A digital twin flanges assembly system for large spherical vessel integrating function blocks and cognitive augmentation

The assembly of flanges on large spherical vessels is challenged by long-cycle operations, thermally induced geometric drift, fragmented workflows, and heavy reliance on manual interpretation of numerical data. To address these limitations, this study proposes a digital twin flange assembly system (DTFAS) integrating function blocks (FBs) and cognitive augmentation. The system establishes a unified framework that couples measurement, pose optimization, and AR-guided execution through a human-in-the-loop closed-loop mechanism. Specifically, measurement data are first registered within a global coordinate system with temperature compensation based on sensor-acquired thermal data and a deformation model. The processed data are then input into a multi-objective pose optimization module, whose results are transformed into intuitive AR guidance via a physics-informed mapping, where geometric deviations between the measured and theoretical flange poses are directly visualized as spatial adjustment vectors and alignment cues, enabling intuitive and physically meaningful human operation. Within this framework, key operations are encapsulated into interoperable FBs, enabling sequential data flow and iterative feedback among measurement, optimization, and execution. This integration resolves the critical barrier of disconnection between multi-source data processing and on-site assembly actions under dynamic environmental conditions. In the experimental validation, the MR-FB decreases the total 3D registration error to 0.039 mm. The PO-FB achieves an eccentric deviation of 1.1667 mm, corresponding to a reduction of 54.2% in overall assembly error compared with the conventional process. In addition, the AR-assisted workflow reduces the total assembly time from 37.0 min to 17.1 min. These results indicate that the proposed system improves measurement consistency, enhances pose determination accuracy, and streamlines execution efficiency under the studied assembly conditions.