From geometric inspection to data-driven assembly decision: A BIM–point cloud–DFMA collaborative framework for steel truss pre-assembly analysis

Abstract

Virtual trial assembly can reduce the cost and time required for assembly verification of large and complex steel structures. However, existing methods cannot simultaneously detect component accuracy, unit geometric states, and interface deviations, nor do they account for the propagation and accumulation of errors across multiple units. To address this, this paper proposes a virtual pre-assembly analysis method for steel trusses that integrates point clouds, Building Information Modelling (BIM), and Design for Manufacture and Assembly (DFMA). The method first employs Light Detection and Ranging (LiDAR) to acquire high-density point cloud data of steel members and then achieves high-precision registration between the point clouds and the BIM model through feature point matching, normal correction, and Iterative Closest Point (ICP) fine registration. On this basis, a deviation detection system covering the component, assembly unit, and key interface levels is established, enabling the identification of bending and torsional deviations of steel members, the analysis of overall geometric deviations of the lifting unit, and the assessment of directional deviations of critical assembly faces. Furthermore, a Pre-assembly Analysis System (PAS) is developed. Based on the deviation characteristics of the top, bottom, left, and right assembly faces, the PAS quantitatively determines the assembly feasibility of key interfaces. Finally, a case study of a large-span cantilever steel truss verifies that the proposed method can effectively identify local manufacturing defects, non-uniform deformation zones, and high-risk interface locations during the pre-assembly process. This research provides technical support for assembly feasibility evaluation, installation scheme optimisation, and construction risk pre-control prior to the lifting of large-span cantilever steel structures.

The authors' abstract, as published at the source. Engineering Structures, 2026 · DOI ↗