AI Assistance for Repair and Remanufacturing – Flexible Support for Technicians

In Detail

The EU project »RemaNet« is building a platform that connects remanufacturing stakeholders with companies looking to enter this field. The platform is also designed to help companies find the right contacts for repairing and improving products or tools.

AI Assistance for Manual Repair and Remanufacturing Processes

As part of the project, Fraunhofer IPA is developing two software tools to support manual repair and remanufacturing processes, as exemplified by the disassembly of electric induction motors. The first tool, »KARA« (Knowledge-based Action Recognition Assistant), guides the worker step by step through the disassembly and inspection process. The second tool, »AiRIEL« (Action Recognition Interactive Expert Learner), enables humans to train the artificial intelligence system and adapt it to different processes and product variants.

From Expert Knowledge to Flexible Activity Recognition

KARA supports untrained workers in the step-by-step disassembly of a motor using instructional videos and additional information. An AI activity recognition system, installed using a webcam mounted at the workstation, automatically tracks which steps have already been completed and highlights deviations from the learned process standard.

This process standard is data-driven and taught by an on-site expert using the AiRIEL tool. The expert’s activities are captured on video and imported into the tool. These recordings are labeled by a human to train the AI and teach it the process for the specific product variant. The trained AI can thus be generalized beyond specific product variants to different product groups, ensuring that KARA’s activity recognition remains broadly applicable and flexible at all times.