Mechanism-enabled digital twin system for thermal error prediction and compensation in the feed system

As a critical subsystem of machine tools, the feed system is crucial to the determination of machining precision, with its thermal deformation directly impacting the tool center point (TCP) error. However, owing to the non-uniform spatial distribution of heat sources and the structural complexity inherent to the feed system, acquiring complete temperature field information remains a significant challenge. Furthermore, existing thermal error prediction methods fail to adequately capture the coupled deformation behavior among structural components, thereby limiting the models’ generalizability and robustness. This study proposes a mechanism-enabled digital twin system to address the challenges of thermal error prediction and compensation in feed systems. The system architecture comprises three layers: the physical layer (PL), the mechanism-enabled layer (MEL), and the virtual interaction layer (VIL). The PL collects information on the thermal characteristics of the feed system. The MEL develops a thermal–mechanical integrated model (TMIM) to reveal the mechanisms driving temperature distribution and thermal deformation, while also estimating unmeasurable temperature data from the screw. This model provides theoretical support for sensor deployment, measurement scheme design, and feature selection in thermal error modeling. The VIL governs information flow and TCP error computation, and integrates a mechanism-informed multi-output broad learning system (MI-MO-BLS) to determine the compensation parameters. Experimental validation demonstrates that the proposed system achieves a thermal error prediction accuracy of 94.27 % and improves machining accuracy by over 64 %, validating its effectiveness and practical applicability in enhancing the precision of machine tools.