Abstract:In the digital transformation of next-generation aircraft assembly, digital-physical assembly plays a critical role. Rapid, high-precision reconstruction of key assembly features is central to accurately predicting on-site assembly outcomes. However, current methods for reconstructing complex hyperbolic ring surfaces face limitations in both accuracy and efficiency. In this study, we focus on point cloud data from the mating surface of an intake duct. First, a radius-based filter is applied to remove obvious noise points, and region-growing techniques are used to determine upper and lower thresholds for passthrough filtering, resulting in the target point cloud for reconstruction. Next, the point cloud is classified by importance, and voxel downsampling based on centroid selection is performed using grids of varying sizes according to priority. Then, segmentation is performed based on direction angle ranges determined through region growing using a normal vector threshold. Subsequently, surface fitting is conducted via rapid NURBS base surface construction and iterative optimization. Finally, the fitted annular mesh surface is merged using a zipper-based stitching approach. To validate the method, a mock intake duct and an annular metal frame were fabricated, and point cloud data of the mating surfaces were collected. Comparative experiments on fitting accuracy and efficiency demonstrated that this method, when compared to approaches in literature and commercial software, effectively reconstructs complex hyperbolic annular surfaces with high efficiency while maintaining fitting accuracy, meeting the high-fidelity requirements for intake duct fitting and assembly simulation.
