Enabling a fully integrated BIM environment using Semantics and Artificial Intelligence

Abstract

This thesis addresses the challenge of data fragmentation in the Architecture, Engineering, Construction, and Operations industry by proposing an integrated framework that combines semantic web technologies and artificial intelligence. Despite the impact of Building Information Modelling in transforming traditional design and construction workflows, the full potential of Building Information Modelling remains limited by the isolation of data across multiple platforms and disciplines. Traditional Computer-Aided Design and early Building Information Modelling systems often rely on visual integration, which lacks the robust semantic connections needed for automated data exchange and cross-disciplinary validation.

The research is organised into three interrelated branches. The first branch focuses on the federation of Building Information Modelling models by converting disparate domain-specific representations into a unified semantic framework. By transforming files based on Industry Foundation Classes into ontological formats, the study identifies and links duplicated elements, such as spatial zones and building storeys, across different models. This process establishes explicit semantic relationships that overcome the limitations of mere visual coordination, enabling a more integrated and machine-readable data environment. The creation of these semantic links not only enhances interoperability but also lays the groundwork for automated reasoning and efficient information retrieval across disciplines.

The second branch introduces novel artificial intelligence methods for the automated validation of Building Information Modelling content. Recognising that manual verification is both error-prone and resource-intensive, the thesis employs deep learning approaches, including graph neural networks and context-aware neural networks, to systematically identify and correct inconsistencies in element classification and property definitions. These artificial intelligence-driven models are trained on publicly available open-source Building Information Modelling datasets, which serve as proxies for real-world applications. The experimental results demonstrate significant improvements in the accuracy and reliability of automated data validation processes. By reducing dependency on manual checks, the proposed methods substantially mitigate risks associated with misclassification and erroneous data entries, thereby enhancing overall model integrity.

The third branch extends the framework by integrating external data sources for cross-validation. Crucial information that is often relevant to a building project, such as schematic drawings, design documents, or ancillary databases, is not embedded within the Building Information Modelling model itself. To address this gap, a conceptual framework is developed to extract, structure, and compare external information with the Building Information Modelling content. This cross-validation process ensures that discrepancies between the digital model and real-world documentation are detected early, allowing for timely corrections that improve project outcomes. Such a systematic approach not only fortifies the reliability of Building Information Modelling data but also supports advanced applications such as digital twin implementations and intelligent asset management.

Overall, the contributions of this thesis offer a robust pathway toward mitigating data fragmentation in the Architecture, Engineering, Construction, and Operations industry. By establishing a fully integrated Building Information Modelling environment through semantic enrichment and artificial intelligence-based validation, the proposed framework enhances the fidelity of digital models and streamlines interdisciplinary collaboration. The research results present that automated semantic integration and data validation can significantly increase operational efficiency and data quality in construction projects. Future research will focus on further improvement of proposed artificial intelligence methodologies and expanding the framework to adapt to evolving industry standards and new technological innovations, ultimately providing the way for smarter, more firm construction environments.