Artificial Intelligence Implementation Roadmap for Automated Heritage Building Information Modelling in UK`s Cultural Heritage Assets

Abstract

Heritage Building Information Modelling (HBIM) plays a critical role in the documentation and management of historic buildings. However, current Scan-to-BIM workflows remain labour-intensive, inconsistent, and insufficiently aligned with classification systems such as NBS Uniclass. These limitations hinder the widespread adoption of HBIM in the UK, particularly in automating the conversion of unstructured point cloud data into semantically rich and structured digital models.
This research develops an AI-enabled implementation roadmap to automate HBIM modelling process, leveraging machine learning and deep learning techniques for point cloud segmentation, classification, and object generation. The study adopts a Design Science Research methodology, combining a systematic literature review, semantic gap analysis, and empirical experimentation. Case study data were drawn from the Maritime Greenwich World Heritage Site to ensure contextual relevance and realworld applicability.
The results demonstrate that Random Forest and Point Transformer algorithms can achieve high levels of classification accuracy when trained on heritage-informed datasets, with Point Transformer model benefiting from bespoke prompt tuning and multi-scale processing. These approaches significantly improved the semantic granularity and consistency of classification outputs across varying building typologies. In addition, procedural geometry generation techniques proved effective in producing HBIM-ready components from segmented data, supporting greater automation and reduced manual effort without compromising data quality or historical fidelity.
This study proposes a structured and practical AI implementation roadmap, offering five adoption scenarios, Revival, Restoration, Retrofit, Restitution and Resilience, tailored to the diverse needs of heritage stakeholders. It delivers both theoretical and practical contributions, advancing the academic discourse on HBIM automation while providing heritage professionals and asset managers with a scalable, interoperable roadmap and execution plan template for AI enabled digital transformation in the UK’s historic built environment.

Date of Award	30 Sept 2025
Original language	English
Awarding Institution	Northumbria University
Supervisor	Yusuf Arayici (Supervisor) & Bimal Kumar (Supervisor)