Automatic detection of production faults using an active vision system

Challenge: Flexible quality control for customised production

While mass production used to be the norm, the emphasis in the industry is increasingly on mass customisation, with small series of customised products - adapted to specific customer requirements. Naturally, the customer also expects the highest quality, with quality control being of paramount importance.

Human quality control is flexible, but it is difficult to maintain the same quality standard. This is because different people make different assessments. Automation with robotic arms equipped with cameras can bring relief, but has shortcomings at the moment. This is because you still need two human experts: a robot expert to programme the robot and a vision expert to instruct the robot on where and how to check for errors.

Solution: Improved automated quality control

Therefore, we developed an improved automated quality control - using an algorithm as a guide and without the help of those two experts. This makes it possible to further streamline production. We validated this innovation together with technology company Atlas Copco, known among other things for its air compressors.

The challenge we faced is that the optimal inspection setup of the cameras (and light sources) can depend heavily on the 3D shape of the product. As a solution, we developed a tool that allows inspection coverage to be determined based on a CAD file, the camera and light setup and defect detection rules. An active vision system that automatically configures itself to quickly scan defect zones on a part with known but complex geometry.

The tool works in three steps. Defects are first analysed and the best inspection configurations translated into so-called detection rules. Then an optimisation algorithm examines the possible locations for the cameras (and light sources) and selects the ideal combination. Finally, a robotic path planning tool determines the optimal path between the inspection points. In the span of just a few minutes, this tool provides extremely detailed advice.

Result: Atlas Copco implements test rig in production

The innovation supports Atlas Copco in detecting defects on different types of air ends for compressors. The company already recently installed cells with a cost-effective setup - with three cameras and an edge device - for quality control. It is now working on an integrated setup, where the cameras and the edge device are installed on the machine itself, which will further bring down costs. This solution could then be widely used in production. With the help of our tool, the company can determine where the cameras are best placed, to accurately and reliably check the largest possible area for errors.

Want to know more?

Would you like more information on this or other cases? Then get in touch or check out the other cases on the website.

Kristof Vrancken, Communication Officer

Kristof Vrancken is Digital Communication Officer at Flanders Make since 2019. As Digital Marketeer with experience in both B2B and B2C environments he writes with a fresh view on technological innovation, about what literally and figuratively moves within our research centre.