Abstract

The need to create and maintain up to date digital copies of infrastructure assets has been well established in literature. These copies are often labelled Digital Twins. Their most basic form includes only their geometry (Geometric Digital Twin-GDT). Yet, very few assets today have a usable digital twin. This occurs because the perceived cost of the creating and maintaining the digital twin greatly counteracts the perceived benefits of the twin. This happens in part because of the labour cost needed to manually build/maintain the digital model. This cost is substantive because state-of-the-art digitisation technologies necessary for automated delivery of the GDT require substantial labour hours to build even a simple, geometry only model. We will tackle this challenge by proposing a digitisation framework, focused on delivering an automated way to create AI-GDTs of rail infrastructure. It starts with registered PCDs and ends with GDTs of railways. The proposed framework combines the strength of the ‘data-driven’ strategy scenarios with very high point densities and ‘model-based’ strategy in scenarios with very low point densities.