Integrating digital twin and large-language model in building life cycle assessment

Traditional Life Cycle Assessment (LCA) in the building sector faces several persistent challenges. It typically relies on manual data collection, static datasets, and time-intensive workflows that require expert knowledge and specialised software. The lack of standardised data formats and interoperability between tools often leads to fragmented LCAs and inconsistent results. Furthermore, traditional LCA is commonly performed at discrete stages of a project, offering limited ability to respond to design changes or provide timely feedback. As a result, LCA is often underutilised during early design phases when the potential for environmental impact reduction is greatest.;;Several recent tools for building LCA have adopted digital twin (DT) interfaces (e.g. AutoBIM Carbon Calculator developed by Balfour Beatty, the AI-based Embodied Carbon Calculator by Costain-Winvic). These tools aim to achieve real-time visualisation, prediction, and automatic feedback to reveal trade-offs. Building Life Cycle Assessment (LCA) is an essential component of many building rating tools and systems, globally. However, the wide range of assumptions devolved along the heterogeneous data points in LCA processes often leads to inconsistent results. As a result, many building stakeholders and building rating tools have depreciated the role of LCA in achieving improved building design. DT allows for the opportunity to collect data in a standard and consistent way and holds promising potential in harmonising the practice of LCA in the building sector. The primary reason for using DT is to actualise a high-fidelity virtual dynamic and interoperable representation of LCA. DTs have the potential to be enriched over time where the monitoring and sensing of physical assets is a key ability. In LCA, this potential can be realised by LCA coefficients. This is usually facilitated by traditional sensors, but there are many other examples of technologies used to monitor and gather data on-site (e.g. 3D scanners, satellite images, etc.). In the case of building DT, large-language models (LLMs) can be used to gather contextual LCA data from external databases and published literature. The aim of this study is to develop a framework that integrates digital twin and large-language models in building life cycle assessment.;;By integrating DT with LLMs, these challenges can be significantly mitigated. When combined with LLMs, this system gains the ability to interpret technical documents, map BIM elements to LCA databases, and extract relevant environmental data from external sources automatically. This integration allows for real-time LCA updates, scenario analysis, and natural language interaction, making the LCA process more responsive, accurate, and accessible to non-experts. Integration of DT and LLMs is expected to transform LCA into an intelligent, adaptive tool that supports informed and sustainable decision-making across all stages of a building’s life.;;The objectives of this study are as follows:;1. To build a Digital Twin (DT) model for selected building projects, enabling real-time representation of LCA data.;2. To use LLMs to collect and synthesize LCA data from available sources;3. To develop a framework integrating Digital Twin (DT) and LLMs to automate the LCA in buildings.;4. To create a Revit API tool to assign environmental impacts to Digital Twin elements in real-time.;