the use of these new teaching technologies has shortened the time a robot needs so that it is ready to conduct the quality controls by the factor 15 to less than 60 hours when compared to conventional solutions. BMW, thus, saved 30 percent of the costs that are usually in-curred. The advantages for quality control are obvious. And just one pilot application won’t be the last one: “It was also examined whether and where this technology can also be used in other places at BMW, and this resulted in identifying about 250 possible examples,” reports Dr. Patrick Grosa, Head of Business Acceleration at Wandelbots.