3D point cloud skeleton-based method for bridge girder virtual trial assembly

Abstract

The growing adoption of large-scale prefabricated girders in bridge construction, driven by technological advancements and accompanied by increasing bridge spans, heavier components, and more complex construction environments, has heightened quality control demands, leading to the widespread application of 3D point cloud-based virtual trial assembly (VTA). However, current VTA research primarily focuses on component assemblability, whereas studies of assembly alignment remain scarce, despite its direct impact on construction quality and the potential for accumulated deviations to cause stress concentrations that undermine long-term performance. This study proposes a 3D point cloud skeleton-based method to evaluate assembly alignment of box girders, offering quantitative insight into this critical performance factor. Specifically, multi-resolution point cloud downsampling is first applied to enhance computational efficiency while considering the missing top and bottom surfaces typically encountered in point clouds collected in beam yards. Then, key geometric features are extracted from individual slices and integrated across slices to construct the 3D point cloud skeleton. Finally, assembly pose adjustment is performed, and alignment is assessed by calculating the clearance at the assembly interface. The proposed method was validated through two world-class sea-crossing bridge projects in China, covering both steel and concrete box girders, demonstrating broad applicability. For the example girders studied, 97.56% and 60.50% of interface valid cells met tolerance requirements for the steel and concrete girders, respectively, with higher percentages reflecting better feasibility of achieving the designed assembly alignment.