SuperComp – Supercomputing

Semiconductors are found in nearly every modern electronic product, and their manufacturing is becoming increasingly complex. In the SuperComp collaborative project, partners from industry and academia—led by Infineon Technologies Dresden—are researching how AI and modern computing methods can make development, manufacturing, and logistics in the semiconductor industry more efficient.

HTW Dresden is leading the work package on digitalization concepts for strategic maintenance and production logistics. We are developing AI-supported computer vision methods that reliably identify and locate parts and components during maintenance work—for human specialists as well as, in the future, for robots. Because manufacturers often do not provide machine-readable design data, we digitize the equipment ourselves using 360° cameras, LiDAR scans, and the Spot mobile robot. This results in photorealistic 3D reconstructions. On our GPU cluster, even large scenes can be broken down into sub-scenes and processed in parallel.

From these digital representations, we generate synthetic training data for AI models and build digital twins. These enable the virtual planning and evaluation of maintenance windows and spare parts logistics. The results are published as open source and open access.