Cloud-based thermal error compensation with a federated learning approach

Thermal error compensation is one of the most research-oriented topics in manufacturing with rising importance industry. This paper presents an innovative Industry 4.0 application thermal for precision engineering. A federated learning-based approach running cloud applied to two machine tools, located at ETH Zürich, and another TU Wien. Although environmental conditions behaviour both machines differ, implemented knowledge transfer across a viable strategy, albeit limited precision. detailed comparison same type under load shows foreseeable similarities behaviour, but also clear differences due different configurations lifetime status. The cloud-based reduced crucial errors best case tools by more than 80% heavier conditions. • new introduced. Federated learning terms edge computing reaching required high level safety security presented compensation. Detailed analysis manufacturer tool wear slightly configuration. Proof concept transferability model calibration period (learning phase) reached (transfer learning). Compensation results already reduce up 80%.