Fleet learning of thermal error compensation in machine tools

Thermal error compensation of machine tools promotes sustainable production. The thermal adaptive learning control (TALC) and machine learning approaches are the required enabling principals. Fleet learnings are key resources to develop sustainable machine tool fleets in terms of thermally induced machine tool error. The target is to integrate each machine tool of the fleet in a learning network. Federated learning with a central cloud server and dedicated edge computing on the one hand keeps the independence of each individual machine tool high and on the other hand leverages the learning of the entire fleet. The outlined concept is based on the TALC, combined with a machine agnostic and machine specific characterization and communication. The proposed system is validated with environmental measurements for two machine tools of the same type, one situated at ETH Zurich and the other one at TU Wien.