"The aim of the project is clear: to move away from time-consuming and expensive programming and towards simply configuring a task for the machine," says Christian Eitzinger. He heads the Machine Vision department at Profactor. The company has experience with both industrial image processing for quality control and robotics. In recent years, Profactor has dedicated itself to linking inspection and robotics. This includes the Thermobot project, in which crankshafts, for example, are inspected robotically. The challenge with components with complex geometries is path planning for the robot. It must guide the sensor in such a way that the entire surface of the component is detected.

Path planning is also highly dependent on the sensor technology used. Eitzinger: "A time delay is essential for heat flow thermography, but this is irrelevant for X-ray technologies, where the requirements for the sensor-guiding robot are completely different to those for cameras used for surface inspection, for example."

Surface inspection, crack detection, completeness check
The Spirit project aims to develop a robot that can be equipped with different inspection systems: Cameras for surface inspection, thermal imaging cameras for crack inspections, X-ray sensors, 3D sensors for completeness checks. Eitzinger: "We are developing the software that enables the robot to automatically plan the path for each of these different inspection tasks." The system is simply fed with the CAD data of the work cell - to avoid collisions - and the CAD model of the component and automatically calculates the path for the robot depending on the sensors and inspection task."

At industry partner Centro Ricerche Fiat, the component is even allowed to move during the test. In this specific case, it is an engine on the assembly line, where a 3D completeness check verifies that plugs, cables and hoses have been assembled correctly and completely.

In addition, the system should be able to react to necessary optimizations in real time using reactive path planning. This is the case if the CAD model deviates from the actual component - due to deformations, for example. This also includes the optimization of the sensor position - for example, in the X-ray inspection of honeycomb structures or similar inside a component.

Although the industrial partners in the research project are well-known corporations, the system is ultimately intended to benefit those companies - including small and medium-sized ones - that cannot afford automation due to small batch sizes and the often expensive programming of robots. as