Creating digital twins of existing bridges through AI-based methods

Abstract

Bridges require regular inspection and maintenance during their service life, which is costly and time-consuming. Digital twins (DT), which incorporate a geometric-semantic model of an existing bridge, can support the operation and maintenance process. The process of creating such DT models can be based on Point cloud data (PCD), created via photogrammetry or laser scanning. However, the semantic segmentation of PCD and parametric modeling is a challenging process, which is nonetheless necessary to support DT modeling. This paper aims to propose a segmentation method that is the basis for a parametric modeling approach to enable the semi-automatic geometric modeling of bridges from PCD. To this end, metaheuristic algorithms, fuzzy C-mean clustering, and signal processing algorithms are used. The results of this paper show that the scan to BIM process of bridges can be automated to a large extent and provide a model that meets the industry’s demand.